diff --git a/Cargo.lock b/Cargo.lock
index e5bcf3376..a2f4c89f2 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -238,9 +238,9 @@ checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
 
 [[package]]
 name = "bitflags"
-version = "2.12.1"
+version = "2.13.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "84d7ced0ae9557296835c32bf1b1e02b44c746701f898460fb000d7eaa84f00a"
+checksum = "b4388bee8683e3d04af747c73422af53102d2bd24d9eadb6cbc100baef4b43f8"
 
 [[package]]
 name = "bitvec"
@@ -872,7 +872,7 @@ version = "0.3.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1e0e367e4e7da84520dedcac1901e4da967309406d1e51017ae1abfb97adbd38"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "objc2",
 ]
 
@@ -1830,7 +1830,7 @@ version = "3.9.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f1d395473824516f38dd1071a1a37bc57daa7be65b293ebba4ead5f7abb017a2"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "ctor",
  "futures",
  "napi-build",
@@ -1894,7 +1894,7 @@ version = "0.28.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ab2156c4fce2f8df6c499cc1c763e4394b7482525bf2a9701c9d79d215f519e4"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "cfg-if",
  "cfg_aliases 0.1.1",
  "libc",
@@ -1906,7 +1906,7 @@ version = "0.31.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cf20d2fde8ff38632c426f1165ed7436270b44f199fc55284c38276f9db47c3d"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "cfg-if",
  "cfg_aliases 0.2.1",
  "libc",
@@ -1997,7 +1997,7 @@ version = "0.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d49e936b501e5c5bf01fda3a9452ff86dc3ea98ad5f283e1455153142d97518c"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "objc2",
  "objc2-core-graphics",
  "objc2-foundation",
@@ -2009,7 +2009,7 @@ version = "0.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2a180dd8642fa45cdb7dd721cd4c11b1cadd4929ce112ebd8b9f5803cc79d536"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "dispatch2",
  "objc2",
 ]
@@ -2020,7 +2020,7 @@ version = "0.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e022c9d066895efa1345f8e33e584b9f958da2fd4cd116792e15e07e4720a807"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "dispatch2",
  "objc2",
  "objc2-core-foundation",
@@ -2039,7 +2039,7 @@ version = "0.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e3e0adef53c21f888deb4fa59fc59f7eb17404926ee8a6f59f5df0fd7f9f3272"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "objc2",
  "objc2-core-foundation",
 ]
@@ -2050,7 +2050,7 @@ version = "0.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "180788110936d59bab6bd83b6060ffdfffb3b922ba1396b312ae795e1de9d81d"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "objc2",
  "objc2-core-foundation",
 ]
@@ -2331,7 +2331,7 @@ dependencies = [
 
 [[package]]
 name = "pi-ast"
-version = "15.9.4"
+version = "15.10.1"
 dependencies = [
  "anyhow",
  "ast-grep-core",
@@ -2399,7 +2399,7 @@ dependencies = [
 
 [[package]]
 name = "pi-iso"
-version = "15.9.4"
+version = "15.10.1"
 dependencies = [
  "async-trait",
  "libc",
@@ -2411,7 +2411,7 @@ dependencies = [
 
 [[package]]
 name = "pi-natives"
-version = "15.9.4"
+version = "15.10.1"
 dependencies = [
  "anyhow",
  "arboard",
@@ -2457,7 +2457,7 @@ dependencies = [
 
 [[package]]
 name = "pi-shell"
-version = "15.9.4"
+version = "15.10.1"
 dependencies = [
  "anyhow",
  "brush-builtins",
@@ -2497,7 +2497,7 @@ version = "0.18.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "60769b8b31b2a9f263dae2776c37b1b28ae246943cf719eb6946a1db05128a61"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "crc32fast",
  "fdeflate",
  "flate2",
@@ -2567,7 +2567,7 @@ version = "0.18.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "25485360a54d6861439d60facef26de713b1e126bf015ec8f98239467a2b82f7"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "chrono",
  "flate2",
  "procfs-core",
@@ -2580,7 +2580,7 @@ version = "0.18.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e6401bf7b6af22f78b563665d15a22e9aef27775b79b149a66ca022468a4e405"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "chrono",
  "hex",
 ]
@@ -2735,7 +2735,7 @@ version = "0.5.18"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ed2bf2547551a7053d6fdfafda3f938979645c44812fbfcda098faae3f1a362d"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
 ]
 
 [[package]]
@@ -2833,7 +2833,7 @@ version = "1.1.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b6fe4565b9518b83ef4f91bb47ce29620ca828bd32cb7e408f0062e9930ba190"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "errno",
  "libc",
  "linux-raw-sys",
@@ -4279,7 +4279,7 @@ version = "0.244.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "47b807c72e1bac69382b3a6fb3dbe8ea4c0ed87ff5629b8685ae6b9a611028fe"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "hashbrown 0.15.5",
  "indexmap",
  "semver",
@@ -4304,7 +4304,7 @@ version = "0.31.14"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "645c7c96bb74690c3189b5c9cb4ca1627062bb23693a4fad9d8c3de958260144"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "rustix",
  "wayland-backend",
  "wayland-scanner",
@@ -4316,7 +4316,7 @@ version = "0.32.12"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "563a85523cade2429938e790815fd7319062103b9f4a2dc806e9b53b95982d8f"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "wayland-backend",
  "wayland-client",
  "wayland-scanner",
@@ -4328,7 +4328,7 @@ version = "0.3.12"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "eb04e52f7836d7c7976c78ca0250d61e33873c34156a2a1fc9474828ec268234"
 dependencies = [
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "wayland-backend",
  "wayland-client",
  "wayland-protocols",
@@ -4836,7 +4836,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9d66ea20e9553b30172b5e831994e35fbde2d165325bec84fc43dbf6f4eb9cb2"
 dependencies = [
  "anyhow",
- "bitflags 2.12.1",
+ "bitflags 2.13.0",
  "indexmap",
  "log",
  "serde",
diff --git a/Cargo.toml b/Cargo.toml
index 49e6ebf38..f6301942c 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -4,7 +4,7 @@ exclude = ["crates/brush-core-vendored", "crates/brush-builtins-vendored"]
 resolver = "3"
 
 [workspace.package]
-version = "15.9.4"
+version = "15.10.1"
 edition = "2024"
 license = "MIT"
 authors = ["Can Boluk"]
diff --git a/bun.lock b/bun.lock
index d2646c93e..3cff53436 100644
--- a/bun.lock
+++ b/bun.lock
@@ -15,7 +15,7 @@
     },
     "packages/agent": {
       "name": "@oh-my-pi/pi-agent-core",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "dependencies": {
         "@oh-my-pi/pi-ai": "catalog:",
         "@oh-my-pi/pi-natives": "catalog:",
@@ -30,7 +30,7 @@
     },
     "packages/ai": {
       "name": "@oh-my-pi/pi-ai",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "dependencies": {
         "@bufbuild/protobuf": "catalog:",
         "@oh-my-pi/pi-utils": "catalog:",
@@ -44,7 +44,7 @@
     },
     "packages/coding-agent": {
       "name": "@oh-my-pi/pi-coding-agent",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "bin": {
         "omp": "src/cli.ts",
       },
@@ -90,7 +90,7 @@
     },
     "packages/hashline": {
       "name": "@oh-my-pi/hashline",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "dependencies": {
         "diff": "catalog:",
         "lru-cache": "catalog:",
@@ -101,7 +101,7 @@
     },
     "packages/mnemopi": {
       "name": "@oh-my-pi/pi-mnemopi",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "bin": {
         "mnemopi": "src/cli.ts",
       },
@@ -118,7 +118,7 @@
     },
     "packages/natives": {
       "name": "@oh-my-pi/pi-natives",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "devDependencies": {
         "@napi-rs/cli": "catalog:",
         "@types/bun": "catalog:",
@@ -126,7 +126,7 @@
     },
     "packages/stats": {
       "name": "@oh-my-pi/omp-stats",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "bin": {
         "omp-stats": "./src/index.ts",
       },
@@ -151,7 +151,7 @@
     },
     "packages/swarm-extension": {
       "name": "@oh-my-pi/swarm-extension",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "bin": {
         "omp-swarm": "src/cli.ts",
       },
@@ -167,7 +167,7 @@
     },
     "packages/tui": {
       "name": "@oh-my-pi/pi-tui",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "dependencies": {
         "@oh-my-pi/pi-natives": "catalog:",
         "@oh-my-pi/pi-utils": "catalog:",
@@ -208,7 +208,7 @@
     },
     "packages/utils": {
       "name": "@oh-my-pi/pi-utils",
-      "version": "15.9.4",
+      "version": "15.10.1",
       "dependencies": {
         "@oh-my-pi/pi-natives": "catalog:",
         "beautiful-mermaid": "catalog:",
@@ -248,15 +248,15 @@
     "@huggingface/transformers": "^4.2.0",
     "@mozilla/readability": "^0.6.0",
     "@napi-rs/cli": "3.7.0",
-    "@oh-my-pi/hashline": "15.9.4",
-    "@oh-my-pi/omp-stats": "15.9.4",
-    "@oh-my-pi/pi-agent-core": "15.9.4",
-    "@oh-my-pi/pi-ai": "15.9.4",
-    "@oh-my-pi/pi-coding-agent": "15.9.4",
-    "@oh-my-pi/pi-mnemopi": "15.9.4",
-    "@oh-my-pi/pi-natives": "15.9.4",
-    "@oh-my-pi/pi-tui": "15.9.4",
-    "@oh-my-pi/pi-utils": "15.9.4",
+    "@oh-my-pi/hashline": "15.10.1",
+    "@oh-my-pi/omp-stats": "15.10.1",
+    "@oh-my-pi/pi-agent-core": "15.10.1",
+    "@oh-my-pi/pi-ai": "15.10.1",
+    "@oh-my-pi/pi-coding-agent": "15.10.1",
+    "@oh-my-pi/pi-mnemopi": "15.10.1",
+    "@oh-my-pi/pi-natives": "15.10.1",
+    "@oh-my-pi/pi-tui": "15.10.1",
+    "@oh-my-pi/pi-utils": "15.10.1",
     "@opentelemetry/api": "^1.9.1",
     "@opentelemetry/context-async-hooks": "^2.7.1",
     "@opentelemetry/exporter-trace-otlp-proto": "^0.218.0",
@@ -395,7 +395,7 @@
 
     "@huggingface/jinja": ["@huggingface/jinja@0.5.9", "", {}, "sha512-uWTG+l3VJRsl7EXxYizuL3P+cCPoc3cRqbWWRcQN0FhejRfbdq0RNhCmbY/YDtnTcz9icdLYuLDjsnz4d8JMuw=="],
 
-    "@huggingface/tasks": ["@huggingface/tasks@0.21.2", "", {}, "sha512-e8dw3tZ7mbZ/mytr9zIFmsr67tMpd9rm/pURCi5ciFqlVvLvdH9FjnoZxVO4KfRXyqzPeV0shXv9z5s2M8Msmw=="],
+    "@huggingface/tasks": ["@huggingface/tasks@0.21.7", "", {}, "sha512-GuEXszIkir4j/Oywp4hXP+wfwojo/SKWA/omroNkzWWgqUGiOQ5p6HuyXcDOcinYnLQW1WsO8fwdEvtLTZbA4w=="],
 
     "@huggingface/tokenizers": ["@huggingface/tokenizers@0.1.3", "", {}, "sha512-8rF/RRT10u+kn7YuUbUg0OF30K8rjTc78aHpxT+qJ1uWSqxT1MHi8+9ltwYfkFYJzT/oS+qw3JVfHtNMGAdqyA=="],
 
@@ -1159,7 +1159,7 @@
 
     "onnxruntime-web": ["onnxruntime-web@1.26.0-dev.20260416-b7804b056c", "", { "dependencies": { "flatbuffers": "^25.1.24", "guid-typescript": "^1.0.9", "long": "^5.2.3", "onnxruntime-common": "1.24.0-dev.20251116-b39e144322", "platform": "^1.3.6", "protobufjs": "^7.2.4" } }, "sha512-MD6Ss4GSpQBo6zqoJzyT9LRbKYs7x/JVN23FT24EcEvlqF4VuzPOeH6X38orZPKHQDbprn7K+SBpu0/mj2CQiw=="],
 
-    "openai": ["openai@6.39.1", "", { "peerDependencies": { "ws": "^8.18.0", "zod": "^3.25 || ^4.0" }, "optionalPeers": ["ws", "zod"], "bin": { "openai": "bin/cli" } }, "sha512-z3dO9fEWOXBzlXynVb/xZ/tujzUjFWQWn3C0n0mw6Vo0zJTbEkaN4b2cLWjhJ6haJQx8LlREoafHRl+Gu/Hl+A=="],
+    "openai": ["openai@6.42.0", "", { "peerDependencies": { "ws": "^8.18.0", "zod": "^3.25 || ^4.0" }, "optionalPeers": ["ws", "zod"] }, "sha512-1WFEt/uXMXOLhYRNkgJWo08Y2YNvNwpVU72K7ibrWgWpNOXd4VojXLbe6SQ4bLiUQ3Y8jz4IiyVkylJCL1DtZg=="],
 
     "option": ["option@0.2.4", "", {}, "sha512-pkEqbDyl8ou5cpq+VsnQbe/WlEy5qS7xPzMS1U55OCG9KPvwFD46zDbxQIj3egJSFc3D+XhYOPUzz49zQAVy7A=="],
 
@@ -1259,7 +1259,7 @@
 
     "string-width": ["string-width@4.2.3", "", { "dependencies": { "emoji-regex": "^8.0.0", "is-fullwidth-code-point": "^3.0.0", "strip-ansi": "^6.0.1" } }, "sha512-wKyQRQpjJ0sIp62ErSZdGsjMJWsap5oRNihHhu6G7JVO/9jIB6UyevL+tXuOqrng8j/cxKTWyWUwvSTriiZz/g=="],
 
-    "string_decoder": ["string_decoder@1.3.0", "", { "dependencies": { "safe-buffer": "~5.2.0" } }, "sha512-hkRX8U1WjJFd8LsDJ2yQ/wWWxaopEsABU1XfkM8A+j0+85JAGppt16cr1Whg6KIbb4okU6Mql6BOj+uup/wKeA=="],
+    "string_decoder": ["string_decoder@1.1.1", "", { "dependencies": { "safe-buffer": "~5.1.0" } }, "sha512-n/ShnvDi6FHbbVfviro+WojiFzv+s8MPMHBczVePfUpDJLwoLT0ht1l4YwBCbi8pJAveEEdnkHyPyTP/mzRfwg=="],
 
     "strip-ansi": ["strip-ansi@7.2.0", "", { "dependencies": { "ansi-regex": "^6.2.2" } }, "sha512-yDPMNjp4WyfYBkHnjIRLfca1i6KMyGCtsVgoKe/z1+6vukgaENdgGBZt+ZmKPc4gavvEZ5OgHfHdrazhgNyG7w=="],
 
@@ -1413,8 +1413,6 @@
 
     "string-width/strip-ansi": ["strip-ansi@6.0.1", "", { "dependencies": { "ansi-regex": "^5.0.1" } }, "sha512-Y38VPSHcqkFrCpFnQ9vuSXmquuv5oXOKpGeT6aGrr3o3Gc9AlVa6JBfUSOCnbxGGZF+/0ooI7KrPuUSztUdU5A=="],
 
-    "string_decoder/safe-buffer": ["safe-buffer@5.2.1", "", {}, "sha512-rp3So07KcdmmKbGvgaNxQSJr7bGVSVk5S9Eq1F+ppbRo70+YeaDxkw5Dd8NPN+GD6bjnYm2VuPuCXmpuYvmCXQ=="],
-
     "wrap-ansi/string-width": ["string-width@8.2.1", "", { "dependencies": { "get-east-asian-width": "^1.5.0", "strip-ansi": "^7.1.2" } }, "sha512-IIaP0g3iy9Cyy18w3M9YcaDudujEAVHKt3a3QJg1+sr/oX96TbaGUubG0hJyCjCBThFH+tFpcIyoUHUn1ogaLA=="],
 
     "xml2js/xmlbuilder": ["xmlbuilder@11.0.1", "", {}, "sha512-fDlsI/kFEx7gLvbecc0/ohLG50fugQp8ryHzMTuW9vSa1GJ0XYWKnhsUx7oie3G98+r56aTQIUB4kht42R3JvA=="],
@@ -1429,8 +1427,6 @@
 
     "fastembed/onnxruntime-node/tar": ["tar@7.5.16", "", { "dependencies": { "@isaacs/fs-minipass": "^4.0.0", "chownr": "^3.0.0", "minipass": "^7.1.2", "minizlib": "^3.1.0", "yallist": "^5.0.0" } }, "sha512-56adEpPMouktRlBLXiaYFFzZ/3+JXa8P9n7WbR+ibIjtviN55mEaOkiysCnPnWm+7kkui1Dn8J9l+g6zV8731w=="],
 
-    "jszip/readable-stream/string_decoder": ["string_decoder@1.1.1", "", { "dependencies": { "safe-buffer": "~5.1.0" } }, "sha512-n/ShnvDi6FHbbVfviro+WojiFzv+s8MPMHBczVePfUpDJLwoLT0ht1l4YwBCbi8pJAveEEdnkHyPyTP/mzRfwg=="],
-
     "log-update/slice-ansi/is-fullwidth-code-point": ["is-fullwidth-code-point@5.1.0", "", { "dependencies": { "get-east-asian-width": "^1.3.1" } }, "sha512-5XHYaSyiqADb4RnZ1Bdad6cPp8Toise4TzEjcOYDHZkTCbKgiUl7WTUCpNWHuxmDt91wnsZBc9xinNzopv3JMQ=="],
 
     "log-update/wrap-ansi/string-width": ["string-width@7.2.0", "", { "dependencies": { "emoji-regex": "^10.3.0", "get-east-asian-width": "^1.0.0", "strip-ansi": "^7.1.0" } }, "sha512-tsaTIkKW9b4N+AEj+SVA+WhJzV7/zMhcSu78mLKWSk7cXMOSHsBKFWUs0fWwq8QyK3MgJBQRX6Gbi4kYbdvGkQ=="],
diff --git a/crates/pi-natives/src/lib.rs b/crates/pi-natives/src/lib.rs
index ae74dbadd..6b9c17c74 100644
--- a/crates/pi-natives/src/lib.rs
+++ b/crates/pi-natives/src/lib.rs
@@ -68,5 +68,5 @@ use napi_derive::napi;
 /// MUST stay in sync with `VERSION_SENTINEL_EXPORT` in
 /// `packages/natives/native/index.js` (which derives the name from
 /// `package.json#version`).
-#[napi(js_name = "__piNativesV15_9_4")]
+#[napi(js_name = "__piNativesV15_10_1")]
 pub const fn pi_natives_version_sentinel() {}
diff --git a/crates/pi-shell/src/fixup.rs b/crates/pi-shell/src/fixup.rs
index 86b6974d4..f9fe9d70a 100644
--- a/crates/pi-shell/src/fixup.rs
+++ b/crates/pi-shell/src/fixup.rs
@@ -154,7 +154,12 @@ fn try_strip_head_tail(
 	// span under-reports when its suffix contains unlocated `IoRedirect`s
 	// (e.g. the synthetic `2>&1` inserted by `|&`).
 	let bytes = cmd.as_bytes();
-	let last_start = last_loc.start.index;
+	let Some(last_start) = byte_offset(cmd, last_loc.start.index) else {
+		return default;
+	};
+	let Some(last_end) = byte_offset(cmd, last_loc.end.index) else {
+		return default;
+	};
 	let Some(head) = cmd.get(..last_start) else {
 		return default;
 	};
@@ -172,7 +177,7 @@ fn try_strip_head_tail(
 	// Reported text starts at the pipe and is right-trimmed. The deletion
 	// range walks back through any leading whitespace so the rewrite is
 	// contiguous.
-	let stripped_text = cmd[pipe_pos..last_loc.end.index].trim_end().to_owned();
+	let stripped_text = cmd[pipe_pos..last_end].trim_end().to_owned();
 	if stripped_text.is_empty() {
 		return default;
 	}
@@ -180,7 +185,7 @@ fn try_strip_head_tail(
 	while delete_start > 0 && matches!(bytes[delete_start - 1], b' ' | b'\t') {
 		delete_start -= 1;
 	}
-	ranges.push((delete_start, last_loc.end.index));
+	ranges.push((delete_start, last_end));
 	stripped.push(stripped_text);
 	HeadTailOutcome { stripped: true, last_idx: n - 2 }
 }
@@ -253,10 +258,14 @@ fn try_strip_2to1(
 	let Some(name_loc) = name_word.loc.as_ref() else {
 		return;
 	};
-	let mut anchor = name_loc.end.index;
+	let Some(mut anchor) = byte_offset(cmd, name_loc.end.index) else {
+		return;
+	};
 	for item in &suffix.0 {
-		if let Some(loc) = item.location() {
-			anchor = anchor.max(loc.end.index);
+		if let Some(loc) = item.location()
+			&& let Some(end) = byte_offset(cmd, loc.end.index)
+		{
+			anchor = anchor.max(end);
 		}
 	}
 	let bytes = cmd.as_bytes();
@@ -283,6 +292,28 @@ fn try_strip_2to1(
 	stripped.push("2>&1".to_owned());
 }
 
+/// Translate a `brush-parser` source-position index into a byte offset in
+/// `cmd`. The parser counts positions in Unicode scalars (one increment per
+/// `char`; see `tokenizer::next_char`), but we slice `cmd` — a `&str` — by
+/// byte index, so the two diverge as soon as the command contains any
+/// multi-byte UTF-8 (e.g. a `✓`/`×` literal inside a `grep` pattern). Without
+/// this conversion the head/tail and `2>&1` strips cut at the wrong place,
+/// corrupting the command (notably orphaning a closing quote).
+///
+/// Returns `None` only when `char_idx` is past the end of the input. The
+/// end-of-input position (`char_idx == cmd.chars().count()`) maps to
+/// `cmd.len()`.
+fn byte_offset(cmd: &str, char_idx: usize) -> Option<usize> {
+	let mut count = 0usize;
+	for (byte, _) in cmd.char_indices() {
+		if count == char_idx {
+			return Some(byte);
+		}
+		count += 1;
+	}
+	(count == char_idx).then_some(cmd.len())
+}
+
 fn is_stderr_to_stdout(io: &IoRedirect) -> bool {
 	let IoRedirect::File(Some(2), IoFileRedirectKind::DuplicateOutput, target) = io else {
 		return false;
@@ -380,6 +411,45 @@ mod tests {
 		}
 	}
 
+	#[test]
+	fn strips_with_multibyte_content() {
+		// brush-parser reports char-indexed positions; multi-byte UTF-8 before
+		// the trailing `| tail` must not shift the byte-level cut. Regression
+		// for a corrupted command that orphaned the grep pattern's closing
+		// quote (`… |✓|×-80`) and broke later re-parsing.
+		let cases: &[(&str, &str, &[&str])] = &[
+			// Mirrors the real bug: `2>&1` sits mid-pipeline (grep becomes the
+			// effective tail) so only `| tail -80` is stripped, leaving the
+			// quoted grep pattern intact.
+			(
+				"xcodebuild 2>&1 | grep -E \"a|✓|×|b\" | tail -80",
+				"xcodebuild 2>&1 | grep -E \"a|✓|×|b\"",
+				&["| tail -80"],
+			),
+			("echo ✓ | head -3", "echo ✓", &["| head -3"]),
+			("printf '日本語' | tail -n 5", "printf '日本語'", &["| tail -n 5"]),
+		];
+		for (input, want_cmd, want_stripped) in cases {
+			let (cmd, stripped) = run(input);
+			assert_eq!(cmd, *want_cmd, "input: {input:?}");
+			assert_eq!(stripped, *want_stripped, "input: {input:?}");
+			// The rewrite must remain valid shell (no orphaned quote).
+			let options = ParserOptions::default();
+			let source_info = SourceInfo::default();
+			let mut reader = BufReader::new(cmd.as_bytes());
+			let mut parser = Parser::new(&mut reader, &options, &source_info);
+			assert!(parser.parse_program().is_ok(), "rewrite not re-parseable: {cmd:?}");
+		}
+	}
+
+	#[test]
+	fn strips_2to1_after_multibyte() {
+		// Multi-byte before a trailing `2>&1` must not misplace the strip.
+		let (cmd, stripped) = run("echo ✓ × 2>&1");
+		assert_eq!(cmd, "echo ✓ ×");
+		assert_eq!(stripped, vec!["2>&1"]);
+	}
+
 	#[test]
 	fn strips_redundant_2to1() {
 		let cases: &[(&str, &str, &[&str])] = &[
diff --git a/docs/extensions.md b/docs/extensions.md
index 119d0f2cb..faedb4c81 100644
--- a/docs/extensions.md
+++ b/docs/extensions.md
@@ -382,7 +382,7 @@ Provide `renderCall` / `renderResult` on `registerTool` definitions for custom t
 - Runtime actions are unavailable during extension load.
 - `tool_call` errors block execution (fail-closed).
 - Command name conflicts with built-ins are skipped with diagnostics.
-- Reserved shortcuts are ignored (`ctrl+c`, `ctrl+d`, `ctrl+z`, `ctrl+k`, `ctrl+p`, `ctrl+l`, `ctrl+o`, `ctrl+t`, `ctrl+g`, `shift+tab`, `shift+ctrl+p`, `alt+enter`, `escape`, `enter`).
+- Reserved shortcuts are ignored (`ctrl+c`, `ctrl+d`, `ctrl+z`, `ctrl+k`, `ctrl+p`, `ctrl+l`, `ctrl+o`, `ctrl+t`, `ctrl+g`, `ctrl+q`, `alt+m`, `shift+tab`, `shift+ctrl+p`, `alt+enter`, `escape`, `enter`).
 - Treat `ctx.reload()` as terminal for the current command handler frame.
 
 ## Extensions vs hooks vs custom-tools
diff --git a/docs/keybindings.md b/docs/keybindings.md
index 8e8158bbf..dfc881bbe 100644
--- a/docs/keybindings.md
+++ b/docs/keybindings.md
@@ -27,20 +27,23 @@ app.stt.toggle: []
 | `app.model.cycleForward`    | `Ctrl+P`                               | Cycle role models forward                     |
 | `app.model.cycleBackward`   | `Shift+Ctrl+P`                         | Cycle role models in temporary mode           |
 | `app.model.selectTemporary` | `Alt+P`                                | Pick a model temporarily for this session     |
-| `app.model.select`          | `Ctrl+L`                               | Open the model selector and set roles         |
+| `app.model.select`          | `Alt+M`                                | Open the model selector and set roles         |
 | `app.plan.toggle`           | `Alt+Shift+P`                          | Toggle plan mode                              |
 | `app.history.search`        | `Ctrl+R`                               | Search prompt history                         |
 | `app.tools.expand`          | `Ctrl+O`                               | Toggle tool-output expansion                  |
 | `app.thinking.toggle`       | `Ctrl+T`                               | Toggle thinking-block visibility              |
 | `app.thinking.cycle`        | `Shift+Tab`                            | Cycle thinking level                          |
 | `app.editor.external`       | `Ctrl+G`                               | Edit the draft in `$VISUAL` / `$EDITOR`       |
-| `app.message.followUp`      | `Ctrl+Enter`                           | Queue a follow-up message                     |
+| `app.message.followUp`      | `Ctrl+Q`, `Ctrl+Enter`                 | Queue a follow-up message                     |
 | `app.message.dequeue`       | `Alt+Up`                               | Dequeue a queued message back into the editor |
+| `app.display.reset`         | `Ctrl+L`                               | Reset terminal display                        |
 | `app.clipboard.copyLine`    | `Alt+Shift+L`                          | Copy the current line                         |
 | `app.clipboard.copyPrompt`  | `Alt+Shift+C`                          | Copy the whole prompt                         |
 | `app.clipboard.pasteImage`  | `Ctrl+V` (`Alt+V` fallback on Windows) | Paste an image from the clipboard             |
 | `app.stt.toggle`            | `Alt+H`                                | Toggle speech-to-text recording               |
 
-On Windows Terminal, `Ctrl+V` may be handled by the terminal paste command before `omp` sees it; use the `Alt+V` fallback when clipboard image paste appears to do nothing.
+On Windows Terminal, `Ctrl+V` may be handled by the terminal paste command before `omp` sees it; use the `Alt+V` fallback when clipboard image paste appears to do nothing. Windows Terminal also swallows `Ctrl+Enter`, so the follow-up shortcut also binds `Ctrl+Q` — the same chord GitHub Copilot CLI uses. If your existing `keybindings.yml` already assigns `Ctrl+Q` to another action, that user remap wins and follow-up keeps `Ctrl+Enter` unless you explicitly bind `app.message.followUp`.
+
+Terminals that implement OSC 5522 enhanced paste can send clipboard MIME data directly to `omp`; image pastes are attached as `[Image #N]`, while text/plain paste events keep normal paste behavior. When OSC 5522 is unavailable, bracketed paste still handles text, and a pasted single image-file path is loaded as an image when the file is readable from the `omp` host.
 
 Older unqualified action names are migrated when `keybindings.yml` is loaded, but new docs and new configs should use the namespaced action IDs above. Existing `keybindings.json` files are still accepted and migrated to `keybindings.yml`; `keybindings.yaml` is also accepted.
diff --git a/docs/models.md b/docs/models.md
index daa9e7a2a..a1ca6d89f 100644
--- a/docs/models.md
+++ b/docs/models.md
@@ -557,9 +557,9 @@ For `anthropic-messages` models the runtime uses a separate `AnthropicCompat` sh
 
 ### Strict tool schemas (`disableStrictTools`)
 
-Anthropic's API supports a `strict` field on tool definitions that forces the model to always follow the provided schema exactly. This is enabled by default for all `anthropic-messages` providers because it guarantees schema conformance in agentic systems.
+Anthropic's API supports a `strict` field on tool definitions that forces the model to always follow the provided schema exactly. OMP enables it by default for a small allowlist of high-frequency built-in `anthropic-messages` tools (`bash`, `python`, `edit`, and `find`) whose schemas fit Anthropic's strict grammar limits; other tools still send normalized schemas but omit `strict`.
 
-Third-party providers that front the Anthropic API (AWS Bedrock, Azure, self-hosted proxies) do not always implement this field and will reject requests that include it. Set `disableStrictTools: true` at the provider level to opt out:
+Third-party providers that front the Anthropic API (AWS Bedrock, Azure, self-hosted proxies) do not always implement this field and will reject requests that include it. Set `disableStrictTools: true` at the provider level to opt out of strict mode for the allowlisted tools:
 
 ```yaml
 providers:
@@ -581,7 +581,7 @@ providers:
           cacheWrite: 3.75
 ```
 
-`disableStrictTools` is a provider-level flag that applies to all models in the provider.
+`disableStrictTools` is a provider-level flag that applies to all models in the provider. It disables the Anthropic `strict` marker only for tools that OMP would otherwise mark strict; it does not change runtime tool argument validation. OMP can automatically retry without strict tools after Anthropic reports a strict-grammar-too-large error before the first streamed token, but proxies that reject the `strict` field for other reasons should set this flag explicitly.
 
 Tool schemas going on the wire are normalized by the unified flow in
 `packages/ai/src/utils/schema/normalize.ts` (Google/CCA/MCP dispatchers
diff --git a/docs/session-operations-export-share-fork-resume.md b/docs/session-operations-export-share-fork-resume.md
index 0173137c1..0d6de4f70 100644
--- a/docs/session-operations-export-share-fork-resume.md
+++ b/docs/session-operations-export-share-fork-resume.md
@@ -1,4 +1,4 @@
-# Session Operations: export, dump, share, fork, resume/continue
+# Session Operations: export, dump, share, fresh, fork, resume/continue
 
 This document describes operator-visible behavior for session export/share/fork/resume operations as currently implemented.
 
@@ -19,6 +19,7 @@ This document describes operator-visible behavior for session export/share/fork/
 | `/export [path]`                        | Interactive slash command | No                                    | No                                                                                 | HTML file                                                       |
 | `--export <session.jsonl> [outputPath]` | CLI startup fast-path     | No runtime session mutation           | No active session; reads target file                                               | HTML file                                                       |
 | `/share`                                | Interactive slash command | No                                    | No                                                                                 | Temp HTML + share URL/gist                                      |
+| `/fresh`                               | Interactive slash command | Yes (provider-facing in-memory id/state only) | No; keeps current session file/header                                               | None                                                            |
 | `/fork`                                 | Interactive slash command | Yes (active session identity changes) | Creates new session file and switches current session to it (persistent mode only) | Copies artifact directory to new session namespace when present |
 | `--fork <id\|path>`                     | CLI startup               | Yes after session creation            | Creates a new session fork from the selected source into current cwd/session dir   | None                                                            |
 | `/resume`                               | Interactive slash command | Yes (active in-memory state replaced) | Switches to selected existing session file                                         | None                                                            |
diff --git a/docs/skills/authoring-extensions.md b/docs/skills/authoring-extensions.md
index 6419f153d..5d6c8367b 100644
--- a/docs/skills/authoring-extensions.md
+++ b/docs/skills/authoring-extensions.md
@@ -242,7 +242,7 @@ The derived name is the filename stem (or directory name for `index.ts`-style en
 - **Do not call runtime actions during load.** Methods like `pi.sendMessage()` throw `ExtensionRuntimeNotInitializedError` if called synchronously during module evaluation (before a session is active). Register handlers/tools/commands during load; perform runtime actions only from event handlers, tools, or commands.
 - **`tool_call` errors are fail-closed.** If a `tool_call` handler throws, the tool is blocked.
 - **Command names must not clash with built-ins.** Conflicts are skipped with a diagnostic log.
-- **Reserved shortcuts are ignored** (`ctrl+c`, `ctrl+d`, `ctrl+z`, `ctrl+k`, `ctrl+p`, `ctrl+l`, `ctrl+o`, `ctrl+t`, `ctrl+g`, `shift+tab`, `shift+ctrl+p`, `alt+enter`, `escape`, `enter`).
+- **Reserved shortcuts are ignored** (`ctrl+c`, `ctrl+d`, `ctrl+z`, `ctrl+k`, `ctrl+p`, `ctrl+l`, `ctrl+o`, `ctrl+t`, `ctrl+g`, `ctrl+q`, `alt+m`, `shift+tab`, `shift+ctrl+p`, `alt+enter`, `escape`, `enter`).
 
 ## Further reading
 
diff --git a/docs/theme.md b/docs/theme.md
index 738ab3b5b..36be98c4e 100644
--- a/docs/theme.md
+++ b/docs/theme.md
@@ -85,7 +85,7 @@ If omitted, export code derives defaults from resolved theme colors.
 
 - `symbols.preset` sets a theme-level default symbol set.
 - `symbols.overrides` can override individual `SymbolKey` values.
-- `symbols.spinnerFrames` overrides the loading spinner frames. Accepts either a flat `string[]` (applied to both spinner types) or an object `{ "status"?: string[], "activity"?: string[] }` to override each type independently. Any type not specified falls back to the symbol preset's default frames. `status` drives the ~12.5fps spinner used by loaders and tool-execution indicators; `activity` drives the ~60fps spinner used by markdown progress bars and similar high-frequency UI.
+- `symbols.spinnerFrames` overrides the loading spinner frames. Accepts either a flat `string[]` (applied to both spinner types) or an object `{ "status"?: string[], "activity"?: string[] }` to override each type independently. Any type not specified falls back to the symbol preset's default frames. `status` drives the ~12.5fps spinner used by loaders and tool-execution indicators; `activity` drives the ~30fps spinner used by markdown progress bars and similar high-frequency UI.
 
 Runtime precedence:
 
diff --git a/docs/tools/search_tool_bm25.md b/docs/tools/search_tool_bm25.md
index 1de189867..ab362da7e 100644
--- a/docs/tools/search_tool_bm25.md
+++ b/docs/tools/search_tool_bm25.md
@@ -42,7 +42,7 @@
 - The renderer shows a status line plus up to 5 collapsed tree items by default (`COLLAPSED_MATCH_LIMIT`), each with label, optional server name, score to 3 decimals, and truncated description. The ranked match list is not serialized into `content`.
 
 ## Flow
-1. `SearchToolBm25Tool.createIf()` in `packages/coding-agent/src/tools/search-tool-bm25.ts` exposes the tool only when `tools.discoveryMode` is set to a non-`"off"` value or legacy `mcp.discoveryMode === true`, and only if the session implements the discovery hooks.
+1. `SearchToolBm25Tool.createIf()` in `packages/coding-agent/src/tools/search-tool-bm25.ts` exposes the tool for explicit discovery modes (`"mcp-only"` / `"all"`) or legacy `mcp.discoveryMode === true`. The default `"auto"` mode is resolved later by `createAgentSession()` after MCP/extension tools are registered.
 2. `description` is rendered from `packages/coding-agent/src/prompts/tools/search-tool-bm25.md` via `renderSearchToolBm25Description()`, using the current discoverable-tool list plus per-server summary/count.
 3. `execute()` re-checks capability and settings:
    - missing discovery hooks -> `ToolError("Tool discovery is unavailable in this session.")`
@@ -59,9 +59,10 @@
 
 ## Modes / Variants
 - Discovery-mode gating:
+  - `tools.discoveryMode = "auto"` (default): when the registered tool set has more than 40 tools, searches hidden MCP tools only; otherwise discovery stays off.
   - `tools.discoveryMode = "all"`: searches hidden discoverable built-ins plus hidden MCP tools.
   - `tools.discoveryMode = "mcp-only"`: searches hidden MCP tools only.
-  - legacy `mcp.discoveryMode = true` with `tools.discoveryMode = "off"`: same as MCP-only.
+  - legacy `mcp.discoveryMode = true`: same as MCP-only.
 - Search-index source:
   - generic cached discoverable index from the session
   - legacy cached MCP index, cast to the generic shape
@@ -113,6 +114,6 @@
   - Built-in entries appear only in `"all"` mode and only for registry tools whose `loadMode === "discoverable"` and are not currently active.
   - Hidden/internal built-ins are intentionally excluded from the built-in corpus: `resolve`, `yield`, `report_finding`, `report_tool_issue` are called out in the `#collectDiscoverableBuiltinTools()` comment.
 - `DiscoverableToolSource` includes `"extension"` and `"custom"`, but `AgentSession.getDiscoverableTools()` currently assembles only built-in and MCP sources.
-- On startup, `packages/coding-agent/src/sdk.ts` hides non-essential discoverable built-ins in `tools.discoveryMode = "all"`; defaults are `read`, `bash`, and `edit` unless `tools.essentialOverride` changes them.
+- On startup, `packages/coding-agent/src/sdk.ts` resolves `"auto"` after the full registry exists and injects `search_tool_bm25` when the count exceeds 40. It hides non-essential discoverable built-ins only in `tools.discoveryMode = "all"`; defaults are `read`, `bash`, and `edit` unless `tools.essentialOverride` changes them.
 - Query tokenization is simple and deterministic: Unicode is NFKD-normalized, combining marks are dropped, acronym/camelCase and digit-to-capital boundaries are split, non-letter/non-number characters become spaces, tokens are lowercased, and only non-empty tokens survive.
 - Scores are rounded differently by surface: `details.tools[].score` keeps 6 decimals; the TUI line renders 3.
diff --git a/docs/tools/web_search.md b/docs/tools/web_search.md
index 5b3aa380b..c62df385d 100644
--- a/docs/tools/web_search.md
+++ b/docs/tools/web_search.md
@@ -161,9 +161,9 @@ Streaming: none. `WebSearchTool.execute()` forwards its `AbortSignal` into `exec
     - Output: `sources`, `requestId`.
   - **Kagi** — `packages/coding-agent/src/web/search/providers/kagi.ts`, `packages/coding-agent/src/web/kagi.ts`
     - Availability: env or `agent.db` credential for `kagi`.
-    - Querying: GET `https://kagi.com/api/v0/search?q=<query>&limit=<n>` with `Authorization: Bot <key>`.
+    - Querying: POST `https://kagi.com/api/v1/search` with `Authorization: Bearer <key>` and JSON body `{ query, workflow: "search", limit, filters?: { after } }`. `recency` maps to `filters.after` as a UTC `YYYY-MM-DD` string (`day`/`week`/`month`/`year`).
     - `limit` and `num_search_results` are collapsed together before dispatch, clamped to `1..40`, default `10`.
-    - Output: `sources`, `relatedQuestions`, `requestId`.
+    - Output: `sources` (concatenated `data.search` + `data.video` + `data.news` + `data.infobox`, with video/news/infobox results tagged in the title), `relatedQuestions` (`data.adjacent_question` + `data.related_search` `props.question`), `answer` (`data.direct_answer[0].snippet ?? title`), `requestId` (`meta.trace`).
   - **Synthetic** — `packages/coding-agent/src/web/search/providers/synthetic.ts`
     - Availability: env or `agent.db` credential for `synthetic`.
     - Querying: POST `https://api.synthetic.new/v2/search` with `{ query }`.
diff --git a/docs/tui-core-renderer.md b/docs/tui-core-renderer.md
new file mode 100644
index 000000000..ba9705e8c
--- /dev/null
+++ b/docs/tui-core-renderer.md
@@ -0,0 +1,389 @@
+# TUI core renderer — invariants & failure modes
+
+What you are dealing with before you touch the rendering engine. This is the
+companion to [`tui-runtime-internals.md`](./tui-runtime-internals.md): that doc
+maps the *flow* (input → component tree → render); this doc explains what
+**does not work, why it keeps breaking, and the invariants you must not
+violate**. Scope is the core engine only:
+
+- [`packages/tui/src/tui.ts`](../packages/tui/src/tui.ts) — render planner, intent emitters, native-scrollback bookkeeping, cursor placement.
+- [`packages/tui/src/terminal.ts`](../packages/tui/src/terminal.ts) — `ProcessTerminal`, capability probes, private-CSI reassembly.
+- [`packages/tui/src/terminal-capabilities.ts`](../packages/tui/src/terminal-capabilities.ts) — `TERMINAL` profile, ED3 risk / sync-output / DECCARA / image detection.
+- [`packages/tui/src/stdin-buffer.ts`](../packages/tui/src/stdin-buffer.ts) — escape-sequence reassembly.
+- [`packages/tui/src/utils.ts`](../packages/tui/src/utils.ts) — width/slice/wrap (the width model).
+- [`packages/tui/src/kitty-graphics.ts`](../packages/tui/src/kitty-graphics.ts) + [`components/image.ts`](../packages/tui/src/components/image.ts) — inline images.
+- [`packages/tui/src/deccara.ts`](../packages/tui/src/deccara.ts) — rectangular-fill optimizer.
+
+Application-layer renderers (transcript, tool calls, session tree, editor,
+widgets) are **out of scope** — they live in `packages/coding-agent`.
+
+---
+
+## 1. The one thing to understand first
+
+> **The renderer cannot observe the terminal's scroll position on most hosts it
+> runs on.** Every decision about rewriting native scrollback is therefore a
+> *guess*, and the guess has two opposite failure modes that cannot both be
+> avoided by a single policy.
+
+We keep our transcript on the **normal screen**. We deliberately have not moved
+the engine to the alternate screen: alt-screen would make the terminal handle
+viewport isolation, but the transcript/resume affordances would disappear with
+the alternate buffer. Keeping the normal screen means
+*we* own native scrollback, which means we must decide, per frame, whether it is
+safe to rebuild it. To rebuild history we emit xterm **ED3** (`CSI 3 J`, erase
+saved lines). Deciding when ED3 is safe requires knowing whether the user has
+scrolled up — and we usually can't:
+
+- **ConPTY hosts** (Windows Terminal, Tabby, Hyper, VS Code, conhost): the
+  pseudo-console buffer is pinned to the visible grid, so any "am I at the
+  bottom?" console query answers "yes" even when the reader scrolled up. The
+  probe *lies*.
+- **POSIX terminals**: there is no scroll-position API at all. The probe is
+  *absent*.
+
+So `Terminal.isNativeViewportAtBottom()` returns `true` / `false` / **`undefined`**,
+and `undefined` ("unknown") is the common case. The whole renderer is built
+around not trusting `undefined`.
+
+### The two-way bind
+
+| If you guess… | …and you're wrong | Symptom |
+|---|---|---|
+| **Eager** (rebuild now → emit `CSI 3 J`) | reader was scrolled up | **YANK** to top + **FLASH** on terminals that snap scroll on ED3 |
+| **Defer** (emit nothing, reconcile later) | viewport really was at the bottom | **CORRUPTION** (stale/duplicated rows) + **invisible-until-resize** |
+
+Yank, flash, and buffer corruption are **the same bug wearing three masks.**
+Historically, every fix that suppressed one mask for one terminal class
+re-enabled the opposite mask for a neighbouring class, and the follow-on
+complaint landed within a day. If you "fix flashing" by making rebuilds more
+eager, you will reintroduce yank. If you "fix yank" by deferring more, you will
+reintroduce corruption / invisibility. **Do not move this lever without the
+fidelity harness (§9) green.**
+
+---
+
+## 2. The render-intent planner (what you are editing)
+
+`#doRender` is split into a **planner** (`#planRender`) that classifies a frame
+into exactly one `RenderIntent`, and one `#emit*` method per intent that owns
+the bytes written and the state update. All state flows through a single
+`#commit` checkpoint at the end of every emitter. The intent union
+(`tui.ts`, search `type RenderIntent`):
+
+| Intent | Emits | When |
+|---|---|---|
+| `noop` | cursor only | nothing visible changed |
+| `initial` | clear viewport, paint transcript, **keep** prior shell scrollback | first paint after `start()` |
+| `sessionReplace` | clear viewport **+ ED3** (outside multiplexers) | caller forced `{ clearScrollback: true }` (switch/branch/reload/resume) |
+| `historyRebuild` | clear viewport **+ ED3** (outside multiplexers) | geometry change rewrapped history, or a proven-at-tail rebuild |
+| `overlayRebuild` | rebuild viewport with overlay composite | overlay visibility changed |
+| `liveRegionPinned` | relative moves + per-row rewrite/suffix-clear + `\r\n` | foreground streaming on an ED3-risk host, commit-as-you-go |
+| `viewportRepaint` | rewrite the visible viewport in place (optional `appendFrom` tail first) | safe non-destructive repaint |
+| `deferredShrink` | padded viewport repaint, history left dirty | bottom-anchored shrink, viewport unobservable |
+| `deferredMutation` | **zero bytes**, history left dirty | row-reindexing edit while possibly scrolled |
+| `shrink` / `diff` | trailing-row clear / changed-line diff | ordinary in-place updates |
+
+**ED3 (`CSI 3 J`) is emitted in exactly one place** — `#emitFullPaint` when
+`clearScrollback: true` (`\x1b[2J\x1b[H\x1b[3J`). The ordinary clear is
+**non-destructive**: `\x1b[22J` (copy-screen-to-scrollback, only when
+`TERMINAL.supportsScreenToScrollback`) then `\x1b[2J\x1b[H`, **no `3J`**. ED3 is
+reached only by `sessionReplace`/`historyRebuild`/`overlayRebuild`, and those
+suppress the scrollback clear inside multiplexers (`isMultiplexerSession()` =
+`TMUX || STY || ZELLIJ`).
+
+### The predicate gates
+
+Three private predicates encode the guessing policy. Do not "simplify" them —
+each branch is load-bearing:
+
+- `#canReplayNativeScrollbackAtCheckpoint(atBottom)` → `atBottom === true`. A
+  rebuild at a **keystroke checkpoint** (prompt submit) is allowed only with a
+  *positive* at-tail proof. A prompt submit is **no longer** treated as implicit
+  proof for an unobservable host.
+- `#canRebuildNativeScrollbackLive(atBottom, allowUnknown)` → `true` iff
+  `atBottom === true`, **or** (`atBottom === undefined && allowUnknown &&
+  platform !== "win32"`). i.e. live ED3 during streaming requires either proof
+  or an explicit direct-user-input opt-in, and **never** on win32.
+- `#nativeViewportIsScrolled(atBottom, allowUnknown)` → `true` if
+  `atBottom === false`, or (`undefined && win32 && !allowUnknown`). Used to
+  decide deferral.
+
+`allowUnknownViewportMutation` is the **direct-user-input opt-in** (autocomplete
+/ IME / a keystroke the user just typed). A keystroke pins the host viewport to
+the bottom, so it is safe to repaint live then. It is **not** set by passive
+streaming. `setEagerNativeScrollbackRebuild(true)` is the streaming opt-in; on
+ED3-risk hosts it is downgraded so it never promotes to a live ED3 clear.
+
+### Deferral + checkpoint discipline
+
+When the viewport is unobservable during **passive streaming**, the planner
+defers (`deferredMutation`/`deferredShrink`/`viewportRepaint`) and marks native
+scrollback dirty (`#markNativeScrollbackDirty()`). Reconciliation happens later
+at a checkpoint via `refreshNativeScrollbackIfDirty()` — and only if
+`#canReplayNativeScrollbackAtCheckpoint` proves at-tail. The streaming-defer +
+live-region-pin seam (`NativeScrollbackLiveRegion`,
+`getNativeScrollbackLiveRegionStart` / `getNativeScrollbackCommitSafeEnd`) is the
+**actively-churning** part of the engine; if you change how transient rows are
+committed, every structural-mutation branch (shrink **and** grow/offscreen-edit)
+must defer **symmetrically**, or you reopen the corruption family.
+
+---
+
+## 3. The five fault families
+
+### YANK — viewport snapped to top — NOT fully converged
+- **Mechanism:** a live `historyRebuild` fires `CSI 3 J` while the reader is
+  scrolled up; ED3-snap terminals reset the visible viewport to the top of the
+  (now-erased) scrollback.
+- **Trigger to avoid:** treating an unobservable probe as "at bottom" during
+  *passive* streaming, or OR-ing an eager-streaming flag into the live ED3 path.
+- **Current stance:** never emit ED3 on an unobservable host during passive
+  streaming; defer and reconcile at a keystroke checkpoint. ConPTY/win32 never
+  trust the probe at all.
+
+### CORRUPTION — duplicated / stale rows — NOT fully converged
+- **Mechanism:** the flip side of the yank fix. A deferred/repainted frame
+  leaves rows already committed to native scrollback out of sync with the live
+  viewport; the scrollback↔viewport seam duplicates (e.g. a 2-row dup, a
+  streaming-tail dup, or an async-expansion dup).
+- **Trigger to avoid:** repainting the viewport over scrollback that still holds
+  the old copy; a frozen/deferred block whose snapshot no longer matches after
+  the region above it reflowed; one mutation branch deferring while its mirror
+  branch repaints.
+- **Current stance:** commit only the **stable prefix** line-count to native
+  history; keep unstable rows out; reconcile drift at the checkpoint; park the
+  hardware cursor at real content bottom, not padded bottom.
+
+### FLASH (and invisible-until-resize) — NOT fully converged
+- **Two distinct causes, one symptom:**
+  - *Flash* = eager ED3 rebuild wrapped in DEC 2026 BSU/ESU fired per streaming
+    frame on a terminal that clamps scroll on ED3 (VTE/GNOME family).
+  - *Invisible-until-resize* = the defer fix over-firing, so a structural frame
+    emits **zero bytes** (`deferredMutation` returns nothing) until a resize
+    forces a repaint.
+- **Trigger to avoid:** env-detection that misses a flashing terminal (SSH
+  strips `VTE_VERSION`; some hosts set no distinguishing var); collapsing an
+  `undefined` probe into a definite scrolled/at-bottom verdict.
+- **Current stance:** confine ED3 to the destructive path; auto-disable DEC 2026
+  at runtime when the terminal reports it unsupported (DECRQM), with
+  `PI_NO_SYNC_OUTPUT` as a manual hatch; keep autowrap discipline regardless.
+
+### WIDTH — measurement crashes / fidelity — crash class dead, accuracy unproven
+- **Mechanism:** the measured column width of a line disagreed with the
+  terminal's painted cells (emoji, wide graphemes, combining marks, Hangul
+  jamo), and the old render loop **threw** on any mismatch — a 1-cell cosmetic
+  error became a fatal whole-agent crash.
+- **Current stance:** **never throw in the render hot path — clamp.** The loop
+  truncates over-wide lines with `truncateToWidth`/`sliceByColumn` and logs
+  (under debug) instead of dying. Width is owned end-to-end by one native UAX#11
+  engine shared by measure/slice/wrap (see §6). Accuracy across all scripts
+  (e.g. RTL/combining marks) is still not proven by a green gate.
+
+### PROBE — stray bytes injected as keystrokes — RESOLVED
+- **Mechanism:** a private-CSI probe reply (DA1 / kitty / mode 2031) split
+  across a stdin flush; the unmatched prefix was dropped and the continuation
+  bytes were forwarded as keystrokes.
+- **Current stance:** buffer-and-reassemble partial CSI responses; give each
+  probe a typed sentinel owner. This is the **one cleanly-closed family** —
+  because its contract is *bounded and observable* (bytes in = bytes out),
+  unlike the unobservable-viewport families. See §7.
+
+---
+
+## 4. Invariants — MUST / NEVER
+
+These are the rules the recurrence taught us. Treat them as load-bearing.
+
+1. **NEVER add a new `CSI 3 J` (ED3) callsite.** ED3 must flow only through
+   `#emitFullPaint({ clearScrollback: true })`, for the existing destructive
+   intents (`sessionReplace`, proven/safe `historyRebuild`, `overlayRebuild`).
+   Ordinary redraws use the non-destructive `\x1b[22J` + `\x1b[2J\x1b[H` clear.
+2. **NEVER trust an unobservable viewport probe (`undefined`) for *passive*
+   streaming.** Only a positive at-tail proof, or a direct-user-input opt-in
+   (`allowUnknownViewportMutation`), authorizes a live rebuild — and never on
+   win32/ConPTY.
+3. **NEVER throw in the render hot path.** Clamp over-wide lines; a width
+   mismatch is cosmetic, not fatal.
+4. **NEVER let a defer path emit a structurally-changed frame as zero bytes
+   while at the bottom** — that is invisible-until-resize. `deferredMutation`/
+   `deferredShrink` are only safe when the viewport is (or may be) scrolled.
+5. **Defer symmetrically.** If one structural-mutation branch (shrink) defers on
+   an unobservable ED3-risk host, the mirror branch (grow / offscreen-edit) must
+   too. Asymmetry reopens corruption.
+6. **Commit only the stable prefix to native history.** Transient/unsettled rows
+   stay out of scrollback until a checkpoint; reconcile drift at the checkpoint.
+7. **Park the hardware cursor at real content bottom**, not the padded viewport
+   bottom, or height shrinks scroll live rows into scrollback and duplicate them
+   per resize step.
+8. **Cursor writes live *inside* the synchronized-output frame**, before ESU —
+   never as a second frame after it (that teleports/blinks the caret).
+9. **Detect terminal *risk*, not terminal *brand*, and default unknown to
+   risky.** Env sniffing is necessarily incomplete (see §5); never assume an
+   un-enumerated host is safe.
+10. **Multiplexers (tmux/screen/zellij) get no destructive scrollback clear and
+    no viewport probe.** ED3 is a no-op there and a full replay duplicates the
+    transcript; repaint in place and rely on the pinned/commit-as-you-go path.
+11. **Any change to the eager/defer lever, the predicates, or the live-region
+    seam must be validated by the render-stress fidelity harness (§9)** across
+    `{win32, POSIX} × {unknown, scrolled, at-bottom}`, not by a single-terminal
+    smoke test.
+
+---
+
+## 5. Terminal capability detection (and why it is fragile)
+
+`TERMINAL` (`terminal-capabilities.ts`) is resolved once at import from
+`TERMINAL_ID` plus environment sniffing. The detection helpers are pure and
+parameterized over `(env, platform)` so they are unit-testable:
+
+- `detectTerminalEagerEraseScrollbackRisk(env, platform)` → is a live ED3
+  rebuild unsafe here? Current policy: `false` on win32 (dedicated ConPTY
+  deferral paths handle it) and when `PI_TUI_ED3_SAFE=1`; otherwise **`true`**
+  for `WT_SESSION` (WT fronting WSL), SSH/tmux/screen/zellij, known
+  ED3-snap/scrollback-clearing terminals (WezTerm, kitty, ghostty, alacritty,
+  VTE, iTerm2, Apple Terminal, GNOME Terminal, Ptyxis, xfce4-terminal), Linux
+  truecolor, **and every other unknown POSIX terminal**. The default is *risky*
+  on purpose.
+- `shouldEnableSynchronizedOutputByDefault(env, id)` → DEC 2026 default. Precedence:
+  user opt-out (`PI_NO_SYNC_OUTPUT`/`PI_TUI_SYNC_OUTPUT=0`) → user force-on
+  (`PI_FORCE_SYNC_OUTPUT=1`/`PI_TUI_SYNC_OUTPUT=1`) → `TERM_FEATURES` advertises
+  `Sy` → `WT_SESSION` (WT/WSL) → known direct terminals
+  (kitty/ghostty/wezterm/iterm2/alacritty/vscode; SSH passes through) → off for
+  risky multiplexers and everything else (VTE-family, GNU screen, Apple Terminal,
+  legacy conhost, unknown). Reconciled at runtime by the DECRQM mode-2026 report:
+  a positive report **enables** sync (upgrading default-off muxes like
+  zellij/tmux-master), a negative one disables it; a user override still wins.
+  `synchronizedOutputUserOverride(env)` is the shared opt-out/force resolver.
+- `detectRectangularSgrSupport(id, env)` → DECCARA fills: **kitty only**
+  (ghostty does not implement the SGR-background extension), off in multiplexers
+  and under `PI_NO_DECCARA`.
+
+**Why this keeps leaking:** terminal class is inferred from env vars that are
+**not durable**. `VTE_VERSION` is stripped by `sshd` (default `AcceptEnv`);
+`COLORTERM` is also not in default `AcceptEnv`; some hosts (Tabby) set no
+distinguishing var; WSL-fronting-WT is neither pure win32 nor pure POSIX. Every
+missed env var is a missed terminal class is a new complaint. The mitigations
+are: (a) **default unknown to risky** rather than safe, and (b) detect by
+*behavior/handshake* (DECRQM) where possible rather than a host allow-list. When
+you add a terminal, add it to the pure detector and add the **SSH-stripped env
+shape** to the test, not just the env-present shape.
+
+---
+
+## 6. Width model
+
+`visibleWidth` / `truncateToWidth` / `sliceByColumn` / `wrapTextWithAnsi`
+(`utils.ts`) all route through **one native UAX#11 engine** (`@oh-my-pi/pi-natives`,
+Rust `unicode-width`). We deliberately dropped `Bun.stringWidth` because it
+disagreed with the engine on combining marks and jamo, and mixing two width
+models in measure-vs-slice produced the crashes.
+
+- Fast path: printable ASCII is one cell per code unit.
+- ZWJ pictographic emoji take the `visibleWidthByGrapheme` override (ANSI spans
+  excised first, then `Intl.Segmenter`), because the native scanner double-counts
+  SGR bytes when a sequence is split by the segmenter.
+- OSC 66 sized text (`\x1b]66;…`) takes the native path.
+
+**Rule:** if you add a code path that measures width, route it through these
+helpers. Never reintroduce `Bun.stringWidth` or a parallel width table — the
+measure model and the slice/wrap model must agree, or you get over-wide lines
+that the hot-path clamp silently truncates (cosmetic loss) or, worse, seam
+duplication.
+
+---
+
+## 7. Capability probes & stdin reassembly
+
+`ProcessTerminal` fuses capability queries with a bare DA1 (`CSI c`) sentinel so
+a non-answering terminal is detected when DA1 returns first. Replies can arrive
+**split across a stdin flush**, so:
+
+- `#privateCsiResponseBuffer` accumulates `\x1b[?…` partials while a sentinel is
+  outstanding, rejoins on the terminator byte (0x40–0x7e), then runs the
+  DA1/kitty/mode-2031 handlers on the **complete** reply. A new `\x1b`
+  mid-reassembly or >256 bytes abandons the partial so real keys (e.g. arrow
+  `\x1b[A`) still reach input.
+- `#da1SentinelOwners` is a **typed FIFO** discriminated by `kind` (`keyboard`,
+  `osc11`, `privateMode`, `kittyGraphicsProbe`, `osc99Probe`) so a keyboard DA1
+  cannot be mistaken for an OSC 11 / DECRQM / graphics-probe sentinel.
+- DECRQM probes (`#queryPrivateMode(2026/2048/2031)`) record support via DECRPM
+  and drive runtime feature gating (e.g. auto-disabling DEC 2026 sync output).
+
+**Rule:** any new probe must own a typed sentinel and survive a split reply. The
+contract is bytes-in = bytes-out; it is testable, so test it (feed the reply
+byte-by-byte and assert nothing leaks to the input handler).
+
+---
+
+## 8. Inline images & memory
+
+Kitty images are **transmit-once, place-many** (`kitty-graphics.ts`):
+`encodeKittyTransmit` (`a=t`, keyed by a stable `i=`) writes the base64 a single
+time; repaints emit only `encodeKittyPlacement` (`a=p`). Text clears
+(`CSI 2 J` / `CSI 3 J`) do **not** purge the terminal's image store — only
+`encodeKittyDeleteImage` (`a=d,d=I`) does. `ImageBudget` (`components/image.ts`)
+keeps only the most-recent N images live; demoted images render their text
+fallback and are explicitly purged.
+
+**Rule:** never re-emit full base64 per frame (it pegged RAM and pinned the UI
+thread). Kitty Unicode placeholders are default-on only for kitty/ghostty
+(`PI_NO_KITTY_PLACEHOLDERS` / `PI_KITTY_PLACEHOLDERS`); other Kitty-protocol
+hosts render placeholder cells as literal PUA glyphs, so they fall back to
+direct `a=p` placement.
+
+---
+
+## 9. The fidelity gate (use it)
+
+`packages/tui/test/render-stress-harness.ts` renders the renderer's **real emitted ANSI** into
+a ghostty-web `VirtualTerminal` and asserts viewport fidelity (a scrolled reader
+stays put), background-column fidelity, and scrollback-buffer fidelity, across
+parameterized terminal shapes and randomized op sequences.
+
+This harness is the structural fix for the whole recurrence: every guess-flip and
+sniffing-gap regression historically **shipped blind and was caught by a user**,
+because no automated "a scrolled-up reader stays pinned across kitty/WT/WSL/
+ConPTY" assertion gated CI. **Before you change the eager/defer lever, a
+predicate, the live-region seam, or width math, run the stress harness and the
+targeted repro tests** (`packages/tui/test/render-regressions.test.ts`,
+`packages/tui/test/streaming-scrollback-defer.test.ts`, the `issue-*-repro.test.ts` files).
+A change that passes one terminal and one seed is not verified.
+
+---
+
+## 10. Escape hatches (env vars)
+
+| Var | Effect |
+|---|---|
+| `PI_NO_SYNC_OUTPUT=1` | Disable DEC 2026 BSU/ESU wrappers (autowrap discipline stays on). For terminals that advertise but mishandle mode 2026. |
+| `PI_TUI_SYNC_OUTPUT=0\|1` / `PI_FORCE_SYNC_OUTPUT=1` | Force sync output off / on. |
+| `PI_TUI_ED3_SAFE=1` | Declare the terminal safe for live ED3 (disables `eagerEraseScrollbackRisk`). |
+| `PI_NO_DECCARA` | Disable Kitty DECCARA rectangular-fill optimization (force padded-string fills). |
+| `PI_FORCE_IMAGE_PROTOCOL=kitty\|iterm2\|sixel\|off` | Override image protocol detection. |
+| `PI_NO_KITTY_PLACEHOLDERS=1` / `PI_KITTY_PLACEHOLDERS=1` | Force Kitty Unicode placeholders off / on. |
+| `PI_CLEAR_ON_SHRINK=1` | Clear empty rows when content shrinks (default off). |
+| `PI_HARDWARE_CURSOR=1` | Show the real hardware cursor instead of a rendered one. |
+| `PI_NOTIFICATIONS=off\|0\|false` | Suppress terminal notifications. |
+| `PI_DEBUG_REDRAW=1` | Log the chosen render intent per frame to the debug log. |
+| `PI_TUI_DEBUG=1` | Dump per-render diff state under `/tmp/tui`. |
+
+---
+
+## 11. Before you touch the render core — checklist
+
+- [ ] Are you about to emit `CSI 3 J` anywhere other than the destructive
+      `clearScrollback` path? **Stop.**
+- [ ] Does your change trust `isNativeViewportAtBottom() === undefined` as
+      "at bottom" during passive streaming? **Stop.**
+- [ ] Did you change one structural-mutation branch without mirroring its
+      sibling (shrink ↔ grow)? **Defer symmetrically.**
+- [ ] Could any frame now emit zero bytes while the viewport is at the bottom?
+      That's invisible-until-resize.
+- [ ] Did you add a terminal by brand instead of by behavior, or skip the
+      SSH-stripped env shape in the test?
+- [ ] Did you run `packages/tui/test/render-stress-harness.ts` + the repro suite across
+      win32/POSIX × unknown/scrolled/at-bottom — not just one terminal?
+- [ ] New probe? Typed sentinel owner + split-reply test.
+- [ ] New width path? Routed through the shared native engine, clamped (never
+      thrown) in the hot path.
diff --git a/docs/tui-runtime-internals.md b/docs/tui-runtime-internals.md
index e5ef9e32d..79df9186d 100644
--- a/docs/tui-runtime-internals.md
+++ b/docs/tui-runtime-internals.md
@@ -2,6 +2,12 @@
 
 This document maps the non-theme runtime path from terminal input to rendered output in interactive mode. It focuses on behavior in `packages/tui` and its integration from `packages/coding-agent` controllers.
 
+> **Editing the rendering engine itself?** Read
+> [`tui-core-renderer.md`](./tui-core-renderer.md) first — it documents the
+> failure modes (yank / corruption / flash / width crashes) and the invariants
+> the render planner, native-scrollback bookkeeping, and capability detection
+> must not violate.
+
 ## Runtime layers and ownership
 
 - **`packages/tui` engine**: terminal lifecycle, stdin normalization, focus routing, render scheduling, differential painting, overlay composition, hardware cursor placement.
@@ -11,44 +17,49 @@ Boundary rule: the TUI engine is message-agnostic. It only knows `Component.rend
 
 ## Implementation files
 
-- [`../src/modes/interactive-mode.ts`](../packages/coding-agent/src/modes/interactive-mode.ts)
-- [`../src/modes/controllers/event-controller.ts`](../packages/coding-agent/src/modes/controllers/event-controller.ts)
-- [`../src/modes/controllers/input-controller.ts`](../packages/coding-agent/src/modes/controllers/input-controller.ts)
-- [`../src/modes/components/custom-editor.ts`](../packages/coding-agent/src/modes/components/custom-editor.ts)
-- [`../../tui/src/tui.ts`](../packages/tui/src/tui.ts)
-- [`../../tui/src/terminal.ts`](../packages/tui/src/terminal.ts)
-- [`../../tui/src/editor-component.ts`](../packages/tui/src/editor-component.ts)
-- [`../../tui/src/stdin-buffer.ts`](../packages/tui/src/stdin-buffer.ts)
-- [`../../tui/src/components/loader.ts`](../packages/tui/src/components/loader.ts)
+- [`packages/coding-agent/src/modes/interactive-mode.ts`](../packages/coding-agent/src/modes/interactive-mode.ts)
+- [`packages/coding-agent/src/modes/controllers/event-controller.ts`](../packages/coding-agent/src/modes/controllers/event-controller.ts)
+- [`packages/coding-agent/src/modes/controllers/input-controller.ts`](../packages/coding-agent/src/modes/controllers/input-controller.ts)
+- [`packages/coding-agent/src/modes/components/custom-editor.ts`](../packages/coding-agent/src/modes/components/custom-editor.ts)
+- [`packages/tui/src/tui.ts`](../packages/tui/src/tui.ts)
+- [`packages/tui/src/terminal.ts`](../packages/tui/src/terminal.ts)
+- [`packages/tui/src/editor-component.ts`](../packages/tui/src/editor-component.ts)
+- [`packages/tui/src/stdin-buffer.ts`](../packages/tui/src/stdin-buffer.ts)
+- [`packages/tui/src/components/loader.ts`](../packages/tui/src/components/loader.ts)
 
 ## Boot and component tree assembly
 
-`InteractiveMode` constructs `TUI(new ProcessTerminal(), settings.get("showHardwareCursor"))`, applies `settings.get("clearOnShrink")`, and creates persistent containers:
+`InteractiveMode` constructs `TUI(new ProcessTerminal(), settings.get("showHardwareCursor"))`, applies `clearOnShrink`, `tui.maxInlineImages`, and Kitty text-sizing settings, then creates persistent containers:
 
 - `chatContainer`
 - `pendingMessagesContainer`
 - `statusContainer`
 - `todoContainer`
 - `btwContainer`
+- `omfgContainer`
+- `errorBannerContainer`
 - `statusLine`
 - `hookWidgetContainerAbove`
 - `editorContainer` (holds `CustomEditor`)
 - `hookWidgetContainerBelow`
 
-`init()` wires the tree in that order, focuses the editor, registers input handlers via `InputController`, subscribes terminal appearance changes into theme auto-detection, starts TUI, and requests a forced render.
-A forced render (`requestRender(true)`) resets previous-line caches and cursor bookkeeping before repainting.
+`init()` wires the tree in that order after any startup warnings/welcome/changelog, focuses the editor, registers input handlers via `InputController`, starts TUI, pushes terminal title state, updates the editor border, and requests a forced render.
+A forced render (`requestRender(true)`) queues a viewport repaint or explicit session replacement; it does **not** throw away previous-line history by default.
 
 ## Terminal lifecycle and stdin normalization
 
 `ProcessTerminal.start()`:
 
 1. Enables raw mode and bracketed paste.
-2. Attaches resize handler.
-3. Creates a `StdinBuffer` to split partial escape chunks into complete sequences.
-4. Queries Kitty keyboard protocol support (`CSI ? u`), then enables protocol flags if supported; otherwise enables modifyOtherKeys fallback after a short timeout.
-5. Queries OSC 11 background color and enables Mode 2031 appearance notifications for dark/light theme detection.
-6. On Windows, attempts VT input enablement via `kernel32` mode flags.
-   `StdinBuffer` behavior:
+2. Attaches resize handler and refreshes dimensions.
+3. Enables Windows VT input mode when running on win32.
+4. Creates a `StdinBuffer` to split partial escape chunks into complete sequences.
+5. Queries Kitty keyboard protocol support (`CSI ? u`), then enables protocol flags if supported; otherwise enables modifyOtherKeys fallback after a short timeout.
+6. Queries OSC 11 background color and Mode 2031 appearance notifications for dark/light theme detection.
+7. Queries OSC 99 notification capabilities.
+8. Starts periodic OSC 11 polling only where safe, then probes DEC private modes 2026/2048/2031 via DECRQM.
+
+`StdinBuffer` behavior:
 
 - Buffers fragmented escape sequences (CSI/OSC/DCS/APC/SS3).
 - Emits `data` only when a sequence is complete or timeout-flushed.
@@ -90,7 +101,7 @@ This keeps key parsing/editor mechanics in `packages/tui` and mode semantics in
 
 `TUI.requestRender()` coalesces render requests and rate-limits ordinary frames:
 
-- forced renders (`requestRender(true, ...)`) reset cached frame/viewport state and run on `process.nextTick`
+- forced renders (`requestRender(true, ...)`) schedule an immediate frame and set `#forceViewportRepaintOnNextRender`; with `clearScrollback`, they also queue `sessionReplace`
 - ordinary renders schedule through `#scheduleRender()` and respect `TUI.#MIN_RENDER_INTERVAL_MS`
 - repeated requests while a render is pending collapse into the same scheduled frame
 
@@ -110,7 +121,7 @@ This keeps key parsing/editor mechanics in `packages/tui` and mode semantics in
    - noop
 6. Emit only the bytes required by the intent and commit cached frame/cursor/viewport state.
 
-Render writes use synchronized output mode (`CSI ? 2026 h/l`) to reduce flicker/tearing.
+Render writes use synchronized output mode (`CSI ? 2026 h/l`) when enabled; capability detection, DECRQM, or `PI_NO_SYNC_OUTPUT` can disable the wrappers while leaving autowrap discipline on.
 
 ## Render safety constraints
 
@@ -123,14 +134,19 @@ Critical safety checks in `TUI`:
 
 These constraints are runtime guards plus component conventions; renderers should still return width-safe lines rather than rely on truncation.
 
+The deeper reasons these guards exist — why the renderer cannot observe scroll
+position, why ED3 (`CSI 3 J`) is confined to one path, and why the hot path
+clamps instead of throwing — are documented in
+[`tui-core-renderer.md`](./tui-core-renderer.md).
+
 ## Resize handling
 
 Resize events are event-driven from `ProcessTerminal` to `TUI.requestRender()`.
 
 Effects:
 
-- Width changes repaint or rebuild because wrapping semantics change.
-- Height-only changes repaint the viewport when needed, but skip repaint in Termux and terminal multiplexers where replays are scrollback-hostile.
+- Width or height changes repaint or rebuild because terminal reflow invalidates wrapping, viewport, and cursor anchors.
+- Inside terminal multiplexers, resize uses viewport repaint instead of destructive native-scrollback replay; pane history cannot be erased safely and a full replay duplicates transcript rows.
 - Viewport/top tracking (`#viewportTopRow`, `#maxLinesRendered`, scrollback high-water state) avoids invalid relative cursor math and defers destructive native scrollback rewrites while the user is scrolled into history.
 - Overlay visibility can depend on terminal dimensions (`OverlayOptions.visible`); focus is corrected when overlays become non-visible after resize.
 
@@ -183,18 +199,6 @@ Escape exits inactive mode by clearing editor text and restoring border color; w
 2. Stops TUI before suspend.
 3. Sends `SIGTSTP` to process group.
 
-### Background mode (`/background` or `/bg`)
-
-`handleBackgroundCommand()`:
-
-- Rejects when idle.
-- Switches tool UI context to non-interactive (`hasUI=false`) so interactive UI tools fail fast.
-- Stops loaders/status line and unsubscribes foreground event handler.
-- Subscribes background event handler (primarily waits for `agent_end`).
-- Stops TUI and sends `SIGTSTP` (POSIX job control path).
-
-On `agent_end` in background with no queued work, controller sends completion notification and shuts down.
-
 ## Cancellation paths
 
 Primary cancellation inputs:
diff --git a/docs/tui.md b/docs/tui.md
index a62b5f0fe..cba829fbe 100644
--- a/docs/tui.md
+++ b/docs/tui.md
@@ -28,7 +28,7 @@ export interface Component {
   render(width: number): string[];
   handleInput?(data: string): void;
   wantsKeyRelease?: boolean;
-  invalidate(): void;
+  invalidate?(): void;
 }
 ```
 
diff --git a/package.json b/package.json
index 7f85dbc86..435a43242 100644
--- a/package.json
+++ b/package.json
@@ -20,15 +20,15 @@
       "@huggingface/transformers": "^4.2.0",
       "@mozilla/readability": "^0.6.0",
       "@napi-rs/cli": "3.7.0",
-      "@oh-my-pi/hashline": "15.9.4",
-      "@oh-my-pi/omp-stats": "15.9.4",
-      "@oh-my-pi/pi-agent-core": "15.9.4",
-      "@oh-my-pi/pi-ai": "15.9.4",
-      "@oh-my-pi/pi-coding-agent": "15.9.4",
-      "@oh-my-pi/pi-mnemopi": "15.9.4",
-      "@oh-my-pi/pi-natives": "15.9.4",
-      "@oh-my-pi/pi-tui": "15.9.4",
-      "@oh-my-pi/pi-utils": "15.9.4",
+      "@oh-my-pi/hashline": "15.10.1",
+      "@oh-my-pi/omp-stats": "15.10.1",
+      "@oh-my-pi/pi-agent-core": "15.10.1",
+      "@oh-my-pi/pi-ai": "15.10.1",
+      "@oh-my-pi/pi-coding-agent": "15.10.1",
+      "@oh-my-pi/pi-mnemopi": "15.10.1",
+      "@oh-my-pi/pi-natives": "15.10.1",
+      "@oh-my-pi/pi-tui": "15.10.1",
+      "@oh-my-pi/pi-utils": "15.10.1",
       "@opentelemetry/api": "^1.9.1",
       "@opentelemetry/context-async-hooks": "^2.7.1",
       "@opentelemetry/exporter-trace-otlp-proto": "^0.218.0",
@@ -85,14 +85,15 @@
   },
   "overrides": {},
   "scripts": {
-    "install:dev": "bun install && bun --cwd=packages/coding-agent link && bun --cwd=packages/ai link",
+    "install:dev": "bun install && bun --cwd=packages/coding-agent link && ln -sfn \"$(pwd)/packages/coding-agent/scripts/dev-launch\" \"$(bun pm -g bin)/omp\"",
     "dev": "bun --cwd=packages/coding-agent src/cli.ts",
+    "dev:timing": "PI_TIMING=x bun --cwd=packages/coding-agent --preload ../utils/src/module-timer.ts src/cli.ts",
     "stats": "bun --cwd=packages/coding-agent src/cli.ts stats",
     "claude:trace": "bun scripts/claude-trace.ts",
     "build": "bun run --workspaces --if-present build",
     "build:native": "bun --cwd=packages/natives run build",
     "test": "bun run --parallel test:ts test:rs",
-    "test:ts": "GITHUB_ACTIONS=0 bun run --workspaces --if-present test -- --only-failures",
+    "test:ts": "GITHUB_ACTIONS= bun run --workspaces --if-present test -- --only-failures",
     "test:rs": "bun scripts/run-rs-task.ts test:rs",
     "check": "bun run --parallel check:ts check:rs",
     "check:ts": "bun run check:tools && bun run --workspaces --if-present check",
diff --git a/packages/agent/CHANGELOG.md b/packages/agent/CHANGELOG.md
index 22d853150..5e4dd0b4f 100644
--- a/packages/agent/CHANGELOG.md
+++ b/packages/agent/CHANGELOG.md
@@ -2,6 +2,26 @@
 
 ## [Unreleased]
 
+## [15.10.1] - 2026-06-07
+
+### Added
+
+- Added optional `promptCacheKey` support to `AgentOptions` and `Agent` via a new `promptCacheKey` property so providers can receive a caller-provided prompt cache key
+- Added optional `ApiKeyResolveContext` parameter to `getApiKey` in `AgentOptions` and `AgentLoopConfig` so key resolvers can receive retry context
+
+### Changed
+
+- Enabled streaming API calls to re-resolve credentials through the `getApiKey` callback when retries occur after authentication-related errors
+- `Agent.abort(reason?)` now forwards `reason` to the underlying `AbortController`, and the synthesized aborted assistant message carries that reason on `errorMessage` (string or non-`AbortError` `Error` message) instead of always defaulting to `"Request was aborted"`. Bare `abort()` is unchanged.
+
+### Fixed
+
+- Fixed handling of short-lived API keys so that expired tokens are retried with a refreshed value during 401/usage-limit failures
+- Ensured fallback API key resolution uses the initially configured static `apiKey` when `getApiKey` is present
+- Wrapped oneshot LLM completions (`instrumentedCompleteSimple`: handoff, compaction/branch summaries) in an `EventLoopKeepalive`. These run outside the agent `#runLoop`, so without the keepalive Bun's event loop stopped servicing timers while parked on the completion promise — freezing host spinners (e.g. the `/handoff` loader) until an unrelated terminal resize poked the loop into rendering again.
+
+## [15.9.5] - 2026-06-05
+
 ### Fixed
 
 - Surfaced Anthropic stream failures whose message starts with `Output blocked by conten` as normal assistant error lifecycle events, so interactive clients render content-filter blocks instead of silently dropping the streaming bubble at `agent_end`.
diff --git a/packages/agent/package.json b/packages/agent/package.json
index cb7560996..6d256bd40 100644
--- a/packages/agent/package.json
+++ b/packages/agent/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-agent-core",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "General-purpose agent with transport abstraction, state management, and attachment support",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
diff --git a/packages/agent/src/agent-loop.ts b/packages/agent/src/agent-loop.ts
index 565420527..c4bddb3cd 100644
--- a/packages/agent/src/agent-loop.ts
+++ b/packages/agent/src/agent-loop.ts
@@ -3,6 +3,7 @@
  * Transforms to Message[] only at the LLM call boundary.
  */
 import {
+	type ApiKeyResolveContext,
 	type AssistantMessage,
 	type AssistantMessageEvent,
 	type Context,
@@ -727,8 +728,9 @@ async function streamAssistantResponse(
 	// Resolve API key (important for expiring tokens) — do this before resolving
 	// metadata so that the session-sticky credential recorded by getApiKey is
 	// visible to metadataResolver (e.g. for the correct account_uuid in metadata.user_id).
+	const staticApiKey = typeof config.apiKey === "string" ? config.apiKey : undefined;
 	const resolvedApiKey =
-		(config.getApiKey ? await config.getApiKey(config.model.provider) : undefined) || config.apiKey;
+		(config.getApiKey ? await config.getApiKey(config.model.provider) : undefined) || staticApiKey;
 
 	// Re-resolve metadata after credential selection so the per-request value
 	// reflects the credential actually used, not the snapshot from AgentLoopConfig construction.
@@ -798,7 +800,19 @@ async function streamAssistantResponse(
 		return await runInActiveSpan(chatSpan, async () => {
 			const response = await streamFunction(config.model, llmContext, {
 				...config,
-				apiKey: resolvedApiKey,
+				// Hand streamSimple a resolver so its central auth-retry policy can
+				// re-resolve on 401 / usage-limit: the initial step reuses the key
+				// already resolved above (which set the session-sticky credential
+				// feeding metadataResolver), and retry steps forward the a/b/c ctx
+				// to config.getApiKey (force-refresh, then rotate). With no
+				// getApiKey hook the caller's own apiKey (string or resolver) flows
+				// through unchanged.
+				apiKey: config.getApiKey
+					? (ctx: ApiKeyResolveContext) =>
+							ctx.error === undefined
+								? resolvedApiKey
+								: Promise.resolve(config.getApiKey!(config.model.provider, ctx))
+					: config.apiKey,
 				metadata: resolvedMetadata,
 				toolChoice: effectiveToolChoice,
 				reasoning: effectiveReasoning,
@@ -839,7 +853,14 @@ async function streamAssistantResponse(
 			let detachAbortListener: (() => void) | undefined;
 			if (requestSignal) {
 				if (requestSignal.aborted) {
-					const aborted = emitAbortedAssistantMessage(partialMessage, addedPartial, context, config, stream);
+					const aborted = emitAbortedAssistantMessage(
+						partialMessage,
+						addedPartial,
+						context,
+						config,
+						stream,
+						requestSignal,
+					);
 					await finishChat(aborted);
 					return aborted;
 				}
@@ -861,7 +882,14 @@ async function streamAssistantResponse(
 								if (capped) return capped;
 							}
 							responseIterator.return?.()?.catch(() => {});
-							const aborted = emitAbortedAssistantMessage(partialMessage, addedPartial, context, config, stream);
+							const aborted = emitAbortedAssistantMessage(
+								partialMessage,
+								addedPartial,
+								context,
+								config,
+								stream,
+								requestSignal,
+							);
 							await finishChat(aborted);
 							return aborted;
 						}
@@ -874,7 +902,14 @@ async function streamAssistantResponse(
 							const capped = await finishCappedAssistantMessage();
 							if (capped) return capped;
 						}
-						const aborted = emitAbortedAssistantMessage(partialMessage, addedPartial, context, config, stream);
+						const aborted = emitAbortedAssistantMessage(
+							partialMessage,
+							addedPartial,
+							context,
+							config,
+							stream,
+							requestSignal,
+						);
 						await finishChat(aborted);
 						return aborted;
 					}
@@ -982,14 +1017,30 @@ async function streamAssistantResponse(
 	}
 }
 
+/** Resolve the human-readable reason an abort carried. A caller that aborts via
+ *  `AbortController.abort(reason)` with a string or a non-`AbortError` `Error`
+ *  (e.g. the coding agent's user-interrupt label) gets that text surfaced on the
+ *  synthesized assistant message's `errorMessage`; a bare `abort()` (whose
+ *  `signal.reason` is the default `AbortError` `DOMException`) falls back to the
+ *  generic sentinel that downstream renderers treat as "no specific reason". */
+export function abortReasonText(signal: AbortSignal | undefined): string {
+	const reason = signal?.reason;
+	if (typeof reason === "string" && reason.trim().length > 0) return reason;
+	if (reason instanceof Error && reason.name !== "AbortError" && reason.message.trim().length > 0) {
+		return reason.message;
+	}
+	return "Request was aborted";
+}
+
 function emitAbortedAssistantMessage(
 	partialMessage: AssistantMessage | null,
 	addedPartial: boolean,
 	context: AgentContext,
 	config: AgentLoopConfig,
 	stream: EventStream<AgentEvent, AgentMessage[]>,
+	requestSignal: AbortSignal | undefined,
 ): AssistantMessage {
-	const errorMessage = "Request was aborted";
+	const errorMessage = abortReasonText(requestSignal);
 	const abortedMessage: AssistantMessage = partialMessage
 		? { ...partialMessage, stopReason: "aborted", errorMessage }
 		: {
diff --git a/packages/agent/src/agent.ts b/packages/agent/src/agent.ts
index 1b3c57f46..99496de45 100644
--- a/packages/agent/src/agent.ts
+++ b/packages/agent/src/agent.ts
@@ -3,6 +3,7 @@
  */
 import { isPromise } from "node:util/types";
 import {
+	type ApiKeyResolveContext,
 	type AssistantMessage,
 	type AssistantMessageEvent,
 	type CursorExecHandlers,
@@ -21,7 +22,7 @@ import {
 	type ToolChoice,
 	type ToolResultMessage,
 } from "@oh-my-pi/pi-ai";
-import { agentLoop, agentLoopContinue } from "./agent-loop";
+import { abortReasonText, agentLoop, agentLoopContinue } from "./agent-loop";
 import type { AppendOnlyContextManager } from "./append-only-context";
 import type { HarmonyAuditEvent } from "./harmony-leak";
 import type {
@@ -132,6 +133,11 @@ export interface AgentOptions {
 	 * Used by providers that support session-based caching (e.g., OpenAI Codex).
 	 */
 	sessionId?: string;
+	/**
+	 * Optional prompt cache key forwarded to LLM providers.
+	 * When omitted, providers may fall back to sessionId.
+	 */
+	promptCacheKey?: string;
 	/**
 	 * Shared provider state map for session-scoped transport/session caches.
 	 */
@@ -141,7 +147,7 @@ export interface AgentOptions {
 	 * Resolves an API key dynamically for each LLM call.
 	 * Useful for expiring tokens (e.g., GitHub Copilot OAuth).
 	 */
-	getApiKey?: (provider: string) => Promise<string | undefined> | string | undefined;
+	getApiKey?: (provider: string, ctx?: ApiKeyResolveContext) => Promise<string | undefined> | string | undefined;
 
 	/**
 	 * Inspect or replace provider payloads before they are sent.
@@ -283,6 +289,7 @@ export class Agent {
 	#interruptMode: "immediate" | "wait";
 	#maxToolCallsPerTurn?: number;
 	#sessionId?: string;
+	#promptCacheKey?: string;
 	#metadata?: Record<string, unknown>;
 	#metadataResolver?: (provider: string) => Record<string, unknown> | undefined;
 	#providerSessionState?: Map<string, ProviderSessionState>;
@@ -319,7 +326,7 @@ export class Agent {
 	#cursorToolResultBuffer: CursorToolResultEntry[] = [];
 
 	streamFn: StreamFn;
-	getApiKey?: (provider: string) => Promise<string | undefined> | string | undefined;
+	getApiKey?: (provider: string, ctx?: ApiKeyResolveContext) => Promise<string | undefined> | string | undefined;
 	/**
 	 * Hook invoked after tool arguments are validated and before execution.
 	 * Reassign at any time to swap the implementation (e.g. on extension reload).
@@ -344,6 +351,7 @@ export class Agent {
 		this.#maxToolCallsPerTurn = opts.maxToolCallsPerTurn;
 		this.streamFn = opts.streamFn || streamSimple;
 		this.#sessionId = opts.sessionId;
+		this.#promptCacheKey = opts.promptCacheKey;
 		this.#providerSessionState = opts.providerSessionState;
 		this.#thinkingBudgets = opts.thinkingBudgets;
 		this.#temperature = opts.temperature;
@@ -390,6 +398,20 @@ export class Agent {
 		this.#sessionId = value;
 	}
 
+	/**
+	 * Get the prompt cache key forwarded to providers.
+	 */
+	get promptCacheKey(): string | undefined {
+		return this.#promptCacheKey;
+	}
+
+	/**
+	 * Set the prompt cache key forwarded to providers.
+	 */
+	set promptCacheKey(value: string | undefined) {
+		this.#promptCacheKey = value;
+	}
+
 	/**
 	 * Static metadata forwarded to every API request when no resolver is installed
 	 * (e.g. `metadata.user_id` for Anthropic session attribution). Setting this
@@ -768,8 +790,8 @@ export class Agent {
 		this.#state.messages.length = 0;
 	}
 
-	abort() {
-		this.#abortController?.abort();
+	abort(reason?: unknown) {
+		this.#abortController?.abort(reason);
 	}
 
 	waitForIdle(): Promise<void> {
@@ -936,6 +958,7 @@ export class Agent {
 			interruptMode: this.#interruptMode,
 			maxToolCallsPerTurn: this.#maxToolCallsPerTurn,
 			sessionId: this.#sessionId,
+			promptCacheKey: this.#promptCacheKey,
 			metadata: this.#metadataResolver ? undefined : this.#metadata,
 			metadataResolver: this.#metadataResolver,
 			providerSessionState: this.#providerSessionState,
@@ -1053,8 +1076,12 @@ export class Agent {
 				}
 			}
 		} catch (err) {
-			const errorMessage = err instanceof Error ? err.message : String(err);
 			const stoppedForAbort = this.#abortController?.signal.aborted === true;
+			const errorMessage = stoppedForAbort
+				? abortReasonText(this.#abortController?.signal)
+				: err instanceof Error
+					? err.message
+					: String(err);
 			const shouldEmitVisibleOutputBlockedError = !stoppedForAbort && isAnthropicOutputBlockedError(errorMessage);
 			const assistantPartial = partial?.role === "assistant" ? partial : undefined;
 			const hadAssistantStart = assistantPartial !== undefined;
diff --git a/packages/agent/src/telemetry.ts b/packages/agent/src/telemetry.ts
index 5cd99a135..8655ccb6c 100644
--- a/packages/agent/src/telemetry.ts
+++ b/packages/agent/src/telemetry.ts
@@ -50,6 +50,7 @@ import {
 } from "@opentelemetry/api";
 import { AgentRunCollector, type AgentRunCoverage, type AgentRunSummary, type ToolStatus } from "./run-collector";
 import type { AgentTool } from "./types";
+import { EventLoopKeepalive } from "./utils/yield";
 
 /** Default tracer name. Override via {@link AgentTelemetryConfig.tracerName}. */
 export const DEFAULT_TRACER_NAME = "@oh-my-pi/pi-agent-core";
@@ -1629,6 +1630,13 @@ export async function instrumentedCompleteSimple<TApi extends Api>(
 	options: SimpleStreamOptions,
 	span: InstrumentedChatSpanOptions,
 ): Promise<AssistantMessage> {
+	// Oneshot LLM calls (handoff, compaction/branch summaries) run outside the
+	// agent `#runLoop`, which is where the EventLoopKeepalive normally lives.
+	// Without it, Bun's JSC loop stops servicing timers while parked on the
+	// long-lived completion promise, freezing any host spinner (e.g. the
+	// `/handoff` Loader) until an unrelated I/O event (a terminal resize)
+	// pokes the loop. Keep the loop healthy for the duration of the call.
+	using _keepalive = new EventLoopKeepalive();
 	const { telemetry, parent, oneshotKind } = span;
 	const stepNumber = span.stepNumber ?? -1;
 	const reasoning = options.reasoning;
diff --git a/packages/agent/src/types.ts b/packages/agent/src/types.ts
index 7ca6a37ec..8ecc1458c 100644
--- a/packages/agent/src/types.ts
+++ b/packages/agent/src/types.ts
@@ -1,4 +1,5 @@
 import type {
+	ApiKeyResolveContext,
 	AssistantMessage,
 	AssistantMessageEvent,
 	AssistantMessageEventStream,
@@ -112,7 +113,7 @@ export interface AgentLoopConfig extends SimpleStreamOptions {
 	 * Useful for short-lived OAuth tokens (e.g., GitHub Copilot) that may expire
 	 * during long-running tool execution phases.
 	 */
-	getApiKey?: (provider: string) => Promise<string | undefined> | string | undefined;
+	getApiKey?: (provider: string, ctx?: ApiKeyResolveContext) => Promise<string | undefined> | string | undefined;
 
 	/**
 	 * Returns steering messages to inject into the conversation mid-run.
diff --git a/packages/agent/test/agent-loop.test.ts b/packages/agent/test/agent-loop.test.ts
index d0913c609..55a9ac7a8 100644
--- a/packages/agent/test/agent-loop.test.ts
+++ b/packages/agent/test/agent-loop.test.ts
@@ -157,6 +157,35 @@ describe("agentLoop with AgentMessage", () => {
 		expect(events.map(event => event.type)).toContain("agent_end");
 	});
 
+	it("surfaces a custom abort reason on the synthesized aborted message", async () => {
+		const context: AgentContext = {
+			systemPrompt: ["You are helpful."],
+			messages: [],
+			tools: [],
+		};
+		const mock = createMockModel();
+		const config: AgentLoopConfig = { model: mock.model, convertToLlm: identityConverter, maxToolCallsPerTurn: 8 };
+		const controller = new AbortController();
+		const streamFn = () => new AssistantMessageEventStream();
+
+		const stream = agentLoop([createUserMessage("Hello")], context, config, controller.signal, streamFn);
+		// Abort with a reason (as the coding agent does for a user Esc interrupt).
+		queueMicrotask(() => controller.abort("Interrupted by user"));
+
+		for await (const _event of stream) {
+			// drain
+		}
+
+		const messages = await stream.result();
+		const finalMessage = messages[messages.length - 1];
+		expect(finalMessage.role).toBe("assistant");
+		if (finalMessage.role !== "assistant") throw new Error("Expected assistant message");
+		expect(finalMessage.stopReason).toBe("aborted");
+		// The reason rides AbortController.abort(reason) onto the message verbatim,
+		// instead of the generic "Request was aborted" default.
+		expect(finalMessage.errorMessage).toBe("Interrupted by user");
+	});
+
 	it("should handle custom message types via convertToLlm", async () => {
 		// Create a custom message type
 		interface CustomNotification {
diff --git a/packages/agent/test/agent.test.ts b/packages/agent/test/agent.test.ts
index c54f9366a..db95e43df 100644
--- a/packages/agent/test/agent.test.ts
+++ b/packages/agent/test/agent.test.ts
@@ -354,6 +354,21 @@ describe("Agent", () => {
 		expect(reasoningPerCall).toEqual([ThinkingLevel.Low, ThinkingLevel.High]);
 	});
 
+	it("forwards distinct provider session id and prompt cache key to the stream", async () => {
+		const mock = createMockModel({ responses: [{ content: ["ok"] }] });
+		const agent = new Agent({
+			initialState: { model: mock.model, messages: [] },
+			streamFn: mock.stream,
+			sessionId: "provider-lineage",
+			promptCacheKey: "parent-cache",
+		});
+
+		await agent.prompt("run");
+
+		expect(mock.calls[0]?.options?.sessionId).toBe("provider-lineage");
+		expect(mock.calls[0]?.options?.promptCacheKey).toBe("parent-cache");
+	});
+
 	it("returns static metadata via the plain setter", () => {
 		const agent = new Agent();
 		expect(agent.metadata).toBeUndefined();
diff --git a/packages/ai/CHANGELOG.md b/packages/ai/CHANGELOG.md
index 83afdf41e..344d76b9a 100644
--- a/packages/ai/CHANGELOG.md
+++ b/packages/ai/CHANGELOG.md
@@ -2,6 +2,56 @@
 
 ## [Unreleased]
 
+## [15.10.1] - 2026-06-07
+
+### Breaking Changes
+
+- Removed the `onAuthError` option from stream request options and shifted auth retry handling to resolver-based `apiKey` behavior, requiring callers using custom auth-retry hooks to migrate
+
+### Added
+
+- Added `ApiKeyResolver` and `ApiKey` auth helpers, including `isApiKeyResolver`, `isAuthRetryableError`, `resolveApiKeyOnce`, and `withAuth`, and exported them from the package root
+- Added support for a function-valued `apiKey` in `SimpleStreamOptions` so a single stream request can refresh or rotate credentials during retry
+- Added `forceRefresh` credential option to `AuthStorage.getApiKey` and `rotateSessionCredential` support for session-level credential rotation after auth failures
+- Added `AuthStorage.resolver(provider, options)` method that builds an `ApiKeyResolver` implementing the a/b/c auth-retry policy directly on the storage instance
+
+### Changed
+
+- Changed gateway and stream auth flows to share the a/b/c retry policy, refreshing the same session credential first and then switching to a sibling credential on repeated auth failures
+
+### Fixed
+
+- Fixed streaming auth retries to handle `401` and usage-limit errors before replay-unsafe content is emitted, including failures surfaced only via `errorStatus`
+- Fixed tool argument validation to coerce singleton non-string values into arrays when the schema expects an array, preventing Anthropic-compatible models that emit `todo.ops` as an object from getting stuck in repeated validation-error loops. ([#2026](https://github.com/can1357/oh-my-pi/issues/2026))
+- Fixed streaming retries to buffer and suppress partial `start` events from failed auth attempts so only clean retried events are delivered
+- Fixed the HTTP 400 raw-request dumper (`appendRawHttpRequestDumpFor400`) littering the real `~/.omp/logs/http-400-requests` directory during tests. Provider suites exercise the 400 error path with mocked `fetch` responses, which the dumper could not distinguish from genuine failures; it now skips persistence under the Bun test runner (`isBunTestRuntime()`).
+- Fixed Anthropic Opus requests unnecessarily forcing `tool_choice.disable_parallel_tool_use`, allowing Claude Opus to use the provider's default parallel tool-calling behavior again.
+- Fixed parallel `function_call` items losing arguments against llama.cpp's OpenAI Responses endpoint (`/v1/responses`), where every call but the last finalized with `{}` and the agent rejected them with `path: Invalid input: expected string, received undefined`. llama.cpp's `to_json_oaicompat_resp` emits `output_item.added` with only `item.call_id` (no `item.id`, no `output_index`) while the matching `function_call_arguments.delta` carries `item_id: "fc_<call_id>"`. `processResponsesStream` now registers function-call and custom-tool-call items under `item.call_id` as a secondary lookup key (alongside `item.id`/`output_index`) so identifier-deviant hosts route deltas and done events to the right block. ([#2015](https://github.com/can1357/oh-my-pi/issues/2015))
+- Fixed `PI_REQ_DEBUG` response recording truncating the captured body when a streamed response was cancelled mid-flight. The response tee in `wrapResponse` could call `FileRequestDebugResponseLog.close()` from both the `cancel` callback and the resumed `pull` (which observes `done` once the source reader is cancelled); the second caller saw the handle already nulled and returned before the first caller's pending write flushed, so the `.res.log` lost the already-buffered chunk. `close()` now memoizes its flush-and-close promise so every caller awaits the same completion.
+
+## [15.10.0] - 2026-06-06
+
+### Added
+
+- Added a dependency-free `@oh-my-pi/pi-ai/effort` module exporting the `Effort` enum and `THINKING_EFFORTS`, split out of `model-thinking` so hot-path consumers can import the thinking levels without pulling in `model-thinking` and its provider-compat dependency graph. The package barrel still re-exports both names, so existing imports are unaffected.
+
+### Fixed
+
+- Fixed Antigravity usage provider emitting one bar per model instead of deduplicating by tier — a single account's 15+ model entries now collapse to one bar per tier, matching the shared-quota reality of the upstream API.
+- Fixed Antigravity usage reports missing `email` and `accountId` in metadata, so the `/usage` display and the deduplicator can associate reports with their credentials.
+- Fixed usage-report dedup ignoring `projectId` for Google Cloud providers, preventing duplicate credential entries from being recognized as the same account.
+
+- Fixed Cloud Code Assist (Antigravity / Gemini CLI) rejecting the `github` tool with HTTP 400 when the `pr` parameter schema contained `anyOf: [string, array]`. The CCA mixed-type combiner collapse picked the first non-null type (`string`) but indiscriminately copied type-specific keys from variant branches — `items` from the array variant leaked onto the string-typed result, producing `{type: "string", items: {...}}` which Google's API rejects as invalid. The collapse now filters merged variant fields against the winning type's allowed key set. ([#2002](https://github.com/can1357/oh-my-pi/pull/2002))
+- Fixed OpenAI Responses-family providers (Codex, OpenAI Responses, Azure Responses) rejecting requests with `400 No tool output found for function call …` after the user branched/navigated the session tree to a node that ends on a tool call (the tool-result child is dropped from the reconstructed history) or after a turn was aborted/crashed between the call streaming and its result persisting. The converters now synthesize a placeholder `function_call_output`/`custom_tool_call_output` immediately after any unpaired `function_call`/`custom_tool_call`, symmetric to the existing orphan-output repair, so the model still sees the call and can recover instead of the whole request 400ing.
+- Fixed Anthropic-compatible reasoning endpoints losing prior-turn reasoning on continuation requests when they emit unsigned `thinking` blocks. `convertAnthropicMessages` treated unknown endpoints as signature-enforcing and demoted unsigned reasoning to `type: "text"`, which destabilized tool-call argument serialization on the next turn — the upstream symptom behind the `args?.ops?.map is not a function` crash reported against the `todo` tool. Official `api.anthropic.com` keeps the conservative text fallback; non-official `anthropic-messages` reasoning models now replay unsigned reasoning as native `type: "thinking"` ([#2005](https://github.com/can1357/oh-my-pi/issues/2005)).
+
+## [15.9.67] - 2026-06-06
+
+### Fixed
+
+- Fixed llama.cpp/OpenAI Responses parallel tool calls losing arguments when `function_call_arguments.done` events omit `output_index` and `item_id`, by routing those identifierless final-argument events through the open function calls in item order. ([#1970](https://github.com/can1357/oh-my-pi/issues/1970))
+- Fixed local Ollama (`openai-responses`) turns failing with HTTP 400 `invalid reasoning value: "minimal"` when a discovered model ran with `minimal` (or `xhigh`) thinking. Ollama's OpenAI-compatible `reasoning.effort` only accepts `high|medium|low|max|none`, so discovered reasoning-capable Ollama models now carry a `compat.reasoningEffortMap` remapping `minimal → low` and `xhigh → max`; non-reasoning models are left untouched.
+
 ## [15.9.2] - 2026-06-05
 
 ### Added
diff --git a/packages/ai/package.json b/packages/ai/package.json
index 708eba2c3..7a108d2e6 100644
--- a/packages/ai/package.json
+++ b/packages/ai/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-ai",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
diff --git a/packages/ai/scripts/generate-models.ts b/packages/ai/scripts/generate-models.ts
index 6431ade6d..497fb7c25 100644
--- a/packages/ai/scripts/generate-models.ts
+++ b/packages/ai/scripts/generate-models.ts
@@ -32,6 +32,7 @@ import {
 	isFireworksKimiK2ModelId,
 	MODELS_DEV_PROVIDER_DESCRIPTORS,
 	mapModelsDevToModels,
+	stripFireworksDeepSeekThinkingToggle,
 	UNK_CONTEXT_WINDOW,
 	UNK_MAX_TOKENS,
 } from "../src/provider-models/openai-compat";
@@ -243,6 +244,18 @@ function applyFireworksKimiMaxTokensCap(models: readonly Model[]): Model[] {
 	});
 }
 
+/**
+ * Fireworks' DeepSeek V4 endpoint accepts the user's effort through
+ * `reasoning_effort` and rejects the DeepSeek-native binary `thinking` toggle
+ * when both are present. Strip stale reference metadata from generated fallbacks.
+ */
+function applyFireworksDeepSeekReasoningShape(models: readonly Model[]): Model[] {
+	return models.map(model => {
+		if (model.provider !== "fireworks" || model.api !== "openai-completions") return model;
+		return stripFireworksDeepSeekThinkingToggle(model, model.id);
+	});
+}
+
 const ANTIGRAVITY_ENDPOINT = "https://daily-cloudcode-pa.sandbox.googleapis.com";
 
 async function getOAuthAccessFromStorage(provider: OAuthProvider): Promise<OAuthAccess | null> {
@@ -416,6 +429,7 @@ async function generateModels() {
 	allModels = applyPremiumMultiplierOverrides(allModels);
 	allModels = applyCodexPricingFallback(allModels);
 	allModels = applyFireworksKimiMaxTokensCap(allModels);
+	allModels = applyFireworksDeepSeekReasoningShape(allModels);
 	applyGeneratedModelPolicies(allModels);
 	linkOpenAIPromotionTargets(allModels);
 
diff --git a/packages/ai/src/auth-gateway/server.ts b/packages/ai/src/auth-gateway/server.ts
index dad2aa394..83a3e4338 100644
--- a/packages/ai/src/auth-gateway/server.ts
+++ b/packages/ai/src/auth-gateway/server.ts
@@ -18,8 +18,9 @@
  *   POST /v1/responses                     → OpenAI Responses in/out
  */
 import { extractRetryHint, logger } from "@oh-my-pi/pi-utils";
+import type { ApiKeyResolver } from "../auth-retry";
 import type { AuthStorage } from "../auth-storage";
-import { Effort } from "../model-thinking";
+import { Effort } from "../effort";
 import * as anthropicMessages from "../providers/anthropic-messages-server";
 import * as openaiChat from "../providers/openai-chat-server";
 import * as openaiResponses from "../providers/openai-responses-server";
@@ -340,6 +341,60 @@ async function refreshGatewayApiKeyAfterAuthError(
 	return storage.getApiKey(provider, sessionId, { modelId: model.id, signal });
 }
 
+/**
+ * Build the {@link ApiKeyResolver} handed to `streamSimple` for a gateway
+ * request. Drives the central a/b/c auth-retry policy server-side:
+ *
+ * - initial resolve → the credential already resolved for this request.
+ * - step (b) `!lastChance` → force-refresh the SAME session-sticky credential
+ *   (a peer/broker may have rotated its token out from under our cached copy).
+ * - step (c) `lastChance` → {@link refreshGatewayApiKeyAfterAuthError} switches
+ *   to a sibling (usage-limit block vs credential invalidation by error class).
+ *
+ * `lastKey` tracks the most recent bearer so the switch step invalidates the
+ * credential that actually failed.
+ */
+function buildGatewayApiKeyResolver(
+	storage: AuthStorage,
+	model: Model<Api>,
+	sessionId: string,
+	initialKey: string,
+	requestSignal: AbortSignal,
+	format: string,
+	peer: string,
+): ApiKeyResolver {
+	let lastKey = initialKey;
+	return async ({ lastChance, error, signal }) => {
+		const sig = signal ?? requestSignal;
+		if (error === undefined) {
+			lastKey = initialKey;
+			return initialKey;
+		}
+		if (!lastChance) {
+			const refreshed = await storage.getApiKey(model.provider, sessionId, {
+				modelId: model.id,
+				signal: sig,
+				forceRefresh: true,
+			});
+			lastKey = refreshed ?? lastKey;
+			return refreshed;
+		}
+		const next = await refreshGatewayApiKeyAfterAuthError(
+			storage,
+			model,
+			sessionId,
+			model.provider,
+			lastKey,
+			error,
+			sig,
+			format,
+			peer,
+		);
+		lastKey = next ?? lastKey;
+		return next;
+	};
+}
+
 function clientClosedResponse(route: { module: FormatModule }): Response {
 	return route.module.formatError(499, "request_aborted", "client closed request");
 }
@@ -447,19 +502,15 @@ async function handleFormatEndpoint(
 	}
 
 	const streamOpts = buildStreamOptions(parsed, model.api, controller.signal);
-	streamOpts.apiKey = apiKey;
-	streamOpts.onAuthError = (provider, oldKey, error) =>
-		refreshGatewayApiKeyAfterAuthError(
-			bootOpts.storage,
-			model,
-			sessionId,
-			provider,
-			oldKey,
-			error,
-			controller.signal,
-			route.label,
-			peer,
-		);
+	streamOpts.apiKey = buildGatewayApiKeyResolver(
+		bootOpts.storage,
+		model,
+		sessionId,
+		apiKey,
+		controller.signal,
+		route.label,
+		peer,
+	);
 
 	logger.info("auth-gateway request", {
 		format: route.label,
@@ -604,18 +655,15 @@ async function handlePiNative(bootOpts: AuthGatewayBootOptions, req: Request, pe
 	// only inject server-controlled fields. The codex temperature/topP strip
 	// matches `buildStreamOptions` — Codex rejects them with a 400.
 	const streamOpts: SimpleStreamOptions = { ...parsed.options, apiKey, signal: controller.signal };
-	streamOpts.onAuthError = (provider, oldKey, error) =>
-		refreshGatewayApiKeyAfterAuthError(
-			bootOpts.storage,
-			model,
-			sessionId,
-			provider,
-			oldKey,
-			error,
-			controller.signal,
-			"pi-native",
-			peer,
-		);
+	streamOpts.apiKey = buildGatewayApiKeyResolver(
+		bootOpts.storage,
+		model,
+		sessionId,
+		apiKey,
+		controller.signal,
+		"pi-native",
+		peer,
+	);
 	if (model.api === "openai-codex-responses") {
 		delete streamOpts.temperature;
 		delete streamOpts.topP;
diff --git a/packages/ai/src/auth-gateway/types.ts b/packages/ai/src/auth-gateway/types.ts
index 0390759c0..bdb563e3b 100644
--- a/packages/ai/src/auth-gateway/types.ts
+++ b/packages/ai/src/auth-gateway/types.ts
@@ -1,4 +1,4 @@
-import type { Effort } from "../model-thinking";
+import type { Effort } from "../effort";
 import type {
 	AssistantMessage,
 	AssistantMessageEventStream,
diff --git a/packages/ai/src/auth-retry.ts b/packages/ai/src/auth-retry.ts
new file mode 100644
index 000000000..e53567f01
--- /dev/null
+++ b/packages/ai/src/auth-retry.ts
@@ -0,0 +1,141 @@
+import { extractHttpStatusFromError } from "@oh-my-pi/pi-utils";
+import { isUsageLimitError } from "./rate-limit-utils";
+
+/**
+ * Context passed to an {@link ApiKeyResolver} on each resolution attempt.
+ *
+ * The `error`/`lastChance` pair drives the central a/b/c retry policy shared by
+ * the streaming ({@link streamSimple}) and non-streaming ({@link withAuth})
+ * drivers:
+ * - `error === undefined` → **initial resolve** (no force-refresh; cheap, may
+ *   return a locally-cached not-yet-expired token).
+ * - `error !== undefined && !lastChance` → **step (b): refresh the SAME
+ *   account** (force a token re-mint / await an in-flight broker refresh).
+ * - `error !== undefined && lastChance` → **step (c): switch account**
+ *   (invalidate/usage-limit the current credential and rotate to a sibling).
+ *
+ * The resolver returns the bearer to send, or `undefined` to stop retrying and
+ * surface the last error to the caller.
+ */
+export interface ApiKeyResolveContext {
+	/** True on the final retry step — the resolver should rotate to a sibling credential. */
+	lastChance: boolean;
+	/** The auth error that triggered this re-resolution, or `undefined` on the initial resolve. */
+	error: unknown;
+	/** Caller cancel signal, threaded into any credential refresh / rotation work. */
+	signal?: AbortSignal;
+}
+
+/**
+ * Resolves the API key to send for a request, retried through the a/b/c policy
+ * described on {@link ApiKeyResolveContext}.
+ */
+export type ApiKeyResolver = (ctx: ApiKeyResolveContext) => Promise<string | undefined> | string | undefined;
+
+/** A static bearer string, or a {@link ApiKeyResolver} that mints/rotates one. */
+export type ApiKey = string | ApiKeyResolver;
+
+/** Narrows {@link ApiKey} to its resolver form. */
+export function isApiKeyResolver(key: ApiKey | undefined): key is ApiKeyResolver {
+	return typeof key === "function";
+}
+
+/**
+ * Performs the initial resolve of an {@link ApiKey} (`error: undefined`,
+ * `lastChance: false`). Static keys pass through unchanged.
+ */
+export async function resolveApiKeyOnce(key: ApiKey | undefined, signal?: AbortSignal): Promise<string | undefined> {
+	if (key === undefined) return undefined;
+	if (isApiKeyResolver(key)) return (await key({ lastChance: false, error: undefined, signal })) || undefined;
+	return key;
+}
+
+/**
+ * Classifies whether an error should trigger a credential refresh/rotation
+ * retry: a hard `401`, or a rotatable usage-limit ("usage_limit_reached",
+ * Codex's "you have hit your ChatGPT usage limit", etc.).
+ */
+export function isAuthRetryableError(error: unknown): boolean {
+	if (extractHttpStatusFromError(error) === 401) return true;
+	const message = error instanceof Error ? error.message : typeof error === "string" ? error : undefined;
+	if (!message) return false;
+	if (extractHttpStatusFromError({ message }) === 401) return true;
+	return isUsageLimitError(message);
+}
+
+/**
+ * The ordered `lastChance` values for the retry steps after the initial
+ * attempt fails: `false` → step (b) refresh-same, `true` → step (c) switch.
+ * Shared by {@link withAuth} and the streaming retry driver so both run the
+ * same policy.
+ */
+export const AUTH_RETRY_STEPS: readonly boolean[] = [false, true];
+
+/** Resolve a single retry step, swallowing resolver failures into `undefined`. */
+export async function resolveRetryKey(
+	resolver: ApiKeyResolver,
+	lastChance: boolean,
+	error: unknown,
+	signal?: AbortSignal,
+): Promise<string | undefined> {
+	try {
+		return (await resolver({ lastChance, error, signal })) || undefined;
+	} catch {
+		return undefined;
+	}
+}
+
+/**
+ * Runs an auth-protected operation through the central a/b/c retry policy.
+ *
+ * - A static string key (or any non-resolver) → a single `attempt` with no
+ *   retry (identical to the legacy static-key path).
+ * - A resolver → initial `attempt`, then on a retryable auth error up to two
+ *   more attempts (refresh-same, then switch). A step is skipped when the
+ *   resolver returns the same key it just tried or `undefined`; non-auth errors
+ *   propagate immediately.
+ *
+ * Used by non-streaming consumers (image generation, web search, completion
+ * helpers). The streaming driver in `stream.ts` implements the same policy with
+ * its replay-safe buffering machinery.
+ */
+export async function withAuth<T>(
+	key: ApiKey | undefined,
+	attempt: (key: string) => Promise<T>,
+	opts?: { isAuthError?: (error: unknown) => boolean; signal?: AbortSignal; missingKeyMessage?: string },
+): Promise<T> {
+	const isAuthError = opts?.isAuthError ?? isAuthRetryableError;
+	const missingKey = (): Error => new Error(opts?.missingKeyMessage ?? "No API key available");
+
+	if (!isApiKeyResolver(key)) {
+		if (key === undefined) throw missingKey();
+		return attempt(key);
+	}
+
+	const resolver = key;
+	const signal = opts?.signal;
+	let lastKey = await resolveRetryKey(resolver, false, undefined, signal);
+	if (lastKey === undefined) throw missingKey();
+
+	let lastError: unknown;
+	try {
+		return await attempt(lastKey);
+	} catch (error) {
+		if (!isAuthError(error)) throw error;
+		lastError = error;
+	}
+
+	for (let i = 0; i < AUTH_RETRY_STEPS.length; i++) {
+		const nextKey = await resolveRetryKey(resolver, AUTH_RETRY_STEPS[i]!, lastError, signal);
+		if (nextKey === undefined || nextKey === lastKey) continue;
+		lastKey = nextKey;
+		try {
+			return await attempt(nextKey);
+		} catch (error) {
+			if (!isAuthError(error)) throw error;
+			lastError = error;
+		}
+	}
+
+	throw lastError;
+}
diff --git a/packages/ai/src/auth-storage.ts b/packages/ai/src/auth-storage.ts
index f03f7b810..b76c2bbd6 100644
--- a/packages/ai/src/auth-storage.ts
+++ b/packages/ai/src/auth-storage.ts
@@ -11,6 +11,8 @@ import { Database, type Statement } from "bun:sqlite";
 import * as fs from "node:fs/promises";
 import * as path from "node:path";
 import { getAgentDbPath, logger } from "@oh-my-pi/pi-utils";
+import type { ApiKeyResolver } from "./auth-retry";
+import { isUsageLimitError } from "./rate-limit-utils";
 import { getEnvApiKey } from "./stream";
 import type { Provider } from "./types";
 import type {
@@ -36,6 +38,8 @@ import { loginOpenAICodexDevice } from "./utils/oauth/openai-codex";
 import type { OAuthController, OAuthCredentials, OAuthProvider, OAuthProviderId } from "./utils/oauth/types";
 import { loginXiaomi, loginXiaomiTokenPlan } from "./utils/oauth/xiaomi";
 
+const USAGE_RANKING_METRIC_EPSILON = 1e-9;
+
 // ─────────────────────────────────────────────────────────────────────────────
 // Credential Types
 // ─────────────────────────────────────────────────────────────────────────────
@@ -544,6 +548,13 @@ type AuthApiKeyOptions = {
 	 * stranding the caller for `timeoutMs * (maxRetries + 1)`.
 	 */
 	signal?: AbortSignal;
+	/**
+	 * Force a re-mint of the session-preferred OAuth credential's access token,
+	 * bypassing the not-yet-expired short-circuit. Powers step (b) of the
+	 * auth-retry policy ("refresh the SAME account") so a locally-cached token
+	 * that a peer/broker rotated out from under us is replaced before retrying.
+	 */
+	forceRefresh?: boolean;
 };
 type OAuthResolutionResult = { apiKey: string; credential: OAuthCredential };
 
@@ -606,6 +617,14 @@ function hasOpenAICodexProPlan(report: UsageReport | null): boolean {
 	return getUsagePlanType(report)?.includes("pro") === true;
 }
 
+function compareUsageRankingMetric(left: number, right: number): number {
+	if (left === right) return 0;
+	if (!Number.isFinite(left) || !Number.isFinite(right)) return left < right ? -1 : 1;
+	const delta = left - right;
+	const tolerance = Math.max(USAGE_RANKING_METRIC_EPSILON, Math.max(Math.abs(left), Math.abs(right)) * 0.000001);
+	return Math.abs(delta) <= tolerance ? 0 : delta;
+}
+
 function resolveDefaultUsageProvider(provider: Provider): UsageProvider | undefined {
 	return DEFAULT_USAGE_PROVIDER_MAP.get(provider);
 }
@@ -2288,6 +2307,16 @@ export class AuthStorage {
 		return undefined;
 	}
 
+	#getUsageReportScopeProjectId(report: UsageReport): string | undefined {
+		const ids = new Set<string>();
+		for (const limit of report.limits) {
+			const projectId = limit.scope.projectId?.trim();
+			if (projectId) ids.add(projectId);
+		}
+		if (ids.size === 1) return [...ids][0];
+		return undefined;
+	}
+
 	#getUsageReportIdentifiers(report: UsageReport): string[] {
 		const identifiers: string[] = [];
 		const email = this.#getUsageReportMetadataValue(report, "email");
@@ -2295,6 +2324,11 @@ export class AuthStorage {
 		if (report.provider === "openai-codex" || report.provider === "anthropic") {
 			return identifiers.map(identifier => `${report.provider}:${identifier.toLowerCase()}`);
 		}
+		const projectId =
+			this.#getUsageReportMetadataValue(report, "projectId") ?? this.#getUsageReportScopeProjectId(report);
+		// Only add project as a fallback when no email is available — two users
+		// with different emails on the same GCP project must not merge.
+		if (projectId && !email) identifiers.push(`project:${projectId}`);
 		const accountId = this.#getUsageReportMetadataValue(report, "accountId");
 		if (accountId) identifiers.push(`account:${accountId}`);
 		const account = this.#getUsageReportMetadataValue(report, "account");
@@ -2781,12 +2815,14 @@ export class AuthStorage {
 			return left.planPriority - right.planPriority;
 		}
 		if (left.hasPriorityBoost !== right.hasPriorityBoost) return left.hasPriorityBoost ? -1 : 1;
-		if (left.secondaryDrainRate !== right.secondaryDrainRate) {
-			return left.secondaryDrainRate - right.secondaryDrainRate;
-		}
-		if (left.secondaryUsed !== right.secondaryUsed) return left.secondaryUsed - right.secondaryUsed;
-		if (left.primaryDrainRate !== right.primaryDrainRate) return left.primaryDrainRate - right.primaryDrainRate;
-		if (left.primaryUsed !== right.primaryUsed) return left.primaryUsed - right.primaryUsed;
+		let metric = compareUsageRankingMetric(left.secondaryDrainRate, right.secondaryDrainRate);
+		if (metric !== 0) return metric;
+		metric = compareUsageRankingMetric(left.secondaryUsed, right.secondaryUsed);
+		if (metric !== 0) return metric;
+		metric = compareUsageRankingMetric(left.primaryDrainRate, right.primaryDrainRate);
+		if (metric !== 0) return metric;
+		metric = compareUsageRankingMetric(left.primaryUsed, right.primaryUsed);
+		if (metric !== 0) return metric;
 		return 0;
 	}
 
@@ -3019,19 +3055,38 @@ export class AuthStorage {
 				candidates.unshift(preferred);
 			}
 		}
+		// Step (b) of the auth-retry policy: when `forceRefresh` is set, re-mint
+		// the session-preferred credential (or the first candidate when no
+		// session preference exists yet) even if its cached token still looks
+		// valid — a peer/broker may have rotated it out from under us.
+		const forceRefreshIndex = options?.forceRefresh
+			? (sessionPreferredIndex ?? candidates[0]?.selection.index)
+			: undefined;
 		await Promise.all(
 			candidates.map(async candidate => {
-				if (Date.now() + OAUTH_REFRESH_SKEW_MS < candidate.selection.credential.expires) return;
+				const force = forceRefreshIndex !== undefined && candidate.selection.index === forceRefreshIndex;
+				if (!force && Date.now() + OAUTH_REFRESH_SKEW_MS < candidate.selection.credential.expires) return;
 				const latestCredential = this.#getCredentialsForProvider(provider)[candidate.selection.index];
-				if (latestCredential?.type === "oauth" && Date.now() + OAUTH_REFRESH_SKEW_MS < latestCredential.expires) {
+				if (
+					!force &&
+					latestCredential?.type === "oauth" &&
+					Date.now() + OAUTH_REFRESH_SKEW_MS < latestCredential.expires
+				) {
 					candidate.selection.credential = latestCredential;
 					return;
 				}
 				try {
 					const credentialId = this.#getStoredCredentials(provider)[candidate.selection.index]?.id;
+					// Hand #refreshOAuthCredential a stale clone (expires:0) so its
+					// not-yet-expired short-circuit doesn't suppress the forced
+					// re-mint; an in-flight peer refresh is still awaited via the
+					// per-credential single-flight.
+					const refreshTarget = force
+						? { ...candidate.selection.credential, expires: 0 }
+						: candidate.selection.credential;
 					const refreshedCredentials = await this.#refreshOAuthCredential(
 						provider,
-						candidate.selection.credential,
+						refreshTarget,
 						credentialId,
 						options?.signal,
 					);
@@ -3643,6 +3698,90 @@ export class AuthStorage {
 		return true;
 	}
 
+	/**
+	 * Rotate away from the session's current credential after a retryable auth
+	 * error — step (c) of the auth-retry policy. Stateless: looks up the
+	 * session-sticky credential (no API-key matching needed), applies the
+	 * storage action for the error class, then clears the sticky so the next
+	 * {@link AuthStorage.getApiKey} for this session picks a sibling.
+	 *
+	 * - usage-limit / account-rate-limit error → {@link AuthStorage.markUsageLimitReached}
+	 *   (temporary block via its own backoff — default plus server usage-report
+	 *   reset; sticky left intact so the next resolve re-ranks around the block).
+	 * - otherwise (hard 401 / auth failure) → mark the credential suspect (or
+	 *   reload when no broker hook is wired) and block it, then drop the sticky.
+	 *
+	 * Returns whether another usable credential of the same type remains.
+	 */
+	async rotateSessionCredential(
+		provider: string,
+		sessionId: string | undefined,
+		options?: { error?: unknown; signal?: AbortSignal },
+	): Promise<boolean> {
+		const sessionCredential = this.#getSessionCredential(provider, sessionId);
+		if (!sessionCredential) return false;
+
+		const error = options?.error;
+		const message = error instanceof Error ? error.message : typeof error === "string" ? error : undefined;
+		if (message && isUsageLimitError(message)) {
+			return this.markUsageLimitReached(provider, sessionId, { signal: options?.signal });
+		}
+
+		const providerKey = this.#getProviderTypeKey(provider, sessionCredential.type);
+		// Snapshot sibling availability before mutating so a soft-deleting
+		// suspect hook can't reindex the answer out from under us.
+		const hasSibling = this.#getCredentialsForProvider(provider).some(
+			(credential, index) =>
+				credential.type === sessionCredential.type &&
+				index !== sessionCredential.index &&
+				!this.#isCredentialBlocked(providerKey, index),
+		);
+		const target = this.#getStoredCredentials(provider)[sessionCredential.index];
+		this.#clearSessionCredential(provider, sessionId);
+		this.#markCredentialBlocked(providerKey, sessionCredential.index, Date.now() + AuthStorage.#defaultBackoffMs);
+
+		if (target) {
+			const markSuspect = this.#store.markCredentialSuspect?.bind(this.#store);
+			if (markSuspect) {
+				await markSuspect(target.id, { signal: options?.signal });
+			} else {
+				await this.reload();
+			}
+			const latestRows = this.#store.listAuthCredentials(provider);
+			this.#setStoredCredentials(
+				provider,
+				latestRows.map(row => ({ id: row.id, credential: row.credential })),
+			);
+		}
+
+		return hasSibling;
+	}
+
+	/**
+	 * Build an {@link ApiKeyResolver} backed by this storage, implementing the
+	 * central a/b/c auth-retry policy:
+	 *
+	 * - initial (`error: undefined`) → resolve the session credential.
+	 * - step (b) `!lastChance` → force-refresh the SAME session-sticky credential.
+	 * - step (c) `lastChance` → rotate to a sibling credential, then re-resolve.
+	 *
+	 * Used by web-search providers and other consumers that hold an AuthStorage
+	 * directly (no ModelRegistry in scope).
+	 */
+	resolver(provider: string, options?: { sessionId?: string; baseUrl?: string; modelId?: string }): ApiKeyResolver {
+		const { sessionId, baseUrl, modelId } = options ?? {};
+		return async ({ lastChance, error, signal }) => {
+			if (error === undefined) {
+				return this.getApiKey(provider, sessionId, { baseUrl, modelId, signal });
+			}
+			if (lastChance) {
+				await this.rotateSessionCredential(provider, sessionId, { error, signal });
+				return this.getApiKey(provider, sessionId, { baseUrl, modelId, signal });
+			}
+			return this.getApiKey(provider, sessionId, { baseUrl, modelId, forceRefresh: true, signal });
+		};
+	}
+
 	// ─── Auth Broker integration ────────────────────────────────────────────
 
 	/**
@@ -3932,6 +4071,8 @@ function resolveProviderCredentialIdentityKey(provider: string, identifiers: str
 	const accountIdentifier = identifiers.find(identifier => identifier.startsWith("account:"));
 	if (accountIdentifier) return accountIdentifier;
 	if (emailIdentifier) return emailIdentifier;
+	const projectIdentifier = identifiers.find(identifier => identifier.startsWith("project:"));
+	if (projectIdentifier) return projectIdentifier;
 	return null;
 }
 
@@ -3967,6 +4108,8 @@ function extractOAuthCredentialIdentifiers(credential: OAuthCredential): string[
 	if (accountId) identifiers.add(`account:${accountId}`);
 	const email = normalizeStoredEmail(credential.email);
 	if (email) identifiers.add(`email:${email}`);
+	const projectId = normalizeStoredAccountId(credential.projectId);
+	if (projectId) identifiers.add(`project:${projectId}`);
 	const accessIdentifiers = extractOAuthTokenIdentifiers(credential.access) ?? [];
 	for (const identifier of accessIdentifiers) {
 		identifiers.add(identifier);
diff --git a/packages/ai/src/effort.ts b/packages/ai/src/effort.ts
new file mode 100644
index 000000000..831a13ede
--- /dev/null
+++ b/packages/ai/src/effort.ts
@@ -0,0 +1,16 @@
+/** User-facing thinking levels, ordered least to most intensive. */
+export const enum Effort {
+	Minimal = "minimal",
+	Low = "low",
+	Medium = "medium",
+	High = "high",
+	XHigh = "xhigh",
+}
+
+export const THINKING_EFFORTS: readonly Effort[] = [
+	Effort.Minimal,
+	Effort.Low,
+	Effort.Medium,
+	Effort.High,
+	Effort.XHigh,
+];
diff --git a/packages/ai/src/index.ts b/packages/ai/src/index.ts
index ecdce180e..a9d3c724a 100644
--- a/packages/ai/src/index.ts
+++ b/packages/ai/src/index.ts
@@ -3,7 +3,9 @@ export * from "./api-registry";
 export * from "./auth-broker";
 export { type AuthGatewayBootOptions, type ModelResolver, startAuthGateway } from "./auth-gateway/server";
 export * from "./auth-gateway/types";
+export * from "./auth-retry";
 export * from "./auth-storage";
+export * from "./effort";
 export * from "./model-cache";
 export * from "./model-manager";
 export * from "./model-thinking";
diff --git a/packages/ai/src/model-thinking.ts b/packages/ai/src/model-thinking.ts
index 75e930315..3758c6d6c 100644
--- a/packages/ai/src/model-thinking.ts
+++ b/packages/ai/src/model-thinking.ts
@@ -1,23 +1,7 @@
+import { Effort, THINKING_EFFORTS } from "./effort";
 import { resolveOpenAICompat } from "./providers/openai-completions-compat";
 import type { Api, Model as ApiModel, ThinkingConfig } from "./types";
 
-/** User-facing thinking levels, ordered least to most intensive. */
-export const enum Effort {
-	Minimal = "minimal",
-	Low = "low",
-	Medium = "medium",
-	High = "high",
-	XHigh = "xhigh",
-}
-
-export const THINKING_EFFORTS: readonly Effort[] = [
-	Effort.Minimal,
-	Effort.Low,
-	Effort.Medium,
-	Effort.High,
-	Effort.XHigh,
-];
-
 const DEFAULT_REASONING_EFFORTS: readonly Effort[] = [Effort.Minimal, Effort.Low, Effort.Medium, Effort.High];
 const DEFAULT_REASONING_EFFORTS_WITH_XHIGH: readonly Effort[] = [
 	Effort.Minimal,
@@ -379,18 +363,6 @@ export function supportsMidConversationSystemMessages(modelId: string): boolean
 	return parsed.kind === "opus" && semverGte(parsed.version, "4.8");
 }
 
-/**
- * Claude Opus 4.8 must emit at most one tool call per turn: the Anthropic
- * Messages provider sends `tool_choice.disable_parallel_tool_use = true` for
- * this model. Scoped to exactly 4.8 — earlier and later Opus versions keep
- * Anthropic's default parallel tool-calling.
- */
-export function disablesParallelToolUse(modelId: string): boolean {
-	const parsed = parseAnthropicModel(getCanonicalModelId(modelId));
-	if (!parsed) return false;
-	return parsed.kind === "opus" && semverEqual(parsed.version, "4.8");
-}
-
 function anthropicModelHasRealXHighEffort<TApi extends Api>(model: ApiModel<TApi>): boolean {
 	if (model.api !== "anthropic-messages") return false;
 	const parsedModel = parseKnownModel(model.id);
diff --git a/packages/ai/src/models.json b/packages/ai/src/models.json
index e98a047fc..d3cd30bf1 100644
--- a/packages/ai/src/models.json
+++ b/packages/ai/src/models.json
@@ -1932,6 +1932,75 @@
 				"maxLevel": "high"
 			}
 		},
+		"openai.gpt-5.4": {
+			"id": "openai.gpt-5.4",
+			"name": "GPT-5.4",
+			"api": "bedrock-converse-stream",
+			"provider": "amazon-bedrock",
+			"baseUrl": "https://bedrock-runtime.us-east-1.amazonaws.com",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 2.75,
+				"output": 16.5,
+				"cacheRead": 0.275,
+				"cacheWrite": 0
+			},
+			"contextWindow": 272000,
+			"maxTokens": 128000,
+			"thinking": {
+				"mode": "budget",
+				"minLevel": "low",
+				"maxLevel": "xhigh"
+			}
+		},
+		"openai.gpt-5.5": {
+			"id": "openai.gpt-5.5",
+			"name": "GPT-5.5",
+			"api": "bedrock-converse-stream",
+			"provider": "amazon-bedrock",
+			"baseUrl": "https://bedrock-runtime.us-east-1.amazonaws.com",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 5.5,
+				"output": 33,
+				"cacheRead": 0.55,
+				"cacheWrite": 0
+			},
+			"contextWindow": 272000,
+			"maxTokens": 128000,
+			"thinking": {
+				"mode": "budget",
+				"minLevel": "low",
+				"maxLevel": "xhigh"
+			}
+		},
+		"openai.gpt-oss-120b": {
+			"id": "openai.gpt-oss-120b",
+			"name": "gpt-oss-120b",
+			"api": "bedrock-converse-stream",
+			"provider": "amazon-bedrock",
+			"baseUrl": "https://bedrock-runtime.us-east-1.amazonaws.com",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.15,
+				"output": 0.6,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 128000,
+			"maxTokens": 16384
+		},
 		"openai.gpt-oss-120b-1:0": {
 			"id": "openai.gpt-oss-120b-1:0",
 			"name": "gpt-oss-120b",
@@ -1951,6 +2020,25 @@
 			"contextWindow": 128000,
 			"maxTokens": 16384
 		},
+		"openai.gpt-oss-20b": {
+			"id": "openai.gpt-oss-20b",
+			"name": "gpt-oss-20b",
+			"api": "bedrock-converse-stream",
+			"provider": "amazon-bedrock",
+			"baseUrl": "https://bedrock-runtime.us-east-1.amazonaws.com",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.07,
+				"output": 0.3,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 128000,
+			"maxTokens": 16384
+		},
 		"openai.gpt-oss-20b-1:0": {
 			"id": "openai.gpt-oss-20b-1:0",
 			"name": "gpt-oss-20b",
@@ -5364,11 +5452,6 @@
 				},
 				"maxTokensField": "max_tokens",
 				"supportsToolChoice": false,
-				"extraBody": {
-					"thinking": {
-						"type": "enabled"
-					}
-				},
 				"reasoningContentField": "reasoning_content",
 				"requiresReasoningContentForToolCalls": true,
 				"requiresAssistantContentForToolCalls": true
@@ -11197,7 +11280,7 @@
 		},
 		"deepseek/deepseek-r1-distill-llama-70b": {
 			"id": "deepseek/deepseek-r1-distill-llama-70b",
-			"name": "DeepSeek: R1 Distill Llama 70B",
+			"name": "DeepSeek: R1 Distill Llama 70B (retires Jun 11)",
 			"api": "openai-completions",
 			"provider": "kilo",
 			"baseUrl": "https://api.kilo.ai/api/gateway",
@@ -12648,7 +12731,7 @@
 		},
 		"meta-llama/llama-3-70b-instruct": {
 			"id": "meta-llama/llama-3-70b-instruct",
-			"name": "Meta: Llama 3 70B Instruct",
+			"name": "Meta: Llama 3 70B Instruct (retires Jun 19)",
 			"api": "openai-completions",
 			"provider": "kilo",
 			"baseUrl": "https://api.kilo.ai/api/gateway",
@@ -13187,6 +13270,25 @@
 		},
 		"minimax/minimax-m3": {
 			"id": "minimax/minimax-m3",
+			"name": "MiniMax: MiniMax M3 (new)",
+			"api": "openai-completions",
+			"provider": "kilo",
+			"baseUrl": "https://api.kilo.ai/api/gateway",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
+		"minimax/minimax-m3:discounted": {
+			"id": "minimax/minimax-m3:discounted",
 			"name": "MiniMax: MiniMax M3 (50% off through 2026-06-07)",
 			"api": "openai-completions",
 			"provider": "kilo",
@@ -14262,6 +14364,63 @@
 			"contextWindow": 222222,
 			"maxTokens": 8888
 		},
+		"nvidia/nemotron-3-ultra-550b-a55b": {
+			"id": "nvidia/nemotron-3-ultra-550b-a55b",
+			"name": "NVIDIA: Nemotron 3 Ultra",
+			"api": "openai-completions",
+			"provider": "kilo",
+			"baseUrl": "https://api.kilo.ai/api/gateway",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
+		"nvidia/nemotron-3-ultra-550b-a55b:free": {
+			"id": "nvidia/nemotron-3-ultra-550b-a55b:free",
+			"name": "NVIDIA: Nemotron 3 Ultra (free)",
+			"api": "openai-completions",
+			"provider": "kilo",
+			"baseUrl": "https://api.kilo.ai/api/gateway",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
+		"nvidia/nemotron-3.5-content-safety:free": {
+			"id": "nvidia/nemotron-3.5-content-safety:free",
+			"name": "NVIDIA: Nemotron 3.5 Content Safety (free)",
+			"api": "openai-completions",
+			"provider": "kilo",
+			"baseUrl": "https://api.kilo.ai/api/gateway",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
 		"nvidia/nemotron-nano-12b-v2-vl": {
 			"id": "nvidia/nemotron-nano-12b-v2-vl",
 			"name": "NVIDIA: Nemotron Nano 12B 2 VL",
@@ -14283,7 +14442,7 @@
 		},
 		"nvidia/nemotron-nano-9b-v2": {
 			"id": "nvidia/nemotron-nano-9b-v2",
-			"name": "NVIDIA: Nemotron Nano 9B V2",
+			"name": "NVIDIA: Nemotron Nano 9B V2 (retires Jun 11)",
 			"api": "openai-completions",
 			"provider": "kilo",
 			"baseUrl": "https://api.kilo.ai/api/gateway",
@@ -16371,7 +16530,7 @@
 		},
 		"qwen/qwen3-30b-a3b": {
 			"id": "qwen/qwen3-30b-a3b",
-			"name": "Qwen: Qwen3 30B A3B (retires Jun 5)",
+			"name": "Qwen: Qwen3 30B A3B",
 			"api": "openai-completions",
 			"provider": "kilo",
 			"baseUrl": "https://api.kilo.ai/api/gateway",
@@ -19993,24 +20152,19 @@
 			"api": "openai-completions",
 			"provider": "mistral",
 			"baseUrl": "https://api.mistral.ai/v1",
-			"reasoning": true,
+			"reasoning": false,
 			"input": [
 				"text",
 				"image"
 			],
 			"cost": {
-				"input": 1.5,
-				"output": 7.5,
+				"input": 0.4,
+				"output": 2,
 				"cacheRead": 0,
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262144,
-			"thinking": {
-				"mode": "effort",
-				"minLevel": "minimal",
-				"maxLevel": "xhigh"
-			}
+			"maxTokens": 262144
 		},
 		"mistral-nemo": {
 			"id": "mistral-nemo",
@@ -25743,6 +25897,82 @@
 			"contextWindow": 222222,
 			"maxTokens": 8888
 		},
+		"linkup-research-high": {
+			"id": "linkup-research-high",
+			"name": "linkup-research-high",
+			"api": "openai-completions",
+			"provider": "nanogpt",
+			"baseUrl": "https://nano-gpt.com/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
+		"linkup-research-low": {
+			"id": "linkup-research-low",
+			"name": "linkup-research-low",
+			"api": "openai-completions",
+			"provider": "nanogpt",
+			"baseUrl": "https://nano-gpt.com/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
+		"linkup-research-medium": {
+			"id": "linkup-research-medium",
+			"name": "linkup-research-medium",
+			"api": "openai-completions",
+			"provider": "nanogpt",
+			"baseUrl": "https://nano-gpt.com/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
+		"linkup-research-xhigh": {
+			"id": "linkup-research-xhigh",
+			"name": "linkup-research-xhigh",
+			"api": "openai-completions",
+			"provider": "nanogpt",
+			"baseUrl": "https://nano-gpt.com/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888
+		},
 		"liquid/lfm-2-24b-a2b": {
 			"id": "liquid/lfm-2-24b-a2b",
 			"name": "liquid/lfm-2-24b-a2b",
@@ -28343,6 +28573,30 @@
 				"maxLevel": "xhigh"
 			}
 		},
+		"nvidia/nemotron-3-ultra-550b-a55b": {
+			"id": "nvidia/nemotron-3-ultra-550b-a55b",
+			"name": "nvidia/nemotron-3-ultra-550b-a55b",
+			"api": "openai-completions",
+			"provider": "nanogpt",
+			"baseUrl": "https://nano-gpt.com/api/v1",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
 		"nvidia/nvidia-nemotron-nano-9b-v2": {
 			"id": "nvidia/nvidia-nemotron-nano-9b-v2",
 			"name": "nvidia-nemotron-nano-9b-v2",
@@ -38214,7 +38468,7 @@
 				"cacheRead": 0.05,
 				"cacheWrite": 0.625
 			},
-			"contextWindow": 262144,
+			"contextWindow": 1000000,
 			"maxTokens": 65536,
 			"thinking": {
 				"mode": "effort",
@@ -38263,7 +38517,7 @@
 				"cacheRead": 0.04,
 				"cacheWrite": 0.5
 			},
-			"contextWindow": 262144,
+			"contextWindow": 1000000,
 			"maxTokens": 65536,
 			"thinking": {
 				"mode": "budget",
@@ -39625,6 +39879,30 @@
 				"maxLevel": "xhigh"
 			}
 		},
+		"nemotron-3-ultra-free": {
+			"id": "nemotron-3-ultra-free",
+			"name": "Nemotron 3 Ultra Free",
+			"api": "openai-completions",
+			"provider": "opencode-zen",
+			"baseUrl": "https://opencode.ai/zen/v1",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1000000,
+			"maxTokens": 128000,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
 		"qwen3.5-plus": {
 			"id": "qwen3.5-plus",
 			"name": "Qwen3.5 Plus",
@@ -41767,12 +42045,12 @@
 			],
 			"cost": {
 				"input": 0.12,
-				"output": 0.37,
-				"cacheRead": 0.019999999499999997,
+				"output": 0.36,
+				"cacheRead": 0.09,
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 16384,
+			"maxTokens": 8192,
 			"thinking": {
 				"mode": "effort",
 				"minLevel": "minimal",
@@ -42135,7 +42413,7 @@
 			],
 			"cost": {
 				"input": 0.02,
-				"output": 0.049999999999999996,
+				"output": 0.03,
 				"cacheRead": 0,
 				"cacheWrite": 0
 			},
@@ -42360,7 +42638,7 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 204800,
-			"maxTokens": 131072,
+			"maxTokens": 196608,
 			"thinking": {
 				"mode": "effort",
 				"minLevel": "minimal",
@@ -43295,6 +43573,54 @@
 				"maxLevel": "high"
 			}
 		},
+		"nvidia/nemotron-3-ultra-550b-a55b": {
+			"id": "nvidia/nemotron-3-ultra-550b-a55b",
+			"name": "NVIDIA: Nemotron 3 Ultra",
+			"api": "openai-completions",
+			"provider": "openrouter",
+			"baseUrl": "https://openrouter.ai/api/v1",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.5,
+				"output": 2.5,
+				"cacheRead": 0.15,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1000000,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "high"
+			}
+		},
+		"nvidia/nemotron-3-ultra-550b-a55b:free": {
+			"id": "nvidia/nemotron-3-ultra-550b-a55b:free",
+			"name": "NVIDIA: Nemotron 3 Ultra (free)",
+			"api": "openai-completions",
+			"provider": "openrouter",
+			"baseUrl": "https://openrouter.ai/api/v1",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1000000,
+			"maxTokens": 65536,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "high"
+			}
+		},
 		"nvidia/nemotron-nano-12b-v2-vl:free": {
 			"id": "nvidia/nemotron-nano-12b-v2-vl:free",
 			"name": "NVIDIA: Nemotron Nano 12B 2 VL (free)",
@@ -45244,7 +45570,7 @@
 				"cacheWrite": 0
 			},
 			"contextWindow": 131072,
-			"maxTokens": 20000,
+			"maxTokens": 16384,
 			"thinking": {
 				"mode": "effort",
 				"minLevel": "minimal",
@@ -46000,13 +46326,13 @@
 				"image"
 			],
 			"cost": {
-				"input": 0.29,
-				"output": 3.1999999999999997,
+				"input": 0.28900000000000003,
+				"output": 2.4,
 				"cacheRead": 0.15,
 				"cacheWrite": 0
 			},
 			"contextWindow": 262144,
-			"maxTokens": 262140,
+			"maxTokens": 131072,
 			"thinking": {
 				"mode": "effort",
 				"minLevel": "minimal",
@@ -49634,6 +49960,28 @@
 				"supportsUsageInStreaming": false
 			}
 		},
+		"nvidia-nemotron-3-ultra-550b-a55b": {
+			"id": "nvidia-nemotron-3-ultra-550b-a55b",
+			"name": "nvidia-nemotron-3-ultra-550b-a55b",
+			"api": "openai-completions",
+			"provider": "venice",
+			"baseUrl": "https://api.venice.ai/api/v1",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0,
+				"output": 0,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 222222,
+			"maxTokens": 8888,
+			"compat": {
+				"supportsUsageInStreaming": false
+			}
+		},
 		"nvidia-nemotron-cascade-2-30b-a3b": {
 			"id": "nvidia-nemotron-cascade-2-30b-a3b",
 			"name": "nvidia-nemotron-cascade-2-30b-a3b",
@@ -52853,6 +53201,30 @@
 				"maxLevel": "xhigh"
 			}
 		},
+		"nvidia/nemotron-3-ultra-550b-a55b": {
+			"id": "nvidia/nemotron-3-ultra-550b-a55b",
+			"name": "Nemotron 3 Ultra",
+			"api": "anthropic-messages",
+			"provider": "vercel-ai-gateway",
+			"baseUrl": "https://ai-gateway.vercel.sh",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.6,
+				"output": 2.4,
+				"cacheRead": 0.12,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1000000,
+			"maxTokens": 65000,
+			"thinking": {
+				"mode": "budget",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
 		"nvidia/nemotron-nano-12b-v2-vl": {
 			"id": "nvidia/nemotron-nano-12b-v2-vl",
 			"name": "Nvidia Nemotron Nano 12B V2 VL",
@@ -57962,7 +58334,7 @@
 			"cost": {
 				"input": 0.5,
 				"output": 1.5,
-				"cacheRead": 0,
+				"cacheRead": 0.05,
 				"cacheWrite": 0
 			},
 			"contextWindow": 256000,
diff --git a/packages/ai/src/provider-models/ollama.ts b/packages/ai/src/provider-models/ollama.ts
index c71fbf617..ed539d924 100644
--- a/packages/ai/src/provider-models/ollama.ts
+++ b/packages/ai/src/provider-models/ollama.ts
@@ -1,6 +1,6 @@
 import { fetchWithRetry } from "@oh-my-pi/pi-utils";
+import { Effort } from "../effort";
 import type { ModelManagerOptions } from "../model-manager";
-import { Effort } from "../model-thinking";
 import type { ThinkingConfig } from "../types";
 import { createBundledReferenceMap, createReferenceResolver } from "./bundled-references";
 
diff --git a/packages/ai/src/provider-models/openai-compat.ts b/packages/ai/src/provider-models/openai-compat.ts
index 7004521cf..c529fb135 100644
--- a/packages/ai/src/provider-models/openai-compat.ts
+++ b/packages/ai/src/provider-models/openai-compat.ts
@@ -1,5 +1,5 @@
+import { Effort } from "../effort";
 import type { ModelManagerOptions } from "../model-manager";
-import { Effort } from "../model-thinking";
 import { getBundledModels } from "../models";
 import type { Api, Model, Provider, ThinkingConfig } from "../types";
 import { isAnthropicOAuthToken, isRecord, toBoolean, toNumber, toPositiveNumber } from "../utils";
@@ -242,6 +242,22 @@ async function fetchOllamaNativeModels(
 const OLLAMA_FALLBACK_CONTEXT_WINDOW = 128_000;
 /** Cap max output tokens at a value that matches OMP's other openai-responses defaults. */
 const OLLAMA_DEFAULT_MAX_TOKENS = 8192;
+/**
+ * Ollama's OpenAI-compatible `reasoning.effort` only accepts
+ * `high|medium|low|max|none`; passing OMP's `minimal`/`xhigh` levels verbatim
+ * makes the server reject the turn with HTTP 400 `invalid reasoning value`.
+ * Map the two unsupported levels onto the closest accepted ones (`low`/`max`).
+ */
+const OLLAMA_REASONING_EFFORT_MAP = { minimal: "low", xhigh: "max" } as const;
+
+/** Stamp the Ollama reasoning-effort map onto a reasoning-capable model. */
+function applyOllamaReasoningCompat(model: Model<"openai-responses">): void {
+	if (!model.reasoning) return;
+	model.compat = {
+		...model.compat,
+		reasoningEffortMap: { ...OLLAMA_REASONING_EFFORT_MAP, ...model.compat?.reasoningEffortMap },
+	};
+}
 
 interface OllamaResolvedMetadata {
 	contextWindow: number;
@@ -955,6 +971,29 @@ export function clampFireworksKimiMaxTokens(modelId: string, candidate: number):
 	return isFireworksKimiK2ModelId(modelId) ? Math.min(candidate, FIREWORKS_KIMI_MAX_TOKENS) : candidate;
 }
 
+/**
+ * Fireworks DeepSeek V4 accepts effort via `reasoning_effort` but rejects the
+ * DeepSeek-native binary `thinking` toggle when both are present.
+ */
+export function stripFireworksDeepSeekThinkingToggle(
+	model: Model<"openai-completions">,
+	publicModelId: string,
+): Model<"openai-completions"> {
+	if (!publicModelId.startsWith("deepseek-v4")) return model;
+	const compat = model.compat;
+	if (!compat?.extraBody || !("thinking" in compat.extraBody)) return model;
+
+	const extraBody = { ...compat.extraBody };
+	delete extraBody.thinking;
+	if (Object.keys(extraBody).length > 0) {
+		return { ...model, compat: { ...compat, extraBody } };
+	}
+
+	const nextCompat = { ...compat };
+	delete nextCompat.extraBody;
+	return { ...model, compat: nextCompat };
+}
+
 export interface FireworksModelManagerConfig {
 	apiKey?: string;
 	baseUrl?: string;
@@ -1024,7 +1063,10 @@ export function fireworksModelManagerOptions(
 					mapModel: (entry, defaults) => {
 						const publicModelId = toFireworksPublicModelId(defaults.id);
 						const reference = modelsDevReferences.get(publicModelId) ?? bundledReferences(publicModelId);
-						const model = mapWithBundledReference(entry, defaults, reference);
+						const model = stripFireworksDeepSeekThinkingToggle(
+							mapWithBundledReference(entry, defaults, reference),
+							publicModelId,
+						);
 						return {
 							...model,
 							id: publicModelId,
@@ -1357,12 +1399,14 @@ export function ollamaModelManagerOptions(config?: OllamaModelManagerConfig): Mo
 						if (metadata.input) {
 							model.input = metadata.input;
 						}
+						applyOllamaReasoningCompat(model);
 					}),
 				);
 				return openAiCompatible;
 			}
 			const nativeFallback = await fetchOllamaNativeModels(baseUrl, resolveMetadata);
 			if (nativeFallback && nativeFallback.length > 0) {
+				for (const model of nativeFallback) applyOllamaReasoningCompat(model);
 				return nativeFallback;
 			}
 			return openAiCompatible;
diff --git a/packages/ai/src/providers/amazon-bedrock.ts b/packages/ai/src/providers/amazon-bedrock.ts
index 7f161f32c..49b1523aa 100644
--- a/packages/ai/src/providers/amazon-bedrock.ts
+++ b/packages/ai/src/providers/amazon-bedrock.ts
@@ -8,7 +8,7 @@
  */
 
 import { $env, $flag, extractHttpStatusFromError, fetchWithRetry } from "@oh-my-pi/pi-utils";
-import type { Effort } from "../model-thinking";
+import type { Effort } from "../effort";
 import { mapEffortToAnthropicAdaptiveEffort, requireSupportedEffort } from "../model-thinking";
 import { calculateCost } from "../models";
 import type {
diff --git a/packages/ai/src/providers/anthropic.ts b/packages/ai/src/providers/anthropic.ts
index 52c0c901e..c8e9f3c35 100644
--- a/packages/ai/src/providers/anthropic.ts
+++ b/packages/ai/src/providers/anthropic.ts
@@ -13,7 +13,6 @@ import {
 	readSseEvents,
 } from "@oh-my-pi/pi-utils";
 import {
-	disablesParallelToolUse,
 	hasOpus47ApiRestrictions,
 	mapEffortToAnthropicAdaptiveEffort,
 	supportsMidConversationSystemMessages,
@@ -1778,16 +1777,12 @@ type SystemBlockOptions = {
 	cacheControl?: AnthropicCacheControl;
 };
 
-function withGlobalCacheScope(cacheControl: AnthropicCacheControl): AnthropicCacheControl {
-	return { ...cacheControl, scope: "global" };
-}
-
 function applyClaudeCodeSystemCache(
 	blocks: AnthropicSystemBlock[],
 	cacheControl: AnthropicCacheControl | undefined,
 ): number {
 	if (!cacheControl || blocks.length <= 2) return 0;
-	blocks[2] = { ...blocks[2], cache_control: withGlobalCacheScope(cacheControl) };
+	blocks[2] = { ...blocks[2], cache_control: cacheControl };
 	if (blocks.length === 3) return 1;
 	const lastIndex = blocks.length - 1;
 	blocks[lastIndex] = { ...blocks[lastIndex], cache_control: cacheControl };
@@ -2375,21 +2370,6 @@ function buildParams(
 		}
 	}
 
-	// Claude Opus 4.8 must emit at most one tool call per turn. Force
-	// `disable_parallel_tool_use` onto the outgoing tool_choice (synthesizing an
-	// `auto` choice when none is set). Gated on tools being present: Anthropic
-	// rejects `tool_choice` without `tools`, and parallelism is moot otherwise.
-	// `none` rejects the field, so leave it untouched. A fresh object is built
-	// rather than mutated so the caller's `options.toolChoice` is never aliased.
-	if (disablesParallelToolUse(model.id) && params.tools && params.tools.length > 0) {
-		const current = params.tool_choice;
-		if (!current) {
-			params.tool_choice = { type: "auto", disable_parallel_tool_use: true };
-		} else if (current.type !== "none") {
-			params.tool_choice = { ...current, disable_parallel_tool_use: true };
-		}
-	}
-
 	const shouldInjectClaudeCodeInstruction = isOAuthToken && !model.id.startsWith("claude-3-5-haiku");
 	const firstUserMessageText = shouldInjectClaudeCodeInstruction
 		? extractClaudeCodeFirstUserMessageText(context.messages)
@@ -2431,22 +2411,31 @@ function isZaiAnthropicEndpoint(model: Model<"anthropic-messages">): boolean {
 }
 
 /**
- * Returns true for providers whose Anthropic-compatible endpoints do NOT
- * implement signature-based thinking-chain integrity (DeepSeek, Z.AI, etc.).
- * For these providers, unsigned thinking blocks must be preserved as
- * `type: "thinking"` instead of being degraded to text.
+ * Returns true when unsigned `thinking` blocks from prior assistant turns should
+ * be replayed as Anthropic-native thinking instead of demoted to text.
+ *
+ * Official Anthropic (matched via `isAnthropicApiBaseUrl`, which intentionally
+ * treats a missing baseUrl as official since `resolveAnthropicBaseUrl` routes
+ * it to `https://api.anthropic.com`) enforces signature-based thinking-chain
+ * integrity, so unsigned blocks must remain text there. Anthropic-compatible
+ * reasoning endpoints commonly emit unsigned thinking blocks while still
+ * expecting them back as `type: "thinking"` on continuation; demoting them
+ * loses the model's reasoning chain and can destabilize the next tool-call
+ * arguments (#2005). Known non-signing hosts are also preserved for
+ * compatibility.
  */
-function isNonSigningAnthropicEndpoint(model: Model<"anthropic-messages">): boolean {
-	// Known non-signing providers
+function shouldReplayUnsignedThinking(model: Model<"anthropic-messages">): boolean {
 	if (model.provider === "zai" || model.provider === "deepseek") return true;
 	const baseUrl = model.baseUrl;
-	if (!baseUrl) return false;
-	try {
-		const hostname = new URL(baseUrl).hostname.toLowerCase();
-		return hostname === "api.deepseek.com" || hostname.endsWith(".deepseek.com");
-	} catch {
-		return false;
+	if (baseUrl) {
+		try {
+			const hostname = new URL(baseUrl).hostname.toLowerCase();
+			if (hostname === "api.deepseek.com" || hostname.endsWith(".deepseek.com")) return true;
+		} catch {
+			// Fall through to the protocol-level reasoning rule below.
+		}
 	}
+	return model.reasoning && !isAnthropicApiBaseUrl(baseUrl);
 }
 
 function buildToolResultBlock(model: Model<"anthropic-messages">, msg: ToolResultMessage): ContentBlockParam {
@@ -2537,7 +2526,7 @@ export function convertAnthropicMessages(
 					}
 					if (block.thinking.trim().length === 0) continue;
 					if (!block.thinkingSignature || block.thinkingSignature.trim().length === 0) {
-						if (isNonSigningAnthropicEndpoint(model)) {
+						if (shouldReplayUnsignedThinking(model)) {
 							blocks.push({
 								type: "thinking",
 								thinking: block.thinking.toWellFormed(),
diff --git a/packages/ai/src/providers/azure-openai-responses.ts b/packages/ai/src/providers/azure-openai-responses.ts
index 04027d02a..26b3f0a16 100644
--- a/packages/ai/src/providers/azure-openai-responses.ts
+++ b/packages/ai/src/providers/azure-openai-responses.ts
@@ -40,6 +40,7 @@ import {
 	isOpenAIResponsesProgressEvent,
 	normalizeResponsesToolCallIdForTransform,
 	processResponsesStream,
+	repairOrphanResponsesToolCalls,
 } from "./openai-responses-shared";
 import { transformMessages } from "./transform-messages";
 
@@ -347,7 +348,7 @@ function convertMessages(
 		msgIndex++;
 	}
 
-	return messages;
+	return repairOrphanResponsesToolCalls(messages);
 }
 
 function convertTools(tools: Tool[]): OpenAITool[] {
diff --git a/packages/ai/src/providers/gitlab-duo.ts b/packages/ai/src/providers/gitlab-duo.ts
index 50c43b372..382ce56c9 100644
--- a/packages/ai/src/providers/gitlab-duo.ts
+++ b/packages/ai/src/providers/gitlab-duo.ts
@@ -234,7 +234,8 @@ export function streamGitLabDuo(
 
 	(async () => {
 		try {
-			if (!options?.apiKey) {
+			const apiKey = typeof options?.apiKey === "string" ? options.apiKey : undefined;
+			if (!apiKey || !options) {
 				throw new Error("Missing GitLab access token. Run /login gitlab-duo or set GITLAB_TOKEN.");
 			}
 
@@ -243,7 +244,7 @@ export function streamGitLabDuo(
 				throw new Error(`Unsupported GitLab Duo model: ${model.id}`);
 			}
 
-			const directAccess = await getDirectAccessToken(options.apiKey, options.fetch);
+			const directAccess = await getDirectAccessToken(apiKey, options.fetch);
 			const headers = {
 				...directAccess.headers,
 				...options.headers,
diff --git a/packages/ai/src/providers/openai-anthropic-shim.ts b/packages/ai/src/providers/openai-anthropic-shim.ts
index f2dad56d7..a4f9b8fac 100644
--- a/packages/ai/src/providers/openai-anthropic-shim.ts
+++ b/packages/ai/src/providers/openai-anthropic-shim.ts
@@ -44,6 +44,9 @@ export function streamOpenAIAnthropicShim(
 ): AssistantMessageEventStream {
 	const stream = new AssistantMessageEventStream();
 	const format = options?.format ?? config.defaultFormat;
+	// The resolver form of `apiKey` is resolved upstream in `streamSimple`;
+	// this shim only ever receives a static bearer string.
+	const apiKey = typeof options?.apiKey === "string" ? options.apiKey : undefined;
 
 	(async () => {
 		try {
@@ -74,7 +77,7 @@ export function streamOpenAIAnthropicShim(
 					: undefined;
 
 				const innerStream = streamAnthropic(anthropicModel, context, {
-					apiKey: options?.apiKey,
+					apiKey,
 					temperature: options?.temperature,
 					topP: options?.topP,
 					topK: options?.topK,
@@ -103,7 +106,7 @@ export function streamOpenAIAnthropicShim(
 
 				const reasoningEffort = options?.reasoning;
 				const innerStream = streamOpenAICompletions(openaiModel, context, {
-					apiKey: options?.apiKey,
+					apiKey,
 					temperature: options?.temperature,
 					topP: options?.topP,
 					topK: options?.topK,
diff --git a/packages/ai/src/providers/openai-codex/request-transformer.ts b/packages/ai/src/providers/openai-codex/request-transformer.ts
index f271bbc81..91abe9dc5 100644
--- a/packages/ai/src/providers/openai-codex/request-transformer.ts
+++ b/packages/ai/src/providers/openai-codex/request-transformer.ts
@@ -1,4 +1,4 @@
-import type { Effort } from "../../model-thinking";
+import type { Effort } from "../../effort";
 import { requireSupportedEffort } from "../../model-thinking";
 import type { Api, Model } from "../../types";
 
@@ -76,6 +76,83 @@ function filterInput(input: InputItem[] | undefined): InputItem[] | undefined {
 		});
 }
 
+const CODEX_ORPHAN_OUTPUT_LIMIT = 16_000;
+/** Placeholder output for a tool call whose result never landed in the input. */
+const CODEX_INTERRUPTED_TOOL_OUTPUT =
+	"[No tool output recorded: the tool call was interrupted before it produced a result.]";
+
+function orphanFunctionOutputToMessage(item: InputItem, callId: string): InputItem {
+	const itemRecord = item as unknown as Record<string, unknown>;
+	const toolName = typeof itemRecord.name === "string" ? itemRecord.name : "tool";
+	let text = "";
+	try {
+		const output = itemRecord.output;
+		text = typeof output === "string" ? output : JSON.stringify(output);
+	} catch {
+		text = String(itemRecord.output ?? "");
+	}
+	if (text.length > CODEX_ORPHAN_OUTPUT_LIMIT) {
+		text = `${text.slice(0, CODEX_ORPHAN_OUTPUT_LIMIT)}\n...[truncated]`;
+	}
+	return {
+		type: "message",
+		role: "assistant",
+		content: `[Previous ${toolName} result; call_id=${callId}]: ${text}`,
+	} as InputItem;
+}
+
+/**
+ * Repair both halves of unpaired tool exchanges so the Responses input grammar
+ * stays valid — the API rejects either orphan with a 400:
+ *
+ * - `function_call_output` with no matching `function_call` → folded into an
+ *   assistant message (`400 No tool call found for function call output …`).
+ *   Regression of #472 / #1351.
+ * - `function_call` / `custom_tool_call` with no matching `*_output` → a
+ *   placeholder output is synthesized immediately after the call
+ *   (`400 No tool output found for function call …`). Hit when the user
+ *   branches/navigates the session tree to a node that ends on a tool call (the
+ *   tool-result child is dropped from the reconstructed history) or when a turn
+ *   is aborted/crashes after the call streamed but before its result persisted.
+ */
+function repairToolCallPairs(input: InputItem[]): InputItem[] {
+	const callIds = new Set<string>();
+	const outputCallIds = new Set<string>();
+	for (const item of input) {
+		const callId = typeof item.call_id === "string" ? item.call_id : undefined;
+		if (callId === undefined) continue;
+		if (item.type === "function_call" || item.type === "custom_tool_call") callIds.add(callId);
+		else if (item.type === "function_call_output" || item.type === "custom_tool_call_output") {
+			outputCallIds.add(callId);
+		}
+	}
+
+	const repaired: InputItem[] = [];
+	for (const item of input) {
+		const callId = typeof item.call_id === "string" ? item.call_id : undefined;
+
+		if (item.type === "function_call_output" && callId !== undefined && !callIds.has(callId)) {
+			repaired.push(orphanFunctionOutputToMessage(item, callId));
+			continue;
+		}
+
+		repaired.push(item);
+
+		if (
+			(item.type === "function_call" || item.type === "custom_tool_call") &&
+			callId !== undefined &&
+			!outputCallIds.has(callId)
+		) {
+			repaired.push({
+				type: item.type === "custom_tool_call" ? "custom_tool_call_output" : "function_call_output",
+				call_id: callId,
+				output: CODEX_INTERRUPTED_TOOL_OUTPUT,
+			} as InputItem);
+		}
+	}
+	return repaired;
+}
+
 export async function transformRequestBody(
 	body: RequestBody,
 	model: Model<Api>,
@@ -87,39 +164,8 @@ export async function transformRequestBody(
 
 	if (body.input && Array.isArray(body.input)) {
 		body.input = filterInput(body.input);
-
 		if (body.input) {
-			const functionCallIds = new Set(
-				body.input
-					.filter(item => item.type === "function_call" && typeof item.call_id === "string")
-					.map(item => item.call_id as string),
-			);
-
-			body.input = body.input.map(item => {
-				if (item.type === "function_call_output" && typeof item.call_id === "string") {
-					const callId = item.call_id as string;
-					if (!functionCallIds.has(callId)) {
-						const itemRecord = item as unknown as Record<string, unknown>;
-						const toolName = typeof itemRecord.name === "string" ? itemRecord.name : "tool";
-						let text = "";
-						try {
-							const output = itemRecord.output;
-							text = typeof output === "string" ? output : JSON.stringify(output);
-						} catch {
-							text = String(itemRecord.output ?? "");
-						}
-						if (text.length > 16000) {
-							text = `${text.slice(0, 16000)}\n...[truncated]`;
-						}
-						return {
-							type: "message",
-							role: "assistant",
-							content: `[Previous ${toolName} result; call_id=${callId}]: ${text}`,
-						} as InputItem;
-					}
-				}
-				return item;
-			});
+			body.input = repairToolCallPairs(body.input);
 		}
 	}
 
diff --git a/packages/ai/src/providers/openai-completions.ts b/packages/ai/src/providers/openai-completions.ts
index b8b5e591c..e58e4b8fd 100644
--- a/packages/ai/src/providers/openai-completions.ts
+++ b/packages/ai/src/providers/openai-completions.ts
@@ -10,7 +10,8 @@ import type {
 	ChatCompletionToolMessageParam,
 } from "openai/resources/chat/completions";
 import packageJson from "../../package.json" with { type: "json" };
-import { type Effort, getSupportedEfforts } from "../model-thinking";
+import type { Effort } from "../effort";
+import { getSupportedEfforts } from "../model-thinking";
 import { calculateCost } from "../models";
 import { getEnvApiKey } from "../stream";
 import {
@@ -1344,6 +1345,11 @@ function buildParams(
 
 	if (compat.extraBody) {
 		Object.assign(params, compat.extraBody);
+		if (model.provider === "fireworks" && params.reasoning_effort !== undefined) {
+			// Fireworks rejects simultaneous DeepSeek-style `thinking` toggles and
+			// OpenAI-style `reasoning_effort`; the effort field carries the user's level.
+			delete params.thinking;
+		}
 	}
 
 	return { params, toolStrictMode };
diff --git a/packages/ai/src/providers/openai-responses-shared.ts b/packages/ai/src/providers/openai-responses-shared.ts
index 135f92589..28532dcc8 100644
--- a/packages/ai/src/providers/openai-responses-shared.ts
+++ b/packages/ai/src/providers/openai-responses-shared.ts
@@ -212,6 +212,59 @@ export function repairOrphanResponsesToolOutputs(input: ResponseInput): Response
 	});
 }
 
+/** Placeholder output for a tool call whose result is absent from the input. */
+const ORPHAN_TOOL_CALL_PLACEHOLDER =
+	"[No tool output recorded: the tool call was interrupted before it produced a result.]";
+
+/**
+ * Synthesize a placeholder `function_call_output` / `custom_tool_call_output`
+ * for every `function_call` / `custom_tool_call` whose `call_id` has no matching
+ * output later in the same input. The Responses API rejects an unpaired call
+ * with `400 No tool output found for function call …`.
+ *
+ * Orphan calls surface when the user branches/navigates the session tree to a
+ * node that ends on a tool call (the tool-result child is excluded from the
+ * reconstructed history) or when a turn is aborted/crashes after the call
+ * streamed but before its result persisted. Dropping the call would erase the
+ * assistant's action; a placeholder output keeps the call visible so the model
+ * can recover (e.g. re-issue the call). Symmetric to
+ * {@link repairOrphanResponsesToolOutputs}.
+ */
+export function repairOrphanResponsesToolCalls(input: ResponseInput): ResponseInput {
+	const outputCallIds = new Set<string>();
+	for (const item of input) {
+		const t = (item as { type?: string }).type;
+		if (t !== "function_call_output" && t !== "custom_tool_call_output") continue;
+		const callId = (item as { call_id?: unknown }).call_id;
+		if (typeof callId === "string") outputCallIds.add(callId);
+	}
+	let hasOrphan = false;
+	for (const item of input) {
+		const t = (item as { type?: string }).type;
+		if (t !== "function_call" && t !== "custom_tool_call") continue;
+		const callId = (item as { call_id?: unknown }).call_id;
+		if (typeof callId === "string" && !outputCallIds.has(callId)) {
+			hasOrphan = true;
+			break;
+		}
+	}
+	if (!hasOrphan) return input;
+	const repaired: ResponseInput = [];
+	for (const item of input) {
+		repaired.push(item);
+		const t = (item as { type?: string }).type;
+		if (t !== "function_call" && t !== "custom_tool_call") continue;
+		const callId = (item as { call_id?: unknown }).call_id;
+		if (typeof callId !== "string" || outputCallIds.has(callId)) continue;
+		repaired.push({
+			type: t === "custom_tool_call" ? "custom_tool_call_output" : "function_call_output",
+			call_id: callId,
+			output: ORPHAN_TOOL_CALL_PLACEHOLDER,
+		} as ResponseInput[number]);
+	}
+	return repaired;
+}
+
 export function convertResponsesInputContent(
 	content: string | Array<TextContent | ImageContent>,
 	supportsImages: boolean,
@@ -395,7 +448,7 @@ export async function processResponsesStream<TApi extends Api>(
 	model: Model<TApi>,
 	options?: ProcessResponsesStreamOptions,
 ): Promise<void> {
-	type StreamingToolCallBlock = ToolCall & { partialJson: string; lastParseLen?: number };
+	type StreamingToolCallBlock = ToolCall & { partialJson: string; lastParseLen?: number; argumentsDone?: boolean };
 	interface StreamingItem {
 		item: ResponseReasoningItem | ResponseOutputMessage | ResponseFunctionToolCall | ResponseCustomToolCall;
 		block: ThinkingContent | TextContent | StreamingToolCallBlock;
@@ -406,17 +459,28 @@ export async function processResponsesStream<TApi extends Api>(
 	// see https://github.com/can1357/oh-my-pi/issues/1880 — llama.cpp emits parallel
 	// function_call deltas interleaved, and a singleton `current` reference would
 	// fold them into the wrong block and drop arguments on every call but the last.
+	//
+	// llama.cpp's `to_json_oaicompat_resp` (issue #2015) compounds this: `output_item.added`
+	// for function_call/custom_tool_call carries `item.call_id` but no `item.id` and no
+	// `output_index`, while the matching `function_call_arguments.delta` carries
+	// `item_id = "fc_<call_id>"`. Registering function-call items by `call_id` as a
+	// secondary key lets the delta lookup find the right block on hosts that emit one
+	// identifier but not the other.
 	const openItemsByOutputIndex = new Map<number, StreamingItem>();
 	const openItemsByItemId = new Map<string, StreamingItem>();
 	let lastOpenItem: StreamingItem | null = null;
+	const openItemsInOrder: StreamingItem[] = [];
 
 	const registerOpenItem = (
 		outputIndex: number | undefined,
 		itemId: string | undefined,
 		entry: StreamingItem,
+		alternateItemKey?: string,
 	): void => {
 		if (typeof outputIndex === "number") openItemsByOutputIndex.set(outputIndex, entry);
 		if (itemId) openItemsByItemId.set(itemId, entry);
+		if (alternateItemKey && alternateItemKey !== itemId) openItemsByItemId.set(alternateItemKey, entry);
+		openItemsInOrder.push(entry);
 		lastOpenItem = entry;
 	};
 	const lookupOpenItem = (event: { output_index?: number; item_id?: string }): StreamingItem | undefined => {
@@ -431,13 +495,37 @@ export async function processResponsesStream<TApi extends Api>(
 		// Fallback for tests / mock providers that omit identifiers on stream events.
 		return lastOpenItem ?? undefined;
 	};
+	const hasOpenItemKey = (event: { output_index?: number; item_id?: string }): boolean =>
+		typeof event.output_index === "number" || event.item_id !== undefined;
+	const lookupOpenFunctionCallItem = (event: {
+		output_index?: number;
+		item_id?: string;
+	}): StreamingItem | undefined => {
+		if (hasOpenItemKey(event)) return lookupOpenItem(event);
+		for (const candidate of openItemsInOrder) {
+			if (
+				candidate.item.type === "function_call" &&
+				candidate.block.type === "toolCall" &&
+				!candidate.block.argumentsDone
+			) {
+				return candidate;
+			}
+		}
+		return lastOpenItem?.item.type === "function_call" ? lastOpenItem : undefined;
+	};
 	const closeOpenItem = (
 		outputIndex: number | undefined,
 		itemId: string | undefined,
 		entry: StreamingItem | undefined,
+		alternateItemKey?: string,
 	): void => {
 		if (typeof outputIndex === "number") openItemsByOutputIndex.delete(outputIndex);
 		if (itemId) openItemsByItemId.delete(itemId);
+		if (alternateItemKey && alternateItemKey !== itemId) openItemsByItemId.delete(alternateItemKey);
+		if (entry) {
+			const index = openItemsInOrder.indexOf(entry);
+			if (index >= 0) openItemsInOrder.splice(index, 1);
+		}
 		if (entry && lastOpenItem === entry) lastOpenItem = null;
 	};
 	const contentIndexOf = (block: ThinkingContent | TextContent | StreamingToolCallBlock): number =>
@@ -473,7 +561,7 @@ export async function processResponsesStream<TApi extends Api>(
 					partialJson: item.arguments || "",
 				};
 				output.content.push(block);
-				registerOpenItem(event.output_index, item.id, { item, block });
+				registerOpenItem(event.output_index, item.id, { item, block }, item.call_id);
 				stream.push({ type: "toolcall_start", contentIndex: contentIndexOf(block), partial: output });
 			} else if (item.type === "custom_tool_call") {
 				const block: StreamingToolCallBlock = {
@@ -491,7 +579,7 @@ export async function processResponsesStream<TApi extends Api>(
 					partialJson: item.input ?? "",
 				};
 				output.content.push(block);
-				registerOpenItem(event.output_index, item.id, { item, block });
+				registerOpenItem(event.output_index, item.id, { item, block }, item.call_id);
 				stream.push({ type: "toolcall_start", contentIndex: contentIndexOf(block), partial: output });
 			}
 		} else if (event.type === "response.reasoning_summary_part.added") {
@@ -584,7 +672,7 @@ export async function processResponsesStream<TApi extends Api>(
 				}
 			}
 		} else if (event.type === "response.function_call_arguments.delta") {
-			const entry = lookupOpenItem(event);
+			const entry = lookupOpenFunctionCallItem(event);
 			if (entry?.item.type === "function_call" && entry.block.type === "toolCall") {
 				const block = entry.block;
 				block.partialJson += event.delta;
@@ -601,11 +689,12 @@ export async function processResponsesStream<TApi extends Api>(
 				});
 			}
 		} else if (event.type === "response.function_call_arguments.done") {
-			const entry = lookupOpenItem(event);
+			const entry = lookupOpenFunctionCallItem(event);
 			if (entry?.item.type === "function_call" && entry.block.type === "toolCall") {
 				const block = entry.block;
 				block.partialJson = event.arguments;
 				block.arguments = parseStreamingJson(block.partialJson);
+				block.argumentsDone = true;
 				delete (block as { partialJson?: string }).partialJson;
 				delete (block as { lastParseLen?: number }).lastParseLen;
 			}
@@ -631,7 +720,10 @@ export async function processResponsesStream<TApi extends Api>(
 		} else if (event.type === "response.output_item.done") {
 			const item = structuredCloneJSON(event.item);
 			options?.onOutputItemDone?.(item);
-			const entry = lookupOpenItem({ output_index: event.output_index, item_id: item.id });
+			const entry =
+				item.type === "function_call" || item.type === "custom_tool_call"
+					? lookupOpenItem({ output_index: event.output_index, item_id: item.id ?? item.call_id })
+					: lookupOpenItem({ output_index: event.output_index, item_id: item.id });
 			if (item.type === "reasoning") {
 				const thinking =
 					item.summary?.length > 0
@@ -668,9 +760,11 @@ export async function processResponsesStream<TApi extends Api>(
 				closeOpenItem(event.output_index, item.id, entry);
 			} else if (item.type === "function_call") {
 				const block = entry?.block.type === "toolCall" ? entry.block : undefined;
-				const args = block?.partialJson
-					? parseStreamingJson(block.partialJson)
-					: parseStreamingJson(item.arguments || "{}");
+				const args = block?.argumentsDone
+					? block.arguments
+					: block?.partialJson
+						? parseStreamingJson(block.partialJson)
+						: parseStreamingJson(item.arguments || "{}");
 				const toolCall: ToolCall = {
 					type: "toolCall",
 					id: encodeResponsesToolCallId(item.call_id, item.id),
@@ -685,9 +779,10 @@ export async function processResponsesStream<TApi extends Api>(
 					block.arguments = args;
 					delete (block as { partialJson?: string }).partialJson;
 					delete (block as { lastParseLen?: number }).lastParseLen;
+					delete (block as { argumentsDone?: boolean }).argumentsDone;
 				}
 				const contentIndex = block ? contentIndexOf(block) : output.content.length - 1;
-				closeOpenItem(event.output_index, item.id, entry);
+				closeOpenItem(event.output_index, item.id, entry, item.call_id);
 				stream.push({ type: "toolcall_end", contentIndex, toolCall, partial: output });
 			} else if (item.type === "custom_tool_call") {
 				const block = entry?.block.type === "toolCall" ? entry.block : undefined;
@@ -700,7 +795,7 @@ export async function processResponsesStream<TApi extends Api>(
 					customWireName: item.name,
 				};
 				const contentIndex = block ? contentIndexOf(block) : output.content.length - 1;
-				closeOpenItem(event.output_index, item.id, entry);
+				closeOpenItem(event.output_index, item.id, entry, item.call_id);
 				stream.push({ type: "toolcall_end", contentIndex, toolCall, partial: output });
 			}
 		} else if (event.type === "response.completed") {
diff --git a/packages/ai/src/providers/openai-responses.ts b/packages/ai/src/providers/openai-responses.ts
index ac1684b43..ed6de0481 100644
--- a/packages/ai/src/providers/openai-responses.ts
+++ b/packages/ai/src/providers/openai-responses.ts
@@ -62,6 +62,7 @@ import {
 	isOpenAIResponsesProgressEvent,
 	normalizeResponsesToolCallIdForTransform,
 	processResponsesStream,
+	repairOrphanResponsesToolCalls,
 	repairOrphanResponsesToolOutputs,
 } from "./openai-responses-shared";
 import { transformMessages } from "./transform-messages";
@@ -614,7 +615,7 @@ function convertConversationMessages(
 		msgIndex++;
 	}
 
-	return repairOrphanResponsesToolOutputs(messages);
+	return repairOrphanResponsesToolCalls(repairOrphanResponsesToolOutputs(messages));
 }
 
 /**
diff --git a/packages/ai/src/providers/pi-native-client.ts b/packages/ai/src/providers/pi-native-client.ts
index b5df79636..68f479d2a 100644
--- a/packages/ai/src/providers/pi-native-client.ts
+++ b/packages/ai/src/providers/pi-native-client.ts
@@ -149,7 +149,10 @@ export function streamPiNative<TApi extends Api>(
 		try {
 			const url = resolveStreamUrl(model as Model<Api>);
 			const fetchImpl = options?.fetch ?? globalThis.fetch;
-			const headers = buildHeaders(model as Model<Api>, options?.apiKey);
+			const headers = buildHeaders(
+				model as Model<Api>,
+				typeof options?.apiKey === "string" ? options.apiKey : undefined,
+			);
 			const body = JSON.stringify({
 				modelId: model.id,
 				context,
diff --git a/packages/ai/src/stream.ts b/packages/ai/src/stream.ts
index 437a49344..f592dcd89 100644
--- a/packages/ai/src/stream.ts
+++ b/packages/ai/src/stream.ts
@@ -3,7 +3,8 @@ import * as os from "node:os";
 import * as path from "node:path";
 import { $env, $pickenv, extractHttpStatusFromError } from "@oh-my-pi/pi-utils";
 import { getCustomApi } from "./api-registry";
-import type { Effort } from "./model-thinking";
+import { AUTH_RETRY_STEPS, isApiKeyResolver, resolveRetryKey } from "./auth-retry";
+import type { Effort } from "./effort";
 import {
 	mapEffortToAnthropicAdaptiveEffort,
 	mapEffortToGoogleThinkingLevel,
@@ -443,12 +444,15 @@ export function streamSimple<TApi extends Api>(
 	options?: SimpleStreamOptions,
 ): AssistantMessageEventStream {
 	const requestOptions = withRequestDebugFetch(options);
-	const retryApiKey = requestOptions?.onAuthError
-		? (requestOptions.apiKey ?? getEnvApiKey(model.provider))
-		: undefined;
-	if (retryApiKey) {
+	const apiKeyResolver = isApiKeyResolver(requestOptions?.apiKey) ? requestOptions.apiKey : undefined;
+	if (apiKeyResolver) {
 		const outer = new AssistantMessageEventStream();
-		const onAuthError = requestOptions!.onAuthError!;
+		const signal = requestOptions?.signal;
+		// One inner attempt against a resolved string key. When
+		// `captureAuthFailure` is set, a retryable auth error that arrives before
+		// any replay-unsafe event is buffered and returned (so the caller can
+		// retry with a fresh key) instead of surfaced. The terminal attempt
+		// clears the flag and emits whatever it gets.
 		const runAttempt = async (apiKey: string, captureAuthFailure: boolean): Promise<AuthRetryFailure | undefined> => {
 			const bufferedEvents: AssistantMessageEvent[] = [];
 			let emittedReplayUnsafeEvent = false;
@@ -458,7 +462,7 @@ export function streamSimple<TApi extends Api>(
 			};
 
 			try {
-				const inner = streamSimple(model, context, { ...requestOptions, apiKey, onAuthError: undefined });
+				const inner = streamSimple(model, context, { ...requestOptions, apiKey });
 				for await (const event of inner) {
 					if (!emittedReplayUnsafeEvent && event.type === "start") {
 						bufferedEvents.push(event);
@@ -510,19 +514,32 @@ export function streamSimple<TApi extends Api>(
 		};
 
 		void (async () => {
-			const failure = await runAttempt(retryApiKey, true);
-			if (!failure) return;
-			let nextKey: string | undefined;
+			let lastKey: string | undefined;
 			try {
-				nextKey = await onAuthError(model.provider, retryApiKey, failure.error);
+				lastKey = (await apiKeyResolver({ lastChance: false, error: undefined, signal })) || undefined;
 			} catch {
-				nextKey = undefined;
+				lastKey = undefined;
 			}
-			if (!nextKey || nextKey === retryApiKey) {
-				emitFailure(failure);
+			if (lastKey === undefined) {
+				outer.fail(new Error(`No API key for provider: ${model.provider}`));
 				return;
 			}
-			await runAttempt(nextKey, false);
+			let failure = await runAttempt(lastKey, true);
+			if (!failure) return;
+			// a/b/c policy: refresh the same account (lastChance=false), then
+			// switch to a sibling (lastChance=true). A step is skipped when the
+			// resolver yields the same key it just tried or `undefined`; the
+			// final step's attempt clears the capture flag so it emits directly.
+			for (let step = 0; step < AUTH_RETRY_STEPS.length; step++) {
+				const nextKey = await resolveRetryKey(apiKeyResolver, AUTH_RETRY_STEPS[step]!, failure.error, signal);
+				if (nextKey === undefined || nextKey === lastKey) continue;
+				lastKey = nextKey;
+				const isLastStep = step === AUTH_RETRY_STEPS.length - 1;
+				const next = await runAttempt(nextKey, !isLastStep);
+				if (!next) return;
+				failure = next;
+			}
+			emitFailure(failure);
 		})();
 		return outer;
 	}
@@ -553,7 +570,10 @@ export function streamSimple<TApi extends Api>(
 		return stream(model, context, providerOptions);
 	}
 
-	const apiKey = requestOptions?.apiKey || getEnvApiKey(model.provider);
+	// The resolver form is handled by the wrapper above; only a static string
+	// key reaches this point.
+	const apiKey =
+		(typeof requestOptions?.apiKey === "string" ? requestOptions.apiKey : undefined) || getEnvApiKey(model.provider);
 	if (!apiKey) {
 		throw new Error(`No API key for provider: ${model.provider}`);
 	}
@@ -724,7 +744,7 @@ function mapOptionsForApi<TApi extends Api>(
 		repetitionPenalty: options?.repetitionPenalty,
 		maxTokens: options?.maxTokens ?? model.maxTokens,
 		signal: options?.signal,
-		apiKey: apiKey || options?.apiKey,
+		apiKey: apiKey ?? (typeof options?.apiKey === "string" ? options.apiKey : undefined),
 		cacheRetention: options?.cacheRetention,
 		headers: options?.headers,
 		initiatorOverride: options?.initiatorOverride,
diff --git a/packages/ai/src/types.ts b/packages/ai/src/types.ts
index 9b03d99d8..566574be4 100644
--- a/packages/ai/src/types.ts
+++ b/packages/ai/src/types.ts
@@ -1,4 +1,5 @@
 import type { ZodType, z } from "zod/v4";
+import type { ApiKey } from "./auth-retry";
 import type { BedrockOptions } from "./providers/amazon-bedrock";
 import type { AnthropicOptions } from "./providers/anthropic";
 import type { AzureOpenAIResponsesOptions } from "./providers/azure-openai-responses";
@@ -151,7 +152,7 @@ export type KnownProvider =
 	| "lm-studio";
 export type Provider = KnownProvider | string;
 
-import type { Effort } from "./model-thinking";
+import type { Effort } from "./effort";
 
 /** Token budgets for each thinking level (token-based providers only) */
 export type ThinkingBudgets = { [key in Effort]?: number };
@@ -298,12 +299,6 @@ export interface StreamOptions {
 	maxTokens?: number;
 	signal?: AbortSignal;
 	apiKey?: string;
-	/**
-	 * Called when a provider returns 401 before any replay-unsafe assistant
-	 * event has been emitted. Returning a different key retries the provider
-	 * request once.
-	 */
-	onAuthError?: (provider: string, apiKey: string, error: unknown) => Promise<string | undefined>;
 	cacheRetention?: CacheRetention;
 	/**
 	 * Additional headers to include in provider requests.
@@ -415,7 +410,15 @@ export interface StreamOptions {
 }
 
 // Unified options with reasoning passed to streamSimple() and completeSimple()
-export interface SimpleStreamOptions extends StreamOptions {
+export interface SimpleStreamOptions extends Omit<StreamOptions, "apiKey"> {
+	/**
+	 * API key for the request: either a static bearer string, or an
+	 * {@link ApiKeyResolver} that mints/rotates the key across the central
+	 * a/b/c auth-retry policy. `streamSimple`/`completeSimple` resolve a
+	 * resolver to a string before per-provider dispatch, so providers only
+	 * ever see the resolved {@link StreamOptions.apiKey} string.
+	 */
+	apiKey?: ApiKey;
 	reasoning?: Effort;
 	/**
 	 * Force-disable reasoning for the request even when the model supports it.
diff --git a/packages/ai/src/usage/google-antigravity.ts b/packages/ai/src/usage/google-antigravity.ts
index 26963d7b5..54e2e11cf 100644
--- a/packages/ai/src/usage/google-antigravity.ts
+++ b/packages/ai/src/usage/google-antigravity.ts
@@ -148,10 +148,17 @@ async function fetchAntigravityUsage(params: UsageFetchParams, ctx: UsageFetchCo
 	}
 
 	const data = (await response.json()) as AntigravityUsageResponse;
-	const limits: UsageLimit[] = [];
+
+	// The API returns per-model quota entries, but quota is shared across
+	// models within the same tier. Deduplicate by (tier, windowId) so one
+	// account doesn't produce 15 redundant bars.
+	const deduped = new Map<
+		string,
+		{ amount: UsageAmount; window: UsageWindow | undefined; tier: string | undefined }
+	>();
 	let earliestReset: number | undefined;
 
-	for (const [modelId, modelInfo] of Object.entries(data.models ?? {})) {
+	for (const [_modelId, modelInfo] of Object.entries(data.models ?? {})) {
 		const quotaInfos = normalizeQuotaInfos(modelInfo);
 		for (const quotaInfo of quotaInfos) {
 			const amount = buildAmount(quotaInfo);
@@ -159,35 +166,81 @@ async function fetchAntigravityUsage(params: UsageFetchParams, ctx: UsageFetchCo
 			if (window?.resetsAt) {
 				earliestReset = earliestReset ? Math.min(earliestReset, window.resetsAt) : window.resetsAt;
 			}
-			const labelBase = modelInfo.displayName || modelId;
-			const label = quotaInfo.tier ? `${labelBase} (${quotaInfo.tier})` : labelBase;
-			const windowId = window?.id ?? "default";
-			limits.push({
-				id: `${modelId}:${quotaInfo.tier ?? "default"}:${windowId}`,
-				label,
-				scope: {
-					provider: params.provider,
-					accountId: credential.accountId,
-					projectId: credential.projectId,
-					modelId,
-					tier: quotaInfo.tier,
-					windowId,
-				},
-				window,
-				amount,
-				status: getUsageStatus(amount.remainingFraction),
-			});
+			const tier = (quotaInfo.tier ?? "default").toLowerCase();
+			// Use quotaInfo.windowId even when parseWindow returns undefined
+			// (no resetTime) — separate windows must not collapse to "default".
+			const windowId = quotaInfo.windowId ?? window?.id ?? "default";
+			const key = `${tier}|${windowId}`;
+			const existing = deduped.get(key);
+			if (!existing) {
+				deduped.set(key, { amount, window, tier: quotaInfo.tier });
+				continue;
+			}
+			// Merge: keep the entry with fraction data for the bar, but
+			// also keep any window with a reset time so "resets in…" survives.
+			const eFrac = existing.amount.remainingFraction;
+			const cFrac = amount.remainingFraction;
+			const eHasFrac = eFrac !== undefined;
+			const cHasFrac = cFrac !== undefined;
+
+			let bestAmount = existing.amount;
+			let bestWindow = existing.window?.resetsAt ? existing.window : (window ?? existing.window);
+			let bestTier = existing.tier ?? quotaInfo.tier;
+
+			if (!eHasFrac && cHasFrac) {
+				bestAmount = amount;
+				bestTier = quotaInfo.tier ?? existing.tier;
+			} else if (eHasFrac && cHasFrac && cFrac! < eFrac!) {
+				bestAmount = amount;
+				bestTier = quotaInfo.tier ?? existing.tier;
+			}
+			// Always merge in window with reset time if the current
+			// best doesn't have one.
+			if (!bestWindow?.resetsAt && window?.resetsAt) {
+				bestWindow = window;
+			}
+			deduped.set(key, { amount: bestAmount, window: bestWindow, tier: bestTier });
 		}
 	}
 
+	const limits: UsageLimit[] = [];
+	for (const [key, entry] of deduped) {
+		const [tier, windowId] = key.split("|") as [string, string];
+		const label = "Usage";
+		limits.push({
+			id: `${params.provider}:${tier}:${windowId}`,
+			label,
+			scope: {
+				provider: params.provider,
+				accountId: credential.accountId,
+				projectId: credential.projectId,
+				tier: entry.tier,
+				windowId,
+			},
+			window: entry.window,
+			amount: entry.amount,
+			status: getUsageStatus(entry.amount.remainingFraction),
+		});
+	}
+
+	limits.sort((a, b) => {
+		const aFraction = a.amount.remainingFraction ?? 1;
+		const bFraction = b.amount.remainingFraction ?? 1;
+		return aFraction - bFraction;
+	});
+
+	const metadata: UsageReport["metadata"] = {
+		endpoint: url,
+		projectId: credential.projectId,
+	};
+	if (credential.email) metadata.email = credential.email;
+	if (credential.accountId) metadata.accountId = credential.accountId;
+
 	const report: UsageReport = {
 		provider: params.provider,
 		fetchedAt: nowMs,
 		limits,
-		metadata: {
-			endpoint: url,
-			projectId: credential.projectId,
-		},
+		metadata,
 		raw: data,
 	};
 
diff --git a/packages/ai/src/utils/http-inspector.ts b/packages/ai/src/utils/http-inspector.ts
index 31a6ccb06..df738c3d8 100644
--- a/packages/ai/src/utils/http-inspector.ts
+++ b/packages/ai/src/utils/http-inspector.ts
@@ -1,5 +1,5 @@
 import * as path from "node:path";
-import { extractHttpStatusFromError, getLogsDir } from "@oh-my-pi/pi-utils";
+import { extractHttpStatusFromError, getLogsDir, isBunTestRuntime } from "@oh-my-pi/pi-utils";
 import { isCopilotTransientModelError } from "./retry.js";
 import { formatErrorMessageWithRetryAfter } from "./retry-after.js";
 
@@ -31,7 +31,9 @@ export async function appendRawHttpRequestDumpFor400(
 	error: unknown,
 	dump: RawHttpRequestDump | undefined,
 ): Promise<string> {
-	if (!dump || extractHttpStatusFromError(error) !== 400) {
+	// Never persist dumps under the test runner: providers exercise the 400 path
+	// with mocked fetch responses, which would otherwise litter the real ~/.omp logs.
+	if (!dump || isBunTestRuntime() || extractHttpStatusFromError(error) !== 400) {
 		return message;
 	}
 
diff --git a/packages/ai/src/utils/request-debug.ts b/packages/ai/src/utils/request-debug.ts
index 193f51be6..daa6831c4 100644
--- a/packages/ai/src/utils/request-debug.ts
+++ b/packages/ai/src/utils/request-debug.ts
@@ -170,6 +170,7 @@ class FileRequestDebugSession implements RequestDebugSession {
 class FileRequestDebugResponseLog implements RequestDebugResponseLog {
 	#handle: fs.FileHandle | undefined;
 	#pending: Promise<void> = Promise.resolve();
+	#closed: Promise<void> | undefined;
 
 	constructor(handle: fs.FileHandle) {
 		this.#handle = handle;
@@ -184,15 +185,19 @@ class FileRequestDebugResponseLog implements RequestDebugResponseLog {
 		});
 	}
 
-	async close(): Promise<void> {
+	close(): Promise<void> {
+		if (this.#closed) return this.#closed;
 		const handle = this.#handle;
-		if (!handle) return;
+		if (!handle) return Promise.resolve();
 		this.#handle = undefined;
-		try {
-			await this.#pending;
-		} finally {
-			await handle.close();
-		}
+		this.#closed = (async () => {
+			try {
+				await this.#pending;
+			} finally {
+				await handle.close();
+			}
+		})();
+		return this.#closed;
 	}
 }
 
diff --git a/packages/ai/src/utils/schema/fields.ts b/packages/ai/src/utils/schema/fields.ts
index 41e9aacf1..b25006248 100644
--- a/packages/ai/src/utils/schema/fields.ts
+++ b/packages/ai/src/utils/schema/fields.ts
@@ -154,6 +154,22 @@ export const CLOUD_CODE_ASSIST_TYPE_SPECIFIC_KEYS: Record<string, Record<string,
 	null: {},
 };
 
+/**
+ * Flat set of every type-specific key across all CCA types.
+ * Used to identify sibling keys that need filtering during mixed-type collapse.
+ */
+export const ALL_CCA_TYPE_SPECIFIC_KEYS: Record<string, true> = buildAllCcaTypeSpecificKeys();
+
+function buildAllCcaTypeSpecificKeys(): Record<string, true> {
+	const all: Record<string, true> = {};
+	for (const typeKeys of Object.values(CLOUD_CODE_ASSIST_TYPE_SPECIFIC_KEYS)) {
+		for (const key in typeKeys) {
+			all[key] = true;
+		}
+	}
+	return all;
+}
+
 /**
  * Cloud Code Assist shared schema keys allowed on any type.
  * Used alongside CLOUD_CODE_ASSIST_TYPE_SPECIFIC_KEYS for CCA combiner collapsing.
diff --git a/packages/ai/src/utils/schema/json-schema-validator.ts b/packages/ai/src/utils/schema/json-schema-validator.ts
index 93a69cf15..79a93052e 100644
--- a/packages/ai/src/utils/schema/json-schema-validator.ts
+++ b/packages/ai/src/utils/schema/json-schema-validator.ts
@@ -20,6 +20,14 @@ export interface JsonSchemaValidationIssue {
 	message: string;
 	expectedTypes?: string[];
 	keyword?: string;
+	/**
+	 * Marks issues that originate inside a failed `anyOf` / `oneOf` branch.
+	 * Consumers such as the tool-argument coercion layer use this to avoid
+	 * applying type repairs (e.g. singleton-array wrapping) that would be
+	 * authoritative outside of a combinator but are only one candidate
+	 * branch's expectation here.
+	 */
+	fromUnionBranch?: boolean;
 }
 
 export interface JsonSchemaValidationResult {
@@ -242,7 +250,17 @@ function validateSchemaNode(
 		const branchValid = keyword === "anyOf" ? matches > 0 : matches === 1;
 		if (!branchValid) {
 			if (matches === 0 && firstIssues && firstIssues.length > 0) {
-				issues.push(...firstIssues);
+				// Only tag issues that sit at the combinator's own path as
+				// union-branch; deeper issues describe a specific field within
+				// the failed branch and should remain individually repairable.
+				const unionDepth = path.length;
+				for (const branchIssue of firstIssues) {
+					if (branchIssue.path.length === unionDepth) {
+						issues.push({ ...branchIssue, fromUnionBranch: true });
+					} else {
+						issues.push(branchIssue);
+					}
+				}
 			} else {
 				pushIssue(
 					issues,
diff --git a/packages/ai/src/utils/schema/normalize.ts b/packages/ai/src/utils/schema/normalize.ts
index 1b21afd67..ac50eccb7 100644
--- a/packages/ai/src/utils/schema/normalize.ts
+++ b/packages/ai/src/utils/schema/normalize.ts
@@ -11,6 +11,7 @@ import { dereferenceJsonSchema } from "./dereference";
 import { upgradeJsonSchemaTo202012 } from "./draft";
 import { areJsonValuesEqual, mergePropertySchemas } from "./equality";
 import {
+	ALL_CCA_TYPE_SPECIFIC_KEYS,
 	CLOUD_CODE_ASSIST_SHARED_SCHEMA_KEYS,
 	CLOUD_CODE_ASSIST_TYPE_SPECIFIC_KEYS,
 	COMBINATOR_KEYS,
@@ -501,12 +502,32 @@ function collapseMixedTypeCombinerVariants(schema: JsonObject, combiner: "anyOf"
 	if (variantTypes.length < 2 || variantTypes.every(type => type === "object")) {
 		return schema;
 	}
-
 	const nextSchema = copySchemaWithout(schema, combiner);
 	const nonNullTypes = variantTypes.filter(t => t !== "null");
-	nextSchema.type = nonNullTypes[0] ?? variantTypes[0];
+	const chosenType: string = nonNullTypes[0] ?? variantTypes[0];
+	nextSchema.type = chosenType;
+	const chosenTypeAllowedKeys = CLOUD_CODE_ASSIST_TYPE_SPECIFIC_KEYS[chosenType] ?? {};
+
+	// Strip sibling keys that were copied from the parent and belong to a
+	// different type (e.g. `items` sibling on a now-string-typed schema).
+	for (const key in nextSchema) {
+		if (!Object.hasOwn(nextSchema, key)) continue;
+		if (key === "type") continue;
+		if (
+			Object.hasOwn(ALL_CCA_TYPE_SPECIFIC_KEYS, key) &&
+			!Object.hasOwn(chosenTypeAllowedKeys, key) &&
+			!Object.hasOwn(CLOUD_CODE_ASSIST_SHARED_SCHEMA_KEYS, key)
+		) {
+			delete nextSchema[key];
+		}
+	}
+
 	for (const key in mergedVariantFields) {
 		if (!Object.hasOwn(mergedVariantFields, key)) continue;
+		// Drop type-specific keys that don't belong to the chosen type
+		if (!Object.hasOwn(chosenTypeAllowedKeys, key) && !Object.hasOwn(CLOUD_CODE_ASSIST_SHARED_SCHEMA_KEYS, key)) {
+			continue;
+		}
 		const value = mergedVariantFields[key];
 		const existingValue = nextSchema[key];
 		if (existingValue !== undefined && !areJsonValuesEqual(existingValue, value)) {
diff --git a/packages/ai/src/utils/validation.ts b/packages/ai/src/utils/validation.ts
index 7f82626ec..506c72978 100644
--- a/packages/ai/src/utils/validation.ts
+++ b/packages/ai/src/utils/validation.ts
@@ -724,6 +724,7 @@ interface FlatIssue {
 	keyword: "type" | "unrecognized" | "other";
 	instancePath: string;
 	expectedTypes: string[];
+	unionBranch: boolean;
 }
 
 /**
@@ -759,12 +760,12 @@ function mapZodExpectedToJsonSchemaType(expected: unknown): string | null {
  */
 function flattenIssues(issues: ReadonlyArray<ZodIssue>): FlatIssue[] {
 	const out: FlatIssue[] = [];
-	const walk = (issue: ZodIssue, prefix: ReadonlyArray<PropertyKey>): void => {
+	const walk = (issue: ZodIssue, prefix: ReadonlyArray<PropertyKey>, unionBranch: boolean): void => {
 		const fullPath = prefix.length === 0 ? issue.path : [...prefix, ...issue.path];
 		if (issue.code === "invalid_type") {
 			const mapped = mapZodExpectedToJsonSchemaType((issue as { expected?: unknown }).expected);
 			if (mapped) {
-				out.push({ keyword: "type", instancePath: pathToPointer(fullPath), expectedTypes: [mapped] });
+				out.push({ keyword: "type", instancePath: pathToPointer(fullPath), expectedTypes: [mapped], unionBranch });
 				return;
 			}
 		}
@@ -775,6 +776,7 @@ function flattenIssues(issues: ReadonlyArray<ZodIssue>): FlatIssue[] {
 					keyword: "unrecognized",
 					instancePath: pathToPointer([...fullPath, key]),
 					expectedTypes: [],
+					unionBranch,
 				});
 			}
 			return;
@@ -782,17 +784,21 @@ function flattenIssues(issues: ReadonlyArray<ZodIssue>): FlatIssue[] {
 		if (issue.code === "invalid_union") {
 			const inner = (issue as unknown as { errors?: ReadonlyArray<ReadonlyArray<ZodIssue>> }).errors;
 			if (inner) {
+				// A union-branch issue only competes with a sibling branch when it
+				// sits at the union node's own path. Issues whose own path is
+				// non-empty live on a deeper field that an already-identified
+				// branch owns, so the singleton-array repair should still apply.
 				for (const branch of inner) {
 					for (const child of branch) {
-						walk(child, fullPath);
+						walk(child, fullPath, child.path.length === 0);
 					}
 				}
 			}
 			return;
 		}
-		out.push({ keyword: "other", instancePath: pathToPointer(fullPath), expectedTypes: [] });
+		out.push({ keyword: "other", instancePath: pathToPointer(fullPath), expectedTypes: [], unionBranch });
 	};
-	for (const issue of issues) walk(issue, []);
+	for (const issue of issues) walk(issue, [], false);
 	return out;
 }
 
@@ -801,7 +807,9 @@ function flattenIssues(issues: ReadonlyArray<ZodIssue>): FlatIssue[] {
  *
  * Two kinds of repair are applied:
  *  - **type**: when a value is a JSON-encoded string and the schema wants
- *    something else, parse it and substitute the parsed value.
+ *    something else, parse it and substitute the parsed value. When a
+ *    non-union schema wants an array but receives a singleton value, wrap that
+ *    value in a one-element array.
  *  - **unrecognized**: when a strict object received an extra key (Zod's
  *    `unrecognized_keys` or JSON Schema's `additionalProperties: false`),
  *    drop that key so re-validation succeeds. This effectively coerces every
@@ -811,6 +819,7 @@ function flattenIssues(issues: ReadonlyArray<ZodIssue>): FlatIssue[] {
  * The function is safe and conservative:
  *   - Only processes "type" and "unrecognized" issues
  *   - Only attempts JSON coercion on string values
+ *   - Only wraps singleton array values for non-union type expectations
  *   - Only accepts parsed results that match the expected type
  *   - Clones the args object before mutation (copy-on-write)
  */
@@ -836,12 +845,16 @@ function coerceArgsFromIssues(args: unknown, issues: FlatIssue[]): { value: unkn
 		if (issue.expectedTypes.length === 0) continue;
 
 		const currentValue = getValueAtPointer(nextArgs, issue.instancePath);
-		if (typeof currentValue !== "string") continue;
-
-		const result = tryParseJsonForTypes(currentValue, issue.expectedTypes);
+		const result =
+			typeof currentValue === "string"
+				? tryParseJsonForTypes(currentValue, issue.expectedTypes)
+				: { value: currentValue, changed: false };
 		const coercedValue = result.changed
 			? result.value
-			: issue.expectedTypes.includes("array")
+			: issue.expectedTypes.includes("array") &&
+					!issue.unionBranch &&
+					currentValue !== undefined &&
+					!Array.isArray(currentValue)
 				? [currentValue]
 				: undefined;
 		if (coercedValue === undefined) continue;
@@ -905,17 +918,20 @@ function preserveUnknownRootFields(input: unknown, parsed: unknown): unknown {
 
 function flattenJsonSchemaIssues(issues: ReadonlyArray<JsonSchemaValidationIssue>): FlatIssue[] {
 	return issues.map(issue => {
+		const unionBranch = issue.fromUnionBranch === true;
 		if (issue.keyword === "additionalProperties") {
 			return {
 				keyword: "unrecognized",
 				instancePath: pathToPointer(issue.path),
 				expectedTypes: [],
+				unionBranch,
 			};
 		}
 		return {
 			keyword: issue.keyword === "type" ? "type" : "other",
 			instancePath: pathToPointer(issue.path),
 			expectedTypes: issue.expectedTypes ?? [],
+			unionBranch,
 		};
 	});
 }
diff --git a/packages/ai/test/anthropic-alignment.test.ts b/packages/ai/test/anthropic-alignment.test.ts
index b40a1b8b9..dd4878a7c 100644
--- a/packages/ai/test/anthropic-alignment.test.ts
+++ b/packages/ai/test/anthropic-alignment.test.ts
@@ -203,6 +203,11 @@ describe("Anthropic request fingerprint alignment", () => {
 	});
 
 	it("matches CC system-block layout: billing and instruction uncached, context cached in order", () => {
+		// We mimic Claude Code's billing+instruction system layout but do NOT emit
+		// the `scope: "global"` field that CC attaches to its middle breakpoint —
+		// `prompt-caching-scope-2026-01-05` only works against canonical
+		// `api.anthropic.com`, and third-party Anthropic-compatible proxies
+		// (z.ai, openrouter, g0i, …) reject the unknown field outright.
 		const blocks = buildAnthropicSystemBlocks(["Stay concise."], {
 			includeClaudeCodeInstruction: true,
 			extraInstructions: ["Use citations when possible"],
@@ -217,7 +222,7 @@ describe("Anthropic request fingerprint alignment", () => {
 		expect(blocks?.[2]).toEqual({
 			type: "text",
 			text: "Use citations when possible",
-			cache_control: { type: "ephemeral", scope: "global" },
+			cache_control: { type: "ephemeral" },
 		});
 		expect(blocks?.[3]).toEqual({
 			type: "text",
@@ -239,7 +244,7 @@ describe("Anthropic request fingerprint alignment", () => {
 		expect(payload.system?.[0]?.cache_control).toBeUndefined();
 		expect(payload.system?.[1]?.text).toBe(claudeCodeSystemInstruction);
 		expect(payload.system?.[1]?.cache_control).toBeUndefined();
-		expect(payload.system?.[2]?.cache_control).toEqual({ type: "ephemeral", ttl: "1h", scope: "global" });
+		expect(payload.system?.[2]?.cache_control).toEqual({ type: "ephemeral", ttl: "1h" });
 		const content = payload.messages?.[0]?.content;
 		expect(Array.isArray(content)).toBe(true);
 		expect(Array.isArray(content) ? content[0]?.cache_control : undefined).toEqual({
diff --git a/packages/ai/test/anthropic-unsigned-thinking-replay.test.ts b/packages/ai/test/anthropic-unsigned-thinking-replay.test.ts
new file mode 100644
index 000000000..9b3f8dcbc
--- /dev/null
+++ b/packages/ai/test/anthropic-unsigned-thinking-replay.test.ts
@@ -0,0 +1,164 @@
+import { describe, expect, it } from "bun:test";
+import { convertAnthropicMessages } from "@oh-my-pi/pi-ai/providers/anthropic";
+import type { AssistantMessage, Message, Model, ToolResultMessage, UserMessage } from "@oh-my-pi/pi-ai/types";
+
+/**
+ * Regression: Anthropic-compatible reasoning endpoints often emit `thinking`
+ * blocks without a first-party Anthropic signature, but still expect those
+ * blocks back as native `type: "thinking"` on continuation. Demoting unsigned
+ * thinking to text strips the reasoning chain and can destabilize follow-up
+ * tool-call argument serialization (the upstream cause behind #2005's `todo`
+ * renderer crash).
+ *
+ * Official Anthropic remains conservative: unsigned thinking is demoted to text
+ * there because the first-party API enforces signature-based integrity.
+ */
+function makeModel(overrides: Partial<Model<"anthropic-messages">> = {}): Model<"anthropic-messages"> {
+	return {
+		api: "anthropic-messages",
+		provider: "custom-anthropic",
+		id: "reasoning-model",
+		name: "Reasoning Anthropic-Compatible Model",
+		baseUrl: "https://llm.example.com/anthropic",
+		input: ["text"],
+		cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+		maxTokens: 8_192,
+		contextWindow: 200_000,
+		reasoning: true,
+		...overrides,
+	};
+}
+
+function makeUser(text = "continue"): UserMessage {
+	return { role: "user", content: text, timestamp: 0 };
+}
+
+function makeAssistantThinking(thinking: string, tail: AssistantMessage["content"][number][] = []): AssistantMessage {
+	return {
+		role: "assistant",
+		content: [{ type: "thinking", thinking, thinkingSignature: "" }, ...tail],
+		api: "anthropic-messages",
+		provider: "custom-anthropic",
+		model: "reasoning-model",
+		usage: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+			totalTokens: 0,
+			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+		},
+		stopReason: "toolUse",
+		timestamp: 0,
+	};
+}
+
+interface WireThinkingBlock {
+	type: "thinking";
+	thinking: string;
+	signature: string;
+}
+interface WireTextBlock {
+	type: "text";
+	text: string;
+}
+interface WireToolUseBlock {
+	type: "tool_use";
+	id: string;
+	name: string;
+	input: Record<string, unknown>;
+}
+type WireBlock = WireThinkingBlock | WireTextBlock | WireToolUseBlock | { type: string; [key: string]: unknown };
+
+function assistantWireBlocks(messages: Message[], model: Model<"anthropic-messages">): WireBlock[] {
+	const params = convertAnthropicMessages(messages, model, false);
+	const assistant = params.find(p => p.role === "assistant");
+	return (assistant?.content as WireBlock[] | undefined) ?? [];
+}
+
+describe("Anthropic-compatible unsigned thinking replay (#2005)", () => {
+	it("preserves unsigned thinking for non-official reasoning endpoints", () => {
+		const blocks = assistantWireBlocks(
+			[
+				makeUser("solve x"),
+				makeAssistantThinking("plan: read the file, then edit", [{ type: "text", text: "Sure." }]),
+			],
+			makeModel(),
+		);
+		expect(blocks[0]).toEqual({
+			type: "thinking",
+			thinking: "plan: read the file, then edit",
+			signature: "",
+		});
+		expect(blocks[1]).toEqual({ type: "text", text: "Sure." });
+	});
+
+	it("covers the Xiaomi MiMo Anthropic-compatible reporter configuration without provider allowlists", () => {
+		const model = makeModel({
+			provider: "user-custom",
+			id: "mimo-v2.5-pro",
+			name: "MiMo V2.5 Pro (Singapore)",
+			baseUrl: "https://token-plan-sgp.xiaomimimo.com/anthropic",
+			maxTokens: 131_072,
+			contextWindow: 1_048_576,
+		});
+		const blocks = assistantWireBlocks([makeUser(), makeAssistantThinking("hidden reasoning")], model);
+		expect(blocks[0]).toEqual({ type: "thinking", thinking: "hidden reasoning", signature: "" });
+	});
+
+	it("preserves legacy known non-signing endpoints even if model.reasoning is false", () => {
+		const model = makeModel({ provider: "custom", baseUrl: "https://api.deepseek.com/v1", reasoning: false });
+		const blocks = assistantWireBlocks([makeUser(), makeAssistantThinking("deepseek reasoning")], model);
+		expect(blocks[0]?.type).toBe("thinking");
+	});
+
+	it("still degrades unsigned thinking to text for official Anthropic", () => {
+		const model = makeModel({ provider: "anthropic", baseUrl: "https://api.anthropic.com" });
+		const blocks = assistantWireBlocks([makeUser(), makeAssistantThinking("internal scratch")], model);
+		expect(blocks[0]?.type).toBe("text");
+		expect((blocks[0] as WireTextBlock).text).toBe("internal scratch");
+	});
+
+	it("treats a missing baseUrl as official Anthropic (resolveAnthropicBaseUrl default)", () => {
+		// `isAnthropicApiBaseUrl(undefined) === true` because the actual HTTP
+		// dispatch falls back to https://api.anthropic.com. Same-id custom
+		// overrides that only tweak model metadata (no baseUrl override) must
+		// not regress to native-thinking replay against the first-party API.
+		const model = { ...makeModel(), provider: "anthropic", baseUrl: "" };
+		const blocks = assistantWireBlocks([makeUser(), makeAssistantThinking("internal scratch")], model);
+		expect(blocks[0]?.type).toBe("text");
+		expect((blocks[0] as WireTextBlock).text).toBe("internal scratch");
+	});
+
+	it("still degrades unsigned thinking to text for non-reasoning unknown endpoints", () => {
+		const model = makeModel({ reasoning: false, baseUrl: "https://plain.example.com/anthropic" });
+		const blocks = assistantWireBlocks([makeUser(), makeAssistantThinking("scratch")], model);
+		expect(blocks[0]?.type).toBe("text");
+		expect((blocks[0] as WireTextBlock).text).toBe("scratch");
+	});
+
+	it("keeps thinking → tool_use pairing intact across continuation conversion", () => {
+		const toolResult: ToolResultMessage = {
+			role: "toolResult",
+			toolCallId: "toolu_reasoning_1",
+			toolName: "read",
+			content: [{ type: "text", text: "file body" }],
+			isError: false,
+			timestamp: 0,
+		};
+		const model = makeModel();
+		const messages: Message[] = [
+			makeUser("read README"),
+			makeAssistantThinking("I need to call the read tool", [
+				{ type: "toolCall", id: "toolu_reasoning_1", name: "read", arguments: { path: "README.md" } },
+			]),
+			toolResult,
+		];
+		const params = convertAnthropicMessages(messages, model, false);
+		expect(params.map(p => p.role)).toEqual(["user", "assistant", "user"]);
+		const assistantBlocks = params[1].content as WireBlock[];
+		expect(assistantBlocks[0]?.type).toBe("thinking");
+		expect(assistantBlocks[1]?.type).toBe("tool_use");
+		expect((assistantBlocks[1] as WireToolUseBlock).id).toBe("toolu_reasoning_1");
+	});
+});
diff --git a/packages/ai/test/auth-gateway-openai-responses.test.ts b/packages/ai/test/auth-gateway-openai-responses.test.ts
index fe3a21801..c6cbd8c9b 100644
--- a/packages/ai/test/auth-gateway-openai-responses.test.ts
+++ b/packages/ai/test/auth-gateway-openai-responses.test.ts
@@ -1,5 +1,5 @@
 import { describe, expect, it } from "bun:test";
-import { Effort } from "../src/model-thinking";
+import { Effort } from "../src/effort";
 import { encodeResponse, encodeStream, parseRequest } from "../src/providers/openai-responses-server";
 import type { AssistantMessage } from "../src/types";
 import { AssistantMessageEventStream } from "../src/utils/event-stream";
diff --git a/packages/ai/test/auth-gateway-pi-native.test.ts b/packages/ai/test/auth-gateway-pi-native.test.ts
index 7c10a65a7..5f3a77a76 100644
--- a/packages/ai/test/auth-gateway-pi-native.test.ts
+++ b/packages/ai/test/auth-gateway-pi-native.test.ts
@@ -1,5 +1,5 @@
 import { describe, expect, it } from "bun:test";
-import { Effort } from "../src/model-thinking";
+import { Effort } from "../src/effort";
 import { encodeStream, formatError, parseRequest } from "../src/providers/pi-native-server";
 import type {
 	AssistantMessage,
diff --git a/packages/ai/test/auth-retry.test.ts b/packages/ai/test/auth-retry.test.ts
new file mode 100644
index 000000000..d992a33c1
--- /dev/null
+++ b/packages/ai/test/auth-retry.test.ts
@@ -0,0 +1,162 @@
+import { describe, expect, it } from "bun:test";
+import type { ApiKeyResolveContext } from "@oh-my-pi/pi-ai";
+import { isApiKeyResolver, isAuthRetryableError, resolveApiKeyOnce, withAuth } from "@oh-my-pi/pi-ai";
+
+function authError(status = 401): Error & { status: number } {
+	return Object.assign(new Error(`${status} authentication_error`), { status });
+}
+
+function usageLimitError(): Error & { status: number } {
+	return Object.assign(new Error("You have hit your ChatGPT usage limit (pro plan). Try again in ~158 min."), {
+		status: 429,
+	});
+}
+
+describe("isApiKeyResolver / resolveApiKeyOnce", () => {
+	it("narrows resolver vs static key and resolves the initial value", async () => {
+		expect(isApiKeyResolver("static")).toBe(false);
+		expect(isApiKeyResolver(undefined)).toBe(false);
+		expect(isApiKeyResolver(() => "k")).toBe(true);
+
+		expect(await resolveApiKeyOnce("static")).toBe("static");
+		expect(await resolveApiKeyOnce(undefined)).toBeUndefined();
+
+		let seen: ApiKeyResolveContext | undefined;
+		const resolved = await resolveApiKeyOnce(ctx => {
+			seen = ctx;
+			return "minted";
+		});
+		expect(resolved).toBe("minted");
+		// Initial resolve must look like an initial resolve, not a retry.
+		expect(seen).toEqual({ lastChance: false, error: undefined, signal: undefined });
+	});
+});
+
+describe("isAuthRetryableError", () => {
+	it("treats 401 and usage-limit phrasing as retryable, everything else as not", () => {
+		expect(isAuthRetryableError(authError(401))).toBe(true);
+		expect(isAuthRetryableError(usageLimitError())).toBe(true);
+		// A 429 whose body names the *account's* rate limit is rotatable (switch
+		// account), even though it isn't a 401 and isn't phrased "usage limit".
+		expect(
+			isAuthRetryableError(
+				Object.assign(
+					new Error(
+						'429 {"type":"error","error":{"type":"rate_limit_error","message":"This request would exceed your account\'s rate limit. Please try again later."}} retry-after-ms=9779000',
+					),
+					{ status: 429 },
+				),
+			),
+		).toBe(true);
+		// A generic (non-account) 429 rate limit is NOT rotatable — switching
+		// credentials won't help an org/global limit.
+		expect(isAuthRetryableError(Object.assign(new Error("429 too many requests"), { status: 429 }))).toBe(false);
+		expect(isAuthRetryableError("Error: 401 unauthorized")).toBe(true);
+		expect(isAuthRetryableError(authError(403))).toBe(false);
+		expect(isAuthRetryableError(authError(500))).toBe(false);
+		expect(isAuthRetryableError(new Error("network blip"))).toBe(false);
+		expect(isAuthRetryableError(undefined)).toBe(false);
+	});
+});
+
+describe("withAuth", () => {
+	it("runs a single attempt for a static string key (no retry)", async () => {
+		const keys: Array<string | undefined> = [];
+		const result = await withAuth("static-key", async key => {
+			keys.push(key);
+			return `ok:${key}`;
+		});
+		expect(result).toBe("ok:static-key");
+		expect(keys).toEqual(["static-key"]);
+	});
+
+	it("throws when a static key is missing", async () => {
+		await expect(withAuth(undefined, async () => "never", { missingKeyMessage: "no key for foo" })).rejects.toThrow(
+			"no key for foo",
+		);
+	});
+
+	it("refreshes the same account, then switches, in order", async () => {
+		const keys: string[] = [];
+		const contexts: ApiKeyResolveContext[] = [];
+		const result = await withAuth(
+			ctx => {
+				contexts.push(ctx);
+				return ctx.error === undefined ? "k0" : ctx.lastChance ? "k2" : "k1";
+			},
+			async key => {
+				keys.push(key);
+				if (key === "k2") return "success";
+				throw authError();
+			},
+		);
+		expect(result).toBe("success");
+		expect(keys).toEqual(["k0", "k1", "k2"]);
+		expect(contexts.map(ctx => ({ lastChance: ctx.lastChance, hasError: ctx.error !== undefined }))).toEqual([
+			{ lastChance: false, hasError: false },
+			{ lastChance: false, hasError: true },
+			{ lastChance: true, hasError: true },
+		]);
+	});
+
+	it("stops retrying when the resolver returns undefined", async () => {
+		const keys: string[] = [];
+		const original = authError();
+		await expect(
+			withAuth(
+				ctx => (ctx.error === undefined ? "k0" : undefined),
+				async key => {
+					keys.push(key);
+					throw original;
+				},
+			),
+		).rejects.toBe(original);
+		expect(keys).toEqual(["k0"]);
+	});
+
+	it("does not re-attempt when the re-resolved key is unchanged", async () => {
+		const keys: string[] = [];
+		const original = authError();
+		// refresh-same returns the same key (skip), switch returns the same key (skip).
+		await expect(
+			withAuth(
+				() => "same",
+				async key => {
+					keys.push(key);
+					throw original;
+				},
+			),
+		).rejects.toBe(original);
+		expect(keys).toEqual(["same"]);
+	});
+
+	it("propagates non-auth errors without retrying", async () => {
+		const keys: string[] = [];
+		const boom = new Error("network blip");
+		await expect(
+			withAuth(
+				ctx => (ctx.error === undefined ? "k0" : "k1"),
+				async key => {
+					keys.push(key);
+					throw boom;
+				},
+			),
+		).rejects.toBe(boom);
+		expect(keys).toEqual(["k0"]);
+	});
+
+	it("honors a custom isAuthError classifier", async () => {
+		const keys: string[] = [];
+		const result = await withAuth(
+			ctx => (ctx.error === undefined ? "k0" : "k1"),
+			async key => {
+				keys.push(key);
+				if (key === "k0") throw new Error("CUSTOM_RETRY");
+				return "ok";
+			},
+			{ isAuthError: error => error instanceof Error && error.message === "CUSTOM_RETRY" },
+		);
+		expect(result).toBe("ok");
+		expect(keys).toEqual(["k0", "k1"]);
+	});
+});
diff --git a/packages/ai/test/auth-storage-force-refresh-rotate.test.ts b/packages/ai/test/auth-storage-force-refresh-rotate.test.ts
new file mode 100644
index 000000000..fc7d0bb01
--- /dev/null
+++ b/packages/ai/test/auth-storage-force-refresh-rotate.test.ts
@@ -0,0 +1,163 @@
+import { afterEach, beforeEach, describe, expect, test, vi } from "bun:test";
+import * as fs from "node:fs/promises";
+import * as os from "node:os";
+import * as path from "node:path";
+import { type AuthCredentialStore, AuthStorage, SqliteAuthCredentialStore } from "../src/auth-storage";
+import { registerOAuthProvider, unregisterOAuthProviders } from "../src/utils/oauth";
+
+const PROVIDER = "unit-rotate-oauth";
+const SOURCE = "auth-storage-force-refresh-rotate-test";
+
+function farExpiry(): number {
+	return Date.now() + 60 * 60_000;
+}
+
+function authError(): Error & { status: number } {
+	return Object.assign(new Error("401 authentication_error"), { status: 401 });
+}
+
+function usageLimitError(): Error & { status: number } {
+	return Object.assign(new Error("You have hit your ChatGPT usage limit (pro plan). Try again in ~158 min."), {
+		status: 429,
+	});
+}
+
+describe("AuthStorage forceRefresh + rotateSessionCredential", () => {
+	let tempDir = "";
+	let store: AuthCredentialStore | undefined;
+	let authStorage: AuthStorage | undefined;
+
+	beforeEach(async () => {
+		tempDir = await fs.mkdtemp(path.join(os.tmpdir(), "pi-ai-rotate-"));
+		store = await SqliteAuthCredentialStore.open(path.join(tempDir, "agent.db"));
+		authStorage = new AuthStorage(store);
+	});
+
+	afterEach(async () => {
+		vi.restoreAllMocks();
+		unregisterOAuthProviders(SOURCE);
+		store?.close();
+		store = undefined;
+		authStorage = undefined;
+		if (tempDir) {
+			await fs.rm(tempDir, { recursive: true, force: true });
+			tempDir = "";
+		}
+	});
+
+	function registerProvider(onRefresh?: () => void): void {
+		registerOAuthProvider({
+			id: PROVIDER,
+			name: "Rotate Unit",
+			sourceId: SOURCE,
+			async login() {
+				return { access: "login", refresh: "login", expires: farExpiry() };
+			},
+			async refreshToken(credentials) {
+				onRefresh?.();
+				return {
+					...credentials,
+					access: "minted-access",
+					refresh: "minted-refresh",
+					expires: farExpiry(),
+				};
+			},
+			getApiKey(credentials) {
+				return credentials.access;
+			},
+		});
+	}
+
+	test("forceRefresh re-mints a not-yet-expired token; a normal resolve uses the cached token", async () => {
+		if (!authStorage) throw new Error("test setup failed");
+		let refreshCalls = 0;
+		registerProvider(() => {
+			refreshCalls += 1;
+		});
+		await authStorage.set(PROVIDER, [
+			{ type: "oauth", access: "cached-access", refresh: "cached-refresh", expires: farExpiry() },
+		]);
+
+		const cached = await authStorage.getApiKey(PROVIDER, "s-control");
+		expect(cached).toBe("cached-access");
+		expect(refreshCalls).toBe(0);
+
+		const forced = await authStorage.getApiKey(PROVIDER, "s-force", { forceRefresh: true });
+		expect(forced).toBe("minted-access");
+		expect(refreshCalls).toBe(1);
+
+		// The re-minted credential is persisted, so the next plain resolve sees it.
+		const after = await authStorage.getApiKey(PROVIDER, "s-after");
+		expect(after).toBe("minted-access");
+	});
+
+	test("rotateSessionCredential(401) blocks + clears the sticky and rotates to a sibling", async () => {
+		if (!authStorage) throw new Error("test setup failed");
+		registerProvider();
+		await authStorage.set(PROVIDER, [
+			{ type: "oauth", access: "acc-A", refresh: "ref-A", expires: farExpiry() },
+			{ type: "oauth", access: "acc-B", refresh: "ref-B", expires: farExpiry() },
+		]);
+
+		const first = await authStorage.getApiKey(PROVIDER, "sess");
+		expect(["acc-A", "acc-B"]).toContain(first ?? "");
+
+		const usageLimitSpy = vi.spyOn(authStorage, "markUsageLimitReached");
+		const rotated = await authStorage.rotateSessionCredential(PROVIDER, "sess", { error: authError() });
+
+		expect(rotated).toBe(true);
+		// A hard 401 must NOT take the usage-limit code path.
+		expect(usageLimitSpy).not.toHaveBeenCalled();
+
+		const second = await authStorage.getApiKey(PROVIDER, "sess");
+		expect(["acc-A", "acc-B"]).toContain(second ?? "");
+		expect(second).not.toBe(first);
+	});
+
+	test("rotateSessionCredential(usage-limit) delegates to markUsageLimitReached", async () => {
+		if (!authStorage) throw new Error("test setup failed");
+		registerProvider();
+		await authStorage.set(PROVIDER, [
+			{ type: "oauth", access: "acc-A", refresh: "ref-A", expires: farExpiry() },
+			{ type: "oauth", access: "acc-B", refresh: "ref-B", expires: farExpiry() },
+		]);
+
+		const first = await authStorage.getApiKey(PROVIDER, "sess");
+		const usageLimitSpy = vi.spyOn(authStorage, "markUsageLimitReached");
+
+		const rotated = await authStorage.rotateSessionCredential(PROVIDER, "sess", {
+			error: usageLimitError(),
+		});
+
+		expect(rotated).toBe(true);
+		// Usage / account-rate-limit errors route to markUsageLimitReached, which
+		// owns the block duration (default + server usage-report reset) — the
+		// resolver never parses retry-after itself.
+		expect(usageLimitSpy).toHaveBeenCalledTimes(1);
+		expect(usageLimitSpy.mock.calls[0]?.[0]).toBe(PROVIDER);
+		expect(usageLimitSpy.mock.calls[0]?.[1]).toBe("sess");
+
+		const second = await authStorage.getApiKey(PROVIDER, "sess");
+		expect(second).not.toBe(first);
+	});
+
+	test("rotateSessionCredential reports no sibling for a single-credential setup", async () => {
+		if (!authStorage) throw new Error("test setup failed");
+		registerProvider();
+		await authStorage.set(PROVIDER, [
+			{ type: "oauth", access: "only-access", refresh: "only-refresh", expires: farExpiry() },
+		]);
+
+		await authStorage.getApiKey(PROVIDER, "sess");
+		expect(await authStorage.rotateSessionCredential(PROVIDER, "sess", { error: authError() })).toBe(false);
+	});
+
+	test("rotateSessionCredential returns false when the session has no sticky credential", async () => {
+		if (!authStorage) throw new Error("test setup failed");
+		registerProvider();
+		await authStorage.set(PROVIDER, [{ type: "oauth", access: "acc-A", refresh: "ref-A", expires: farExpiry() }]);
+
+		// Never resolved a key for this session → nothing to rotate away from.
+		expect(await authStorage.rotateSessionCredential(PROVIDER, "untouched", { error: authError() })).toBe(false);
+	});
+});
diff --git a/packages/ai/test/github-copilot-model-limits.test.ts b/packages/ai/test/github-copilot-model-limits.test.ts
index c46f14487..07d76ee87 100644
--- a/packages/ai/test/github-copilot-model-limits.test.ts
+++ b/packages/ai/test/github-copilot-model-limits.test.ts
@@ -2,8 +2,8 @@ import { afterEach, describe, expect, it, vi } from "bun:test";
 import * as fs from "node:fs/promises";
 import * as os from "node:os";
 import * as path from "node:path";
+import { Effort } from "../src/effort";
 import { createModelManager } from "../src/model-manager";
-import { Effort } from "../src/model-thinking";
 import { getBundledModel } from "../src/models";
 import { githubCopilotModelManagerOptions } from "../src/provider-models/openai-compat";
 
diff --git a/packages/ai/test/github-copilot-reasoning.test.ts b/packages/ai/test/github-copilot-reasoning.test.ts
index 32a88c006..28746c85d 100644
--- a/packages/ai/test/github-copilot-reasoning.test.ts
+++ b/packages/ai/test/github-copilot-reasoning.test.ts
@@ -1,5 +1,5 @@
 import { describe, expect, it } from "bun:test";
-import { Effort } from "../src/model-thinking";
+import { Effort } from "../src/effort";
 import { getBundledModel } from "../src/models";
 import { streamAnthropic } from "../src/providers/anthropic";
 import { streamOpenAIResponses } from "../src/providers/openai-responses";
diff --git a/packages/ai/test/google-antigravity-usage.test.ts b/packages/ai/test/google-antigravity-usage.test.ts
new file mode 100644
index 000000000..4fdec1ae0
--- /dev/null
+++ b/packages/ai/test/google-antigravity-usage.test.ts
@@ -0,0 +1,196 @@
+/**
+ * Antigravity usage provider contract tests. The merge logic
+ * deduplicates per-model quota entries by (tier, windowId),
+ * preserves reset times when bar data and window data come from
+ * different model entries, and handles mixed-case tier names.
+ */
+import { describe, expect, it } from "bun:test";
+import type { UsageFetchContext, UsageFetchParams } from "../src/usage";
+import { antigravityUsageProvider } from "../src/usage/google-antigravity";
+
+const accessTokenFixture = (() => {
+	const header = Buffer.from(JSON.stringify({ alg: "none", typ: "JWT" })).toString("base64url");
+	const body = Buffer.from(JSON.stringify({ sub: "user-fixture" })).toString("base64url");
+	return `${header}.${body}.sig`;
+})();
+
+function makeCredential(overrides?: Partial<UsageFetchParams["credential"]>) {
+	return {
+		type: "oauth" as const,
+		accessToken: accessTokenFixture,
+		refreshToken: "refresh-fixture",
+		expiresAt: Date.now() + 3600_000,
+		projectId: "test-project",
+		email: "test@example.com",
+		accountId: "acct-1",
+		...overrides,
+	} satisfies UsageFetchParams["credential"];
+}
+
+function fakeFetch(json: unknown): typeof fetch {
+	const fn = async () =>
+		new Response(JSON.stringify(json), {
+			status: 200,
+			headers: { "content-type": "application/json" },
+		});
+	return fn as unknown as typeof fetch;
+}
+
+function makeCtx(fetchImpl?: typeof fetch): UsageFetchContext {
+	return { fetch: fetchImpl ?? fakeFetch({}) };
+}
+
+// ── helpers ──────────────────────────────────────────────────────────
+
+function makeApiModel(
+	displayName: string,
+	quota: { remainingFraction?: number; resetTime?: string; tier?: string; windowId?: string },
+) {
+	return {
+		displayName,
+		quotaInfo: {
+			remainingFraction: quota.remainingFraction,
+			resetTime: quota.resetTime,
+			tier: quota.tier,
+			windowId: quota.windowId,
+		},
+	};
+}
+
+// ── tests ────────────────────────────────────────────────────────────
+
+describe("antigravity usage provider", () => {
+	it("merges two models with same tier into one limit", async () => {
+		const payload = {
+			models: {
+				modelA: makeApiModel("Model A", { remainingFraction: 0.3, tier: "premium" }),
+				modelB: makeApiModel("Model B", { remainingFraction: 0.5, tier: "premium" }),
+			},
+		};
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{ provider: "google-antigravity", credential: makeCredential(), signal: undefined },
+			makeCtx(fakeFetch(payload)),
+		);
+		expect(report).not.toBeNull();
+		expect(report!.limits.length).toBe(1);
+	});
+
+	it("keeps the worst remainingFraction when merging same tier", async () => {
+		const payload = {
+			models: {
+				modelA: makeApiModel("Model A", { remainingFraction: 0.1, tier: "premium" }),
+				modelB: makeApiModel("Model B", { remainingFraction: 0.8, tier: "premium" }),
+			},
+		};
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{ provider: "google-antigravity", credential: makeCredential(), signal: undefined },
+			makeCtx(fakeFetch(payload)),
+		);
+		expect(report!.limits.length).toBe(1);
+		expect(report!.limits[0]!.amount.remainingFraction).toBe(0.1);
+	});
+
+	it("merges mixed-case tier names under lowercased key", async () => {
+		const payload = {
+			models: {
+				modelA: makeApiModel("Model A", { remainingFraction: 0.3, tier: "Default" }),
+				modelB: makeApiModel("Model B", { remainingFraction: 0.6, tier: "default" }),
+			},
+		};
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{ provider: "google-antigravity", credential: makeCredential(), signal: undefined },
+			makeCtx(fakeFetch(payload)),
+		);
+		expect(report!.limits.length).toBe(1);
+	});
+
+	it("preserves reset time from an entry even when bar data comes from another", async () => {
+		const now = Date.now();
+		const resetTime = new Date(now + 4 * 3600_000).toISOString();
+		const payload = {
+			models: {
+				modelA: makeApiModel("Model A", { remainingFraction: 0.3, tier: "default" }),
+				modelB: makeApiModel("Model B", { remainingFraction: undefined, tier: "default", resetTime }),
+			},
+		};
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{ provider: "google-antigravity", credential: makeCredential(), signal: undefined },
+			makeCtx(fakeFetch(payload)),
+		);
+		expect(report!.limits.length).toBe(1);
+		expect(report!.limits[0]!.amount.remainingFraction).toBe(0.3);
+		expect(report!.limits[0]!.window).toBeDefined();
+		expect(report!.limits[0]!.window!.resetsAt).toBeGreaterThan(now);
+	});
+
+	it("separates models with different windowIds in the same tier", async () => {
+		const now = Date.now();
+		const t1 = new Date(now + 5 * 3600_000).toISOString();
+		const t2 = new Date(now + 24 * 3600_000).toISOString();
+		const payload = {
+			models: {
+				modelA: makeApiModel("Model A", { remainingFraction: 0.3, tier: "premium", windowId: "5h", resetTime: t1 }),
+				modelB: makeApiModel("Model B", {
+					remainingFraction: 0.7,
+					tier: "premium",
+					windowId: "daily",
+					resetTime: t2,
+				}),
+			},
+		};
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{ provider: "google-antigravity", credential: makeCredential(), signal: undefined },
+			makeCtx(fakeFetch(payload)),
+		);
+		expect(report!.limits.length).toBe(2);
+	});
+
+	it("includes email and projectId in report metadata", async () => {
+		const payload = { models: { m: makeApiModel("M", { remainingFraction: 1 }) } };
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{
+				provider: "google-antigravity",
+				credential: makeCredential({ email: "user@example.com", projectId: "proj-1" }),
+				signal: undefined,
+			},
+			makeCtx(fakeFetch(payload)),
+		);
+		expect(report!.metadata?.email).toBe("user@example.com");
+		expect(report!.metadata?.projectId).toBe("proj-1");
+	});
+
+	it("does not include email when credential has none", async () => {
+		const payload = { models: { m: makeApiModel("M", { remainingFraction: 1 }) } };
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{ provider: "google-antigravity", credential: makeCredential({ email: undefined }), signal: undefined },
+			makeCtx(fakeFetch(payload)),
+		);
+		expect(report!.metadata?.email).toBeUndefined();
+	});
+
+	it("sorts limits by remainingFraction ascending (worst first)", async () => {
+		const payload = {
+			models: {
+				modelA: makeApiModel("Model A", { remainingFraction: 0.9, tier: "high" }),
+				modelB: makeApiModel("Model B", { remainingFraction: 0.2, tier: "low" }),
+				modelC: makeApiModel("Model C", { remainingFraction: 0.5, tier: "mid" }),
+			},
+		};
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{ provider: "google-antigravity", credential: makeCredential(), signal: undefined },
+			makeCtx(fakeFetch(payload)),
+		);
+		expect(report!.limits.length).toBe(3);
+		expect(report!.limits[0]!.amount.remainingFraction).toBe(0.2);
+		expect(report!.limits[1]!.amount.remainingFraction).toBe(0.5);
+		expect(report!.limits[2]!.amount.remainingFraction).toBe(0.9);
+	});
+
+	it("returns null when credential has no projectId", async () => {
+		const report = await antigravityUsageProvider.fetchUsage!(
+			{ provider: "google-antigravity", credential: makeCredential({ projectId: undefined }), signal: undefined },
+			makeCtx(),
+		);
+		expect(report).toBeNull();
+	});
+});
diff --git a/packages/ai/test/issue-1207-repro.test.ts b/packages/ai/test/issue-1207-repro.test.ts
index 85e31868d..d71c91b3d 100644
--- a/packages/ai/test/issue-1207-repro.test.ts
+++ b/packages/ai/test/issue-1207-repro.test.ts
@@ -91,6 +91,19 @@ describe("issue #1207 — DeepSeek V4 keeps reasoning with tools", () => {
 		expect(body.max_completion_tokens).toBeUndefined();
 	});
 
+	it("does not mix Fireworks DeepSeek effort with the native thinking toggle", async () => {
+		const model = getBundledModel("fireworks", "deepseek-v4-pro") as Model<"openai-completions">;
+		const compat = resolveOpenAICompat(model);
+		const body = await capturePayload(model);
+
+		expect(compat.extraBody).toBeUndefined();
+		expect(body.tools).toBeDefined();
+		expect(body.tool_choice).toBeUndefined();
+		expect(body.reasoning_effort).toBe("high");
+		expect(body.thinking).toBeUndefined();
+		expect(body.max_tokens).toBe(123);
+	});
+
 	it("preserves OpenRouter reasoning when tool_choice auto is present", async () => {
 		const model = getBundledModel("openrouter", "deepseek/deepseek-v4-flash") as Model<"openai-completions">;
 		const compat = detectOpenAICompat(model);
diff --git a/packages/ai/test/issue-1373-repro.test.ts b/packages/ai/test/issue-1373-repro.test.ts
index 986acbf3c..88e855f4e 100644
--- a/packages/ai/test/issue-1373-repro.test.ts
+++ b/packages/ai/test/issue-1373-repro.test.ts
@@ -1,5 +1,5 @@
 import { afterAll, beforeAll, describe, expect, it } from "bun:test";
-import { Effort } from "../src/model-thinking";
+import { Effort } from "../src/effort";
 import { streamBedrock } from "../src/providers/amazon-bedrock";
 import type { Context, Model } from "../src/types";
 
diff --git a/packages/ai/test/issue-826-repro.test.ts b/packages/ai/test/issue-826-repro.test.ts
index efcf9ccea..dd78aeda0 100644
--- a/packages/ai/test/issue-826-repro.test.ts
+++ b/packages/ai/test/issue-826-repro.test.ts
@@ -1,5 +1,5 @@
 import { describe, expect, it } from "bun:test";
-import { Effort } from "../src/model-thinking";
+import { Effort } from "../src/effort";
 import { streamAnthropic } from "../src/providers/anthropic";
 import type { Context, Model, Tool } from "../src/types";
 
diff --git a/packages/ai/test/issue-969-repro.test.ts b/packages/ai/test/issue-969-repro.test.ts
index 1b347470e..9f42a85bb 100644
--- a/packages/ai/test/issue-969-repro.test.ts
+++ b/packages/ai/test/issue-969-repro.test.ts
@@ -1,5 +1,6 @@
 import { afterEach, describe, expect, it } from "bun:test";
-import { Effort, getSupportedEfforts } from "../src/model-thinking";
+import { Effort } from "../src/effort";
+import { getSupportedEfforts } from "../src/model-thinking";
 import { streamOpenAICompletions } from "../src/providers/openai-completions";
 import type { Context, Model } from "../src/types";
 
diff --git a/packages/ai/test/model-thinking.test.ts b/packages/ai/test/model-thinking.test.ts
index 89b90da8e..e8475974c 100644
--- a/packages/ai/test/model-thinking.test.ts
+++ b/packages/ai/test/model-thinking.test.ts
@@ -1,8 +1,8 @@
 import { describe, expect, it } from "bun:test";
+import { Effort } from "@oh-my-pi/pi-ai/effort";
 import {
 	applyGeneratedModelPolicies,
 	clampThinkingLevelForModel,
-	Effort,
 	enrichModelThinking,
 	linkOpenAIPromotionTargets,
 	mapEffortToAnthropicAdaptiveEffort,
diff --git a/packages/ai/test/nanogpt-model-limits.test.ts b/packages/ai/test/nanogpt-model-limits.test.ts
index 03d7b8bab..0270b18ff 100644
--- a/packages/ai/test/nanogpt-model-limits.test.ts
+++ b/packages/ai/test/nanogpt-model-limits.test.ts
@@ -1,5 +1,5 @@
 import { afterEach, describe, expect, it, vi } from "bun:test";
-import { Effort } from "../src/model-thinking";
+import { Effort } from "../src/effort";
 import { nanoGptModelManagerOptions } from "../src/provider-models/openai-compat";
 
 const originalFetch = global.fetch;
diff --git a/packages/ai/test/ollama-provider.test.ts b/packages/ai/test/ollama-provider.test.ts
index 164d17812..4fa4866c2 100644
--- a/packages/ai/test/ollama-provider.test.ts
+++ b/packages/ai/test/ollama-provider.test.ts
@@ -1,5 +1,5 @@
 import { afterEach, describe, expect, test, vi } from "bun:test";
-import { Effort } from "../src/model-thinking";
+import { Effort } from "../src/effort";
 import { ollamaModelManagerOptions } from "../src/provider-models/openai-compat";
 import { streamOllama } from "../src/providers/ollama";
 import type { Context, Model, Tool } from "../src/types";
@@ -53,6 +53,48 @@ describe("ollama local provider discovery", () => {
 		expect(model?.thinking).toEqual({ mode: "effort", minLevel: Effort.Minimal, maxLevel: Effort.High });
 		expect(model?.input).toEqual(["text", "image"]);
 	});
+
+	test("remaps Ollama's unsupported reasoning levels and skips non-reasoning models", async () => {
+		global.fetch = vi.fn(async (input, init) => {
+			const url = String(input);
+			if (url === "http://127.0.0.1:11434/v1/models") {
+				return new Response(
+					JSON.stringify({
+						object: "list",
+						data: [
+							{ id: "gemma4:e4b", object: "model" },
+							{ id: "llama-plain:latest", object: "model" },
+						],
+					}),
+					{ status: 200, headers: { "Content-Type": "application/json" } },
+				);
+			}
+			if (url === "http://127.0.0.1:11434/api/show") {
+				const body = JSON.parse(String(init?.body ?? "{}")) as { model?: string };
+				const thinking = body.model === "gemma4:e4b";
+				return new Response(
+					JSON.stringify({
+						capabilities: thinking ? ["completion", "tools", "thinking"] : ["completion", "tools"],
+						model_info: {},
+					}),
+					{ status: 200, headers: { "Content-Type": "application/json" } },
+				);
+			}
+			throw new Error(`Unexpected URL: ${url}`);
+		}) as unknown as typeof fetch;
+
+		const models = await ollamaModelManagerOptions().fetchDynamicModels?.();
+		const reasoningModel = models?.find(candidate => candidate.id === "gemma4:e4b");
+		const plainModel = models?.find(candidate => candidate.id === "llama-plain:latest");
+
+		// Ollama's OpenAI-compatible endpoint rejects "minimal"/"xhigh" with HTTP 400;
+		// reasoning models must remap them onto accepted levels (low/max).
+		expect(reasoningModel?.reasoning).toBe(true);
+		expect(reasoningModel?.compat?.reasoningEffortMap).toMatchObject({ minimal: "low", xhigh: "max" });
+		// Non-reasoning models never send an effort, so they carry no remap.
+		expect(plainModel?.reasoning).toBe(false);
+		expect(plainModel?.compat?.reasoningEffortMap).toBeUndefined();
+	});
 });
 
 describe("ollama tool forcing", () => {
diff --git a/packages/ai/test/openai-codex.test.ts b/packages/ai/test/openai-codex.test.ts
index 4dbc616a7..ed81d7c78 100644
--- a/packages/ai/test/openai-codex.test.ts
+++ b/packages/ai/test/openai-codex.test.ts
@@ -71,6 +71,62 @@ describe("openai-codex request transformer", () => {
 	});
 });
 
+describe("openai-codex orphan tool-call repair", () => {
+	it("synthesizes a function_call_output for a function_call with no result", async () => {
+		const body: RequestBody = {
+			model: "gpt-5.1-codex",
+			input: [
+				{ type: "message", role: "user", content: [{ type: "input_text", text: "hi" }] },
+				{ type: "function_call", call_id: "call_orphan", name: "read", arguments: "{}" },
+				{ type: "message", role: "user", content: [{ type: "input_text", text: "next" }] },
+			],
+		};
+
+		const transformed = await transformRequestBody(body, createCodexModel(body.model), {});
+		const input = transformed.input || [];
+
+		const callIndex = input.findIndex(item => item.type === "function_call" && item.call_id === "call_orphan");
+		expect(callIndex).toBeGreaterThanOrEqual(0);
+		// The synthesized output sits immediately after the orphan call.
+		const output = input[callIndex + 1];
+		expect(output?.type).toBe("function_call_output");
+		expect(output?.call_id).toBe("call_orphan");
+		expect(typeof output?.output).toBe("string");
+		expect(output?.output as string).toMatch(/interrupted/i);
+	});
+
+	it("leaves a paired function_call untouched", async () => {
+		const body: RequestBody = {
+			model: "gpt-5.1-codex",
+			input: [
+				{ type: "function_call", call_id: "call_paired", name: "read", arguments: "{}" },
+				{ type: "function_call_output", call_id: "call_paired", output: "real result" },
+			],
+		};
+
+		const transformed = await transformRequestBody(body, createCodexModel(body.model), {});
+		const input = transformed.input || [];
+
+		const outputs = input.filter(item => item.type === "function_call_output" && item.call_id === "call_paired");
+		expect(outputs).toHaveLength(1);
+		expect(outputs[0]?.output).toBe("real result");
+	});
+
+	it("synthesizes a custom_tool_call_output for an orphan custom_tool_call", async () => {
+		const body: RequestBody = {
+			model: "gpt-5.1-codex",
+			input: [{ type: "custom_tool_call", call_id: "call_custom", name: "apply_patch" }],
+		};
+
+		const transformed = await transformRequestBody(body, createCodexModel(body.model), {});
+		const input = transformed.input || [];
+
+		const output = input.find(item => item.type === "custom_tool_call_output" && item.call_id === "call_custom");
+		expect(output).toBeDefined();
+		expect(output?.output as string).toMatch(/interrupted/i);
+	});
+});
+
 describe("openai-codex reasoning effort validation", () => {
 	it("rejects gpt-5.1 xhigh when metadata does not list it", async () => {
 		const body: RequestBody = { model: "gpt-5.1", input: [] };
diff --git a/packages/ai/test/openai-completions-disable-reasoning.test.ts b/packages/ai/test/openai-completions-disable-reasoning.test.ts
index f2d9107f6..629a74887 100644
--- a/packages/ai/test/openai-completions-disable-reasoning.test.ts
+++ b/packages/ai/test/openai-completions-disable-reasoning.test.ts
@@ -1,5 +1,5 @@
 import { afterEach, describe, expect, it } from "bun:test";
-import { Effort } from "../src/model-thinking";
+import { Effort } from "../src/effort";
 import { streamOpenAICompletions } from "../src/providers/openai-completions";
 import type { Context, Model } from "../src/types";
 
diff --git a/packages/ai/test/openai-responses-orphan-repair.test.ts b/packages/ai/test/openai-responses-orphan-repair.test.ts
new file mode 100644
index 000000000..014d6988f
--- /dev/null
+++ b/packages/ai/test/openai-responses-orphan-repair.test.ts
@@ -0,0 +1,69 @@
+import { describe, expect, it } from "bun:test";
+import {
+	repairOrphanResponsesToolCalls,
+	repairOrphanResponsesToolOutputs,
+} from "@oh-my-pi/pi-ai/providers/openai-responses-shared";
+import type { ResponseInput } from "openai/resources/responses/responses";
+
+describe("repairOrphanResponsesToolCalls", () => {
+	it("appends a synthetic function_call_output after a call with no result", () => {
+		const input: ResponseInput = [
+			{ type: "function_call", call_id: "call_a", name: "read", arguments: "{}" },
+			{ role: "user", content: [{ type: "input_text", text: "continue" }] },
+		];
+
+		const repaired = repairOrphanResponsesToolCalls(input);
+		const callIndex = repaired.findIndex(
+			item =>
+				(item as { type?: string }).type === "function_call" && (item as { call_id?: string }).call_id === "call_a",
+		);
+		const output = repaired[callIndex + 1] as { type?: string; call_id?: string; output?: unknown };
+		expect(output.type).toBe("function_call_output");
+		expect(output.call_id).toBe("call_a");
+		expect(output.output).toMatch(/interrupted/i);
+	});
+
+	it("uses custom_tool_call_output for an orphan custom_tool_call", () => {
+		const input: ResponseInput = [
+			{ type: "custom_tool_call", call_id: "call_c", name: "apply_patch", input: "patch" } as ResponseInput[number],
+		];
+
+		const repaired = repairOrphanResponsesToolCalls(input);
+		const output = repaired.find(item => (item as { type?: string }).type === "custom_tool_call_output") as
+			| { call_id?: string }
+			| undefined;
+		expect(output?.call_id).toBe("call_c");
+	});
+
+	it("returns the input unchanged when every call is paired", () => {
+		const input: ResponseInput = [
+			{ type: "function_call", call_id: "call_a", name: "read", arguments: "{}" },
+			{ type: "function_call_output", call_id: "call_a", output: "ok" } as ResponseInput[number],
+		];
+
+		const repaired = repairOrphanResponsesToolCalls(input);
+		expect(repaired).toBe(input);
+	});
+
+	it("composes with output repair so a tree-branch snapshot stays API-valid", () => {
+		// Branching to a node that ends on a tool call drops the result child:
+		// the assistant turn keeps the call, but no matching output remains.
+		const input: ResponseInput = [
+			{ role: "user", content: [{ type: "input_text", text: "do it" }] },
+			{ type: "function_call", call_id: "call_x", name: "bash", arguments: "{}" },
+		];
+
+		const repaired = repairOrphanResponsesToolCalls(repairOrphanResponsesToolOutputs(input));
+		const callIds = new Set(
+			repaired
+				.filter(i => (i as { type?: string }).type === "function_call")
+				.map(i => (i as { call_id: string }).call_id),
+		);
+		const outputIds = new Set(
+			repaired
+				.filter(i => (i as { type?: string }).type === "function_call_output")
+				.map(i => (i as { call_id: string }).call_id),
+		);
+		for (const id of callIds) expect(outputIds.has(id)).toBe(true);
+	});
+});
diff --git a/packages/ai/test/openai-responses-parallel-tool-calls.test.ts b/packages/ai/test/openai-responses-parallel-tool-calls.test.ts
index 56e076e6d..dbff9a657 100644
--- a/packages/ai/test/openai-responses-parallel-tool-calls.test.ts
+++ b/packages/ai/test/openai-responses-parallel-tool-calls.test.ts
@@ -202,4 +202,137 @@ describe("processResponsesStream: parallel function_call items", () => {
 		expect(byCallId.get("call_a")?.toolCall.arguments).toEqual({ path: "test.txt" });
 		expect(byCallId.get("call_b")?.toolCall.arguments).toEqual({ path: "test.md" });
 	});
+
+	test("routes identifierless final argument events in item order", async () => {
+		const output = makeOutput();
+		const emitted: EmittedEvent[] = [];
+		const stream = { push: (e: unknown) => emitted.push(e as EmittedEvent), end: () => {} } as never;
+
+		const argsA = JSON.stringify({ command: "printf a" });
+		const argsB = JSON.stringify({ command: "printf b" });
+
+		await processResponsesStream(
+			makeStream([
+				{
+					type: "response.output_item.added",
+					output_index: 0,
+					item: { type: "function_call", id: "fc_a", call_id: "call_a", name: "bash", arguments: "" },
+				},
+				{
+					type: "response.output_item.added",
+					output_index: 1,
+					item: { type: "function_call", id: "fc_b", call_id: "call_b", name: "bash", arguments: "" },
+				},
+				{
+					type: "response.function_call_arguments.done",
+					arguments: argsA,
+				},
+				{
+					type: "response.function_call_arguments.done",
+					arguments: argsB,
+				},
+				{
+					type: "response.output_item.done",
+					output_index: 0,
+					item: { type: "function_call", id: "fc_a", call_id: "call_a", name: "bash", arguments: "" },
+				},
+				{
+					type: "response.output_item.done",
+					output_index: 1,
+					item: { type: "function_call", id: "fc_b", call_id: "call_b", name: "bash", arguments: "" },
+				},
+			]),
+			output,
+			stream,
+			makeModel(),
+		);
+
+		const [blockA, blockB] = output.content;
+		if (blockA?.type !== "toolCall" || blockB?.type !== "toolCall") throw new Error("expected toolCalls");
+		expect(blockA.arguments).toEqual({ command: "printf a" });
+		expect(blockB.arguments).toEqual({ command: "printf b" });
+
+		const ends = emitted.filter(e => e.type === "toolcall_end") as Array<{
+			toolCall: { id: string; arguments: Record<string, unknown> };
+		}>;
+		expect(ends).toHaveLength(2);
+		const byCallId = new Map(ends.map(e => [e.toolCall.id.split("|")[0], e]));
+		expect(byCallId.get("call_a")?.toolCall.arguments).toEqual({ command: "printf a" });
+		expect(byCallId.get("call_b")?.toolCall.arguments).toEqual({ command: "printf b" });
+	});
+
+	test("routes deltas by item.call_id when llama.cpp omits item.id and output_index (issue #2015)", async () => {
+		// llama.cpp's `to_json_oaicompat_resp` (tools/server/server-task.cpp) emits a
+		// function_call's `output_item.added` with only `item.call_id` — no `item.id`,
+		// no `output_index`. The matching `function_call_arguments.delta` then carries
+		// `item_id: "fc_<call_id>"` and again no `output_index`. Without secondary
+		// indexing on `call_id`, `processResponsesStream`'s lookup map stays empty and
+		// every delta lands on the trailing block, leaving earlier calls with empty
+		// arguments (= `{}`) — the read tool then rejects them with
+		// `path: Invalid input: expected string, received undefined`.
+		const output = makeOutput();
+		const emitted: EmittedEvent[] = [];
+		const stream = { push: (e: unknown) => emitted.push(e as EmittedEvent), end: () => {} } as never;
+
+		const argsA = JSON.stringify({ path: "a.txt" });
+		const argsB = JSON.stringify({ path: "b.txt" });
+		const argsC = JSON.stringify({ path: "c.txt" });
+
+		await processResponsesStream(
+			makeStream([
+				{
+					type: "response.output_item.added",
+					item: { type: "function_call", call_id: "fc_a", name: "read", arguments: "" },
+				},
+				{
+					type: "response.output_item.added",
+					item: { type: "function_call", call_id: "fc_b", name: "read", arguments: "" },
+				},
+				{
+					type: "response.output_item.added",
+					item: { type: "function_call", call_id: "fc_c", name: "read", arguments: "" },
+				},
+				{ type: "response.function_call_arguments.delta", item_id: "fc_a", delta: argsA },
+				{ type: "response.function_call_arguments.delta", item_id: "fc_b", delta: argsB },
+				{ type: "response.function_call_arguments.delta", item_id: "fc_c", delta: argsC },
+				{
+					type: "response.output_item.done",
+					item: { type: "function_call", call_id: "fc_a", name: "read", arguments: argsA },
+				},
+				{
+					type: "response.output_item.done",
+					item: { type: "function_call", call_id: "fc_b", name: "read", arguments: argsB },
+				},
+				{
+					type: "response.output_item.done",
+					item: { type: "function_call", call_id: "fc_c", name: "read", arguments: argsC },
+				},
+			]),
+			output,
+			stream,
+			makeModel(),
+		);
+
+		expect(output.content).toHaveLength(3);
+		const [a, b, c] = output.content;
+		if (a?.type !== "toolCall" || b?.type !== "toolCall" || c?.type !== "toolCall") {
+			throw new Error("expected toolCalls");
+		}
+		expect(a.arguments).toEqual({ path: "a.txt" });
+		expect(b.arguments).toEqual({ path: "b.txt" });
+		expect(c.arguments).toEqual({ path: "c.txt" });
+
+		const ends = emitted.filter(e => e.type === "toolcall_end") as Array<{
+			toolCall: { id: string; arguments: Record<string, unknown> };
+			contentIndex: number;
+		}>;
+		expect(ends).toHaveLength(3);
+		const byCallId = new Map(ends.map(e => [e.toolCall.id.split("|")[0], e]));
+		expect(byCallId.get("fc_a")?.toolCall.arguments).toEqual({ path: "a.txt" });
+		expect(byCallId.get("fc_b")?.toolCall.arguments).toEqual({ path: "b.txt" });
+		expect(byCallId.get("fc_c")?.toolCall.arguments).toEqual({ path: "c.txt" });
+		expect(byCallId.get("fc_a")?.contentIndex).toBe(0);
+		expect(byCallId.get("fc_b")?.contentIndex).toBe(1);
+		expect(byCallId.get("fc_c")?.contentIndex).toBe(2);
+	});
 });
diff --git a/packages/ai/test/schema-normalization.test.ts b/packages/ai/test/schema-normalization.test.ts
index 1f6ba7688..22efcb2cd 100644
--- a/packages/ai/test/schema-normalization.test.ts
+++ b/packages/ai/test/schema-normalization.test.ts
@@ -728,6 +728,18 @@ describe("stripResidualCombiners", () => {
 		expect(normalized.anyOf).toBeUndefined();
 		expect(normalized.oneOf).toBeUndefined();
 	});
+
+	it("drops array-only keys when mixed-type collapse picks string from anyOf fixpoint", () => {
+		const stripped = stripResidualCombiners({
+			anyOf: [{ type: "string" }, { type: "array", items: { type: "string" } }],
+			description: "pr number, url, or branch",
+		}) as Record<string, unknown>;
+
+		expect(stripped.type).toBe("string");
+		expect(stripped.items).toBeUndefined();
+		expect(stripped.anyOf).toBeUndefined();
+		expect(stripped.description).toBe("pr number, url, or branch");
+	});
 });
 
 // ---------------------------------------------------------------------------
@@ -952,6 +964,36 @@ describe("normalizeSchemaForCCA", () => {
 			properties: {},
 		});
 	});
+
+	it("strips array-only keys when mixed-type collapse picks a non-array type", () => {
+		// Regression: anyOf [{type:"string"}, {type:"array", items:{type:"string"}}]
+		// collapsed to {type:"string", items:{type:"string"}} which is invalid.
+		// The fix filters mergedVariantFields against the chosen type's allowed keys.
+		const normalized = normalizeSchemaForCCA({
+			anyOf: [{ type: "string" }, { type: "array", items: { type: "string" } }],
+			description: "pr number, url, or branch",
+		});
+
+		expect(normalized).toEqual({
+			type: "string",
+			description: "pr number, url, or branch",
+		});
+	});
+
+	it("strips sibling type-specific keys copied from parent when mixed-type collapse picks opposing type", () => {
+		// Edge case: parent has a sibling `items` outside the anyOf,
+		// and the chosen type is string. The sibling must be stripped.
+		const normalized = normalizeSchemaForCCA({
+			anyOf: [{ type: "string" }, { type: "array", items: { type: "number" } }],
+			items: { type: "string" },
+			description: "pr number, url, or branch",
+		});
+
+		expect(normalized).toEqual({
+			type: "string",
+			description: "pr number, url, or branch",
+		});
+	});
 });
 
 // ---------------------------------------------------------------------------
diff --git a/packages/ai/test/stream-auth-retry.test.ts b/packages/ai/test/stream-auth-retry.test.ts
index d019e57ce..7192cb8c8 100644
--- a/packages/ai/test/stream-auth-retry.test.ts
+++ b/packages/ai/test/stream-auth-retry.test.ts
@@ -1,4 +1,5 @@
 import { afterEach, describe, expect, it } from "bun:test";
+import type { ApiKeyResolveContext } from "@oh-my-pi/pi-ai";
 import { registerCustomApi, unregisterCustomApis } from "@oh-my-pi/pi-ai";
 import { streamSimple } from "@oh-my-pi/pi-ai/stream";
 import type { Api, AssistantMessage, Context, Model, SimpleStreamOptions, Usage } from "@oh-my-pi/pi-ai/types";
@@ -25,9 +26,9 @@ function assistant(content: string[] = []): AssistantMessage {
 		api: API,
 		provider: "test-provider",
 		model: "test-model",
-		usage: usage(),
+		timestamp: 1,
 		stopReason: "stop",
-		timestamp: Date.now(),
+		usage: usage(),
 	};
 }
 
@@ -39,19 +40,21 @@ function authError(): Error & { status: number } {
 	return Object.assign(new Error("401 authentication_error"), { status: 401 });
 }
 
+function usageLimitError(): Error & { status: number } {
+	return Object.assign(new Error("You have hit your ChatGPT usage limit (pro plan). Try again in ~158 min."), {
+		status: 429,
+	});
+}
+
 function model(): Model<Api> {
 	return {
 		id: "test-model",
-		name: "test-model",
+		name: "Test Model",
 		api: API,
 		provider: "test-provider",
-		baseUrl: "mock://",
-		reasoning: false,
-		input: ["text"],
-		cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
-		contextWindow: 1024,
-		maxTokens: 1024,
-	};
+		contextWindow: 1000,
+		maxTokens: 100,
+	} as Model<Api>;
 }
 
 const context: Context = {
@@ -59,61 +62,64 @@ const context: Context = {
 	messages: [{ role: "user", content: "hello", timestamp: 1 }],
 };
 
-describe("streamSimple auth retry", () => {
+/** Records the static key each inner attempt actually received. */
+function pushKey(keys: unknown[], options?: SimpleStreamOptions): void {
+	keys.push(options?.apiKey);
+}
+
+function ok(stream: AssistantMessageEventStream): void {
+	const message = assistant(["ok"]);
+	stream.push({ type: "start", partial: message });
+	stream.push({ type: "done", reason: "stop", message });
+}
+
+describe("streamSimple resolver auth retry", () => {
 	afterEach(() => {
 		unregisterCustomApis(SOURCE_ID);
 	});
 
-	it("retries once with a fresh key when 401 happens before the first event", async () => {
-		const keys: Array<string | undefined> = [];
-		let authCalls = 0;
+	it("retries with a refreshed key when a 401 is thrown before the first event", async () => {
+		const keys: unknown[] = [];
+		const contexts: ApiKeyResolveContext[] = [];
 		registerCustomApi(
 			API,
 			(_model: Model<Api>, _context: Context, options?: SimpleStreamOptions) => {
-				keys.push(options?.apiKey);
+				pushKey(keys, options);
 				const stream = new AssistantMessageEventStream();
-				queueMicrotask(() => {
-					if (keys.length === 1) {
-						stream.fail(authError());
-						return;
-					}
-					const message = assistant(["ok"]);
-					stream.push({ type: "start", partial: message });
-					stream.push({ type: "done", reason: "stop", message });
-				});
+				queueMicrotask(() => (keys.length === 1 ? stream.fail(authError()) : ok(stream)));
 				return stream;
 			},
 			SOURCE_ID,
 		);
 
 		const stream = streamSimple(model(), context, {
-			apiKey: "old-key",
-			onAuthError: async (provider, oldKey, error) => {
-				authCalls += 1;
-				expect(provider).toBe("test-provider");
-				expect(oldKey).toBe("old-key");
-				expect((error as { status?: number }).status).toBe(401);
-				return "new-key";
+			apiKey: async ctx => {
+				contexts.push(ctx);
+				return ctx.error === undefined ? "old-key" : ctx.lastChance ? "switch-key" : "refresh-key";
 			},
 		});
-
 		for await (const _event of stream) {
 			// drain
 		}
 
 		expect((await stream.result()).content).toEqual([{ type: "text", text: "ok" }]);
-		expect(keys).toEqual(["old-key", "new-key"]);
-		expect(authCalls).toBe(1);
+		// Initial resolve, then step (b) refresh-same — the switch step is never reached.
+		expect(keys).toEqual(["old-key", "refresh-key"]);
+		expect(keys.every(key => typeof key === "string")).toBe(true);
+		expect(contexts.map(ctx => ({ lastChance: ctx.lastChance, hasError: ctx.error !== undefined }))).toEqual([
+			{ lastChance: false, hasError: false },
+			{ lastChance: false, hasError: true },
+		]);
+		expect((contexts[1]?.error as { status?: number }).status).toBe(401);
 	});
 
-	it("retries when a provider emits start then a 401 error event before content", async () => {
-		const keys: Array<string | undefined> = [];
+	it("buffers the start event and retries on a 401 error event before content", async () => {
+		const keys: unknown[] = [];
 		const eventTypes: string[] = [];
-		let authCalls = 0;
 		registerCustomApi(
 			API,
 			(_model: Model<Api>, _context: Context, options?: SimpleStreamOptions) => {
-				keys.push(options?.apiKey);
+				pushKey(keys, options);
 				const stream = new AssistantMessageEventStream();
 				queueMicrotask(() => {
 					if (keys.length === 1) {
@@ -127,9 +133,7 @@ describe("streamSimple auth retry", () => {
 						});
 						return;
 					}
-					const message = assistant(["ok"]);
-					stream.push({ type: "start", partial: message });
-					stream.push({ type: "done", reason: "stop", message });
+					ok(stream);
 				});
 				return stream;
 			},
@@ -137,28 +141,59 @@ describe("streamSimple auth retry", () => {
 		);
 
 		const stream = streamSimple(model(), context, {
-			apiKey: "old-key",
-			onAuthError: async (provider, oldKey, error) => {
-				authCalls += 1;
-				expect(provider).toBe("test-provider");
-				expect(oldKey).toBe("old-key");
-				expect((error as { status?: number }).status).toBe(401);
-				return "new-key";
-			},
+			apiKey: async ctx => (ctx.error === undefined ? "old-key" : "new-key"),
 		});
-
 		for await (const event of stream) {
 			eventTypes.push(event.type);
 		}
 
 		expect((await stream.result()).content).toEqual([{ type: "text", text: "ok" }]);
 		expect(keys).toEqual(["old-key", "new-key"]);
+		// The buffered `start` of the failed attempt must not leak — the user
+		// sees exactly one clean start/done pair.
 		expect(eventTypes).toEqual(["start", "done"]);
-		expect(authCalls).toBe(1);
+	});
+
+	it("retries on a 401 carried only via errorStatus", async () => {
+		const keys: unknown[] = [];
+		registerCustomApi(
+			API,
+			(_model: Model<Api>, _context: Context, options?: SimpleStreamOptions) => {
+				pushKey(keys, options);
+				const stream = new AssistantMessageEventStream();
+				queueMicrotask(() => {
+					if (keys.length === 1) {
+						stream.push({ type: "start", partial: assistant() });
+						stream.push({
+							type: "error",
+							reason: "error",
+							error: assistantError(
+								'{"type":"error","error":{"type":"authentication_error","message":"Invalid authentication credentials"}}',
+								401,
+							),
+						});
+						return;
+					}
+					ok(stream);
+				});
+				return stream;
+			},
+			SOURCE_ID,
+		);
+
+		const stream = streamSimple(model(), context, {
+			apiKey: async ctx => (ctx.error === undefined ? "old-key" : "new-key"),
+		});
+		for await (const _event of stream) {
+			// drain
+		}
+
+		expect((await stream.result()).content).toEqual([{ type: "text", text: "ok" }]);
+		expect(keys).toEqual(["old-key", "new-key"]);
 	});
 
 	it("does not retry after replay-unsafe content has been emitted", async () => {
-		let authCalls = 0;
+		let retryResolves = 0;
 		const failure = authError();
 		registerCustomApi(
 			API,
@@ -175,10 +210,9 @@ describe("streamSimple auth retry", () => {
 		);
 
 		const stream = streamSimple(model(), context, {
-			apiKey: "old-key",
-			onAuthError: async () => {
-				authCalls += 1;
-				return "new-key";
+			apiKey: async ctx => {
+				if (ctx.error !== undefined) retryResolves += 1;
+				return ctx.error === undefined ? "old-key" : "new-key";
 			},
 		});
 
@@ -192,94 +226,79 @@ describe("streamSimple auth retry", () => {
 		}
 
 		expect(caught).toBe(failure);
-		expect(authCalls).toBe(0);
+		// The resolver is never asked for a retry key once a replay-unsafe event shipped.
+		expect(retryResolves).toBe(0);
 	});
 
-	it("retries on 401 carried via errorStatus when the message has no parseable status", async () => {
-		const keys: Array<string | undefined> = [];
-		let authCalls = 0;
+	it("escalates refresh-same then switch in order (2-retry ordering)", async () => {
+		const keys: unknown[] = [];
 		registerCustomApi(
 			API,
 			(_model: Model<Api>, _context: Context, options?: SimpleStreamOptions) => {
-				keys.push(options?.apiKey);
+				pushKey(keys, options);
 				const stream = new AssistantMessageEventStream();
-				queueMicrotask(() => {
-					if (keys.length === 1) {
-						stream.push({ type: "start", partial: assistant() });
-						stream.push({
-							type: "error",
-							reason: "error",
-							// Realistic Anthropic SDK shape: message begins with `<status> <body>`
-							// and the regex fallback in extractHttpStatusFromError cannot find 401
-							// inside the JSON body. Only `errorStatus` carries the signal.
-							error: assistantError(
-								'{"type":"error","error":{"type":"authentication_error","message":"Invalid authentication credentials"}}',
-								401,
-							),
-						});
-						return;
-					}
-					const message = assistant(["ok"]);
-					stream.push({ type: "start", partial: message });
-					stream.push({ type: "done", reason: "stop", message });
-				});
+				queueMicrotask(() => (options?.apiKey === "switch-key" ? ok(stream) : stream.fail(authError())));
 				return stream;
 			},
 			SOURCE_ID,
 		);
 
 		const stream = streamSimple(model(), context, {
-			apiKey: "old-key",
-			onAuthError: async () => {
-				authCalls += 1;
-				return "new-key";
-			},
+			apiKey: async ctx => (ctx.error === undefined ? "old-key" : ctx.lastChance ? "switch-key" : "refresh-key"),
 		});
-
 		for await (const _event of stream) {
 			// drain
 		}
 
 		expect((await stream.result()).content).toEqual([{ type: "text", text: "ok" }]);
-		expect(keys).toEqual(["old-key", "new-key"]);
-		expect(authCalls).toBe(1);
+		expect(keys).toEqual(["old-key", "refresh-key", "switch-key"]);
 	});
 
-	it("retries on a thrown usage_limit_reached error before any event has been emitted", async () => {
-		const keys: Array<string | undefined> = [];
-		const errors: unknown[] = [];
+	it("skips the refresh-same step when the resolver returns an unchanged key", async () => {
+		const keys: unknown[] = [];
 		registerCustomApi(
 			API,
 			(_model: Model<Api>, _context: Context, options?: SimpleStreamOptions) => {
-				keys.push(options?.apiKey);
+				pushKey(keys, options);
 				const stream = new AssistantMessageEventStream();
-				queueMicrotask(() => {
-					if (keys.length === 1) {
-						stream.fail(
-							Object.assign(
-								new Error("You have hit your ChatGPT usage limit (pro plan). Try again in ~158 min."),
-								{ status: 429 },
-							),
-						);
-						return;
-					}
-					const message = assistant(["ok"]);
-					stream.push({ type: "start", partial: message });
-					stream.push({ type: "done", reason: "stop", message });
-				});
+				queueMicrotask(() => (options?.apiKey === "switch-key" ? ok(stream) : stream.fail(authError())));
 				return stream;
 			},
 			SOURCE_ID,
 		);
 
 		const stream = streamSimple(model(), context, {
-			apiKey: "old-key",
-			onAuthError: async (_provider, _oldKey, error) => {
-				errors.push(error);
-				return "new-key";
+			// refresh-same yields the same failing key → that attempt is skipped.
+			apiKey: async ctx => (ctx.error === undefined ? "old-key" : ctx.lastChance ? "switch-key" : "old-key"),
+		});
+		for await (const _event of stream) {
+			// drain
+		}
+
+		expect((await stream.result()).content).toEqual([{ type: "text", text: "ok" }]);
+		expect(keys).toEqual(["old-key", "switch-key"]);
+	});
+
+	it("retries a thrown usage-limit error and passes the cause to the resolver", async () => {
+		const keys: unknown[] = [];
+		const errors: unknown[] = [];
+		registerCustomApi(
+			API,
+			(_model: Model<Api>, _context: Context, options?: SimpleStreamOptions) => {
+				pushKey(keys, options);
+				const stream = new AssistantMessageEventStream();
+				queueMicrotask(() => (keys.length === 1 ? stream.fail(usageLimitError()) : ok(stream)));
+				return stream;
+			},
+			SOURCE_ID,
+		);
+
+		const stream = streamSimple(model(), context, {
+			apiKey: async ctx => {
+				if (ctx.error !== undefined) errors.push(ctx.error);
+				return ctx.error === undefined ? "old-key" : "new-key";
 			},
 		});
-
 		for await (const _event of stream) {
 			// drain
 		}
@@ -287,18 +306,17 @@ describe("streamSimple auth retry", () => {
 		expect((await stream.result()).content).toEqual([{ type: "text", text: "ok" }]);
 		expect(keys).toEqual(["old-key", "new-key"]);
 		expect(errors).toHaveLength(1);
-		// The error surfaced to onAuthError carries the original 429 status so
-		// the gateway's refresh hook can branch on it.
+		// The cause carries the original 429 so the resolver can branch usage-limit vs 401.
 		expect((errors[0] as { status?: number }).status).toBe(429);
 		expect((errors[0] as Error).message).toMatch(/usage limit/i);
 	});
 
-	it("retries when a provider emits a usage_limit_reached error event before content", async () => {
-		const keys: Array<string | undefined> = [];
+	it("retries a usage-limit error event before content", async () => {
+		const keys: unknown[] = [];
 		registerCustomApi(
 			API,
 			(_model: Model<Api>, _context: Context, options?: SimpleStreamOptions) => {
-				keys.push(options?.apiKey);
+				pushKey(keys, options);
 				const stream = new AssistantMessageEventStream();
 				queueMicrotask(() => {
 					if (keys.length === 1) {
@@ -306,16 +324,11 @@ describe("streamSimple auth retry", () => {
 						stream.push({
 							type: "error",
 							reason: "error",
-							// errorStatus deliberately omitted: matches how the codex
-							// provider serializes the message-only path that used to
-							// 502 in the gateway.
 							error: assistantError("You have hit your ChatGPT usage limit (pro plan). Try again in ~158 min."),
 						});
 						return;
 					}
-					const message = assistant(["ok"]);
-					stream.push({ type: "start", partial: message });
-					stream.push({ type: "done", reason: "stop", message });
+					ok(stream);
 				});
 				return stream;
 			},
@@ -323,10 +336,8 @@ describe("streamSimple auth retry", () => {
 		);
 
 		const stream = streamSimple(model(), context, {
-			apiKey: "old-key",
-			onAuthError: async () => "new-key",
+			apiKey: async ctx => (ctx.error === undefined ? "old-key" : "new-key"),
 		});
-
 		for await (const _event of stream) {
 			// drain
 		}
@@ -335,13 +346,13 @@ describe("streamSimple auth retry", () => {
 		expect(keys).toEqual(["old-key", "new-key"]);
 	});
 
-	it("surfaces the original usage_limit error when the retry callback declines", async () => {
-		const keys: Array<string | undefined> = [];
-		const original = Object.assign(new Error("You have hit your ChatGPT usage limit (pro plan)."), { status: 429 });
+	it("surfaces the original error when the resolver declines every retry", async () => {
+		const keys: unknown[] = [];
+		const original = usageLimitError();
 		registerCustomApi(
 			API,
 			(_model: Model<Api>, _context: Context, options?: SimpleStreamOptions) => {
-				keys.push(options?.apiKey);
+				pushKey(keys, options);
 				const stream = new AssistantMessageEventStream();
 				queueMicrotask(() => stream.fail(original));
 				return stream;
@@ -349,12 +360,9 @@ describe("streamSimple auth retry", () => {
 			SOURCE_ID,
 		);
 
-		// Callback returns undefined → no sibling credential to rotate to.
-		// The original failure must reach the caller untouched so the client
-		// can decide what to do (back off, surface to user, …).
 		const stream = streamSimple(model(), context, {
-			apiKey: "old-key",
-			onAuthError: async () => undefined,
+			// Decline all retries: no sibling credential to rotate to.
+			apiKey: async ctx => (ctx.error === undefined ? "old-key" : undefined),
 		});
 
 		let caught: unknown;
@@ -367,8 +375,32 @@ describe("streamSimple auth retry", () => {
 		}
 
 		expect(caught).toBe(original);
-		// Single attempt — the inner stream is only re-invoked when a new key
-		// is provided.
 		expect(keys).toEqual(["old-key"]);
 	});
+
+	it("fails the stream when the initial resolve yields no key", async () => {
+		let attempts = 0;
+		registerCustomApi(
+			API,
+			() => {
+				attempts += 1;
+				return new AssistantMessageEventStream();
+			},
+			SOURCE_ID,
+		);
+
+		const stream = streamSimple(model(), context, { apiKey: async () => undefined });
+
+		let caught: unknown;
+		try {
+			for await (const _event of stream) {
+				// drain
+			}
+		} catch (error) {
+			caught = error;
+		}
+
+		expect((caught as Error).message).toMatch(/No API key for provider/);
+		expect(attempts).toBe(0);
+	});
 });
diff --git a/packages/ai/test/tool-argument-coercion.test.ts b/packages/ai/test/tool-argument-coercion.test.ts
index 224159d5b..e0ca64378 100644
--- a/packages/ai/test/tool-argument-coercion.test.ts
+++ b/packages/ai/test/tool-argument-coercion.test.ts
@@ -78,6 +78,178 @@ describe("Tool argument coercion", () => {
 		expect(result.paths).toEqual(["src/**/*.ts"]);
 	});
 
+	it("wraps a singleton object in an array when schema expects object array", () => {
+		const tool: Tool = {
+			name: "todo_like",
+			description: "",
+			parameters: z.object({
+				ops: z.array(
+					z.object({
+						op: z.literal("init"),
+						list: z.array(
+							z.object({
+								phase: z.string(),
+								items: z.array(z.string()),
+							}),
+						),
+					}),
+				),
+			}),
+		};
+
+		const result = validateToolArguments(tool, {
+			type: "toolCall",
+			id: "call-singleton-object-array",
+			name: "todo_like",
+			arguments: {
+				ops: {
+					op: "init",
+					list: [{ phase: "Repro", items: ["capture"] }],
+				},
+			},
+		});
+
+		expect(result).toEqual({
+			ops: [{ op: "init", list: [{ phase: "Repro", items: ["capture"] }] }],
+		});
+	});
+
+	it("wraps a singleton number in an array when schema expects number array", () => {
+		const tool: Tool = {
+			name: "numeric_list",
+			description: "",
+			parameters: z.object({ values: z.array(z.number()) }),
+		};
+
+		const result = validateToolArguments(tool, {
+			type: "toolCall",
+			id: "call-singleton-number-array",
+			name: "numeric_list",
+			arguments: { values: 7 },
+		});
+
+		expect(result).toEqual({ values: [7] });
+	});
+
+	it("does not wrap singleton values for array expectations from failed union branches", () => {
+		const entry = z.object({ id: z.number() });
+		const tool: Tool = {
+			name: "union_shape",
+			description: "",
+			parameters: z.union([z.array(entry), entry]),
+		};
+
+		const result = validateToolArguments(tool, {
+			type: "toolCall",
+			id: "call-union-shape",
+			name: "union_shape",
+			arguments: { id: "1" },
+		});
+
+		expect(result).toEqual({ id: 1 });
+	});
+
+	it("does not wrap singleton values for JSON Schema anyOf array branches", () => {
+		const tool: Tool = {
+			name: "json_schema_union",
+			description: "",
+			parameters: {
+				type: "object",
+				properties: {
+					target: {
+						anyOf: [
+							{
+								type: "array",
+								items: {
+									type: "object",
+									properties: { a: { type: "boolean" } },
+									required: ["a"],
+									additionalProperties: false,
+								},
+							},
+							{
+								type: "object",
+								properties: { a: { type: "boolean" } },
+								required: ["a"],
+								additionalProperties: false,
+							},
+						],
+					},
+				},
+				required: ["target"],
+				additionalProperties: false,
+			},
+		};
+
+		// The bug would silently coerce `{ a: "true" }` into `[{ a: true }]` by
+		// wrapping the object to satisfy the failed `anyOf` array branch and
+		// then coercing the inner string into a boolean. Branch tracking keeps
+		// the wrap from firing so the wrong shape never makes it through.
+		expect(() =>
+			validateToolArguments(tool, {
+				type: "toolCall",
+				id: "call-jsonschema-union",
+				name: "json_schema_union",
+				arguments: { target: { a: "true" } },
+			}),
+		).toThrow("Validation failed");
+	});
+
+	it("still wraps nested array fields inside a tag-selected Zod union branch", () => {
+		const tool: Tool = {
+			name: "tagged_union",
+			description: "",
+			parameters: z.union([
+				z.object({ type: z.literal("indices"), indices: z.array(z.number()) }),
+				z.object({ type: z.literal("all") }),
+			]),
+		};
+
+		const result = validateToolArguments(tool, {
+			type: "toolCall",
+			id: "call-tagged-union",
+			name: "tagged_union",
+			arguments: { type: "indices", indices: 1 },
+		});
+
+		expect(result).toEqual({ type: "indices", indices: [1] });
+	});
+
+	it("still wraps nested array fields inside a tag-selected JSON Schema anyOf branch", () => {
+		const tool: Tool = {
+			name: "tagged_json_union",
+			description: "",
+			parameters: {
+				anyOf: [
+					{
+						type: "object",
+						properties: {
+							type: { const: "indices" },
+							indices: { type: "array", items: { type: "number" } },
+						},
+						required: ["type", "indices"],
+						additionalProperties: false,
+					},
+					{
+						type: "object",
+						properties: { type: { const: "all" } },
+						required: ["type"],
+						additionalProperties: false,
+					},
+				],
+			},
+		};
+
+		const result = validateToolArguments(tool, {
+			type: "toolCall",
+			id: "call-tagged-json-union",
+			name: "tagged_json_union",
+			arguments: { type: "indices", indices: 1 },
+		});
+
+		expect(result).toEqual({ type: "indices", indices: [1] });
+	});
+
 	it("parses JSON objects in string values when schema expects object", () => {
 		const tool: Tool = {
 			name: "t4",
diff --git a/packages/coding-agent/CHANGELOG.md b/packages/coding-agent/CHANGELOG.md
index 95733447a..0243ed841 100644
--- a/packages/coding-agent/CHANGELOG.md
+++ b/packages/coding-agent/CHANGELOG.md
@@ -1,6 +1,178 @@
 # Changelog
 
 ## [Unreleased]
+
+## [15.10.1] - 2026-06-07
+
+### Added
+
+- Added `display.smoothStreaming` setting (default `true`) to let users enable or disable smooth assistant-stream text reveal
+- Added `/tan <work>` slash command to fork the current conversation into a background agent so tangential work can continue asynchronously while your main session stays active
+- Added a background `/tan` dispatch message that records the handoff in the transcript and marks the delegated work as non-blocking
+- Added `providerPromptCacheKey` support to `CreateAgentSessionOptions` so `/tan` background sessions can reuse the parent session’s prompt-cache lineage
+- Added session cloning for `/tan` runs with copied artifacts and shared MCP proxy tools
+- Added `SessionManager.forkFrom`’s optional `suppressBreadcrumb` mode to avoid breadcrumb updates when forking background `/tan` sessions
+- Added OSC 5522 enhanced paste handling in `InputController`, so terminal clipboard events are decoded as image or text payloads and inserted without passing raw paste sequences to the editor
+- Added bracketed image-path paste support in `CustomEditor` so a single pasted image file path (PNG/JPEG/GIF/WEBP) is loaded from disk and inserted as an image candidate
+- Added direct support for `Image #N` insertion from pasted local image paths by routing successful image-path pastes through the same image normalization and resize flow as clipboard image pastes
+- Added `/fresh` to rotate the provider-facing session id and clear in-memory provider stream/cache state without changing the local session file.
+- Added a `ChatBlock` transcript primitive (`modes/components/chat-block.ts`) and a single `ctx.present(...)` sink (with `ctx.resetTranscript()`) so chat output is mounted in one place instead of the repeated `chatContainer.addChild(...)` + `ui.requestRender()` pattern scattered across controllers. `ChatBlock` carries a React/Svelte-style lifecycle — `onMount` starts effects, `onCleanup` registers teardown, `finish()` self-completes (stops timers and freezes the block at its final content), and `dispose()`/`resetTranscript()` tears everything down — so animated blocks own their own resources instead of leaking `setInterval`/`requestRender` bookkeeping into callers. The MCP "Connecting…" spinner is now such a block.
+- Added a `framedBlock` output-block helper (`tui/output-block.ts`) plus a `borderColor` override and `applyBg: false` (no background fill) on output blocks, a `renderStatusLine` `iconOverride`, and an `icon.search` (magnifier) theme symbol — so tool renderers can draw self-contained muted-outline frames and search-family tools can show a magnifier instead of a checkmark.
+
+### Changed
+
+- Changed the bash tool frame to use a plain top rule instead of repeating "Bash" in the title bar, and folded minimizer raw-output artifact links into the status footer as `Artifact: <id>`.
+
+- Changed grouped `read` output to use a white filled-circle mark for the group/single-read success state and omit duplicate per-file success marks inside multi-read groups.
+
+- Changed assistant streaming output to reveal text incrementally at 30 FPS with grapheme-safe adaptive catch-up, instead of replacing the whole message chunk-by-chunk
+- Changed shimmer-driven TUI animations (working text, pending bash/eval borders, and theme activity-spinner documentation) to render at 30fps instead of 60fps.
+- Changed running `task` tool agent rows to use a static `•` marker and shimmer only the subagent name, leaving descriptions, stats, and nested tool detail text solid while removing the rotating status glyph from those rows.
+- Changed settings singleton method access to reuse bound methods for the active instance instead of allocating a new bound function on every `settings.get` lookup.
+- Changed plan-mode approval to keep the drafted `local://<slug>-plan.md` file at its original name as the canonical plan path, so approved plans are no longer renamed when leaving plan mode
+- Changed plan-mode write enforcement so only `local://` artifact files are writable during planning, blocking working-tree edits and allowing scratch or draft plan files in the local artifact area
+- Changed the `todo` tool result renderer to stop redrawing every phase's full task list on each update: when a multi-phase list is rendered collapsed (the default, not manually expanded), only phases the latest update touched — the phase holding the in_progress task, any phase with a just-completed task, and phases named by the ops that ran (`init` counts as touching all) — render their tasks; untouched phases collapse to a one-line `N. Name  done/total` summary. When call args are unavailable (e.g. transcript rebuilds) it falls back to the in_progress/completed-transition signals, and the manual expand toggle still shows every task. Also dropped the blank separator line previously inserted between phases.
+- Changed non-agent API operations (title and commit-message generation, image generation, web search, eval `llm()`, auto-thinking classifier, memory consolidation) to use session-aware API key resolution with auth retries via `registry.resolver()` / `authStorage.resolver()`, refreshing the active credential before rotating to another account
+- Changed image generation to wrap every provider fetch branch in `withAuth`, so 401 / usage-limit errors trigger credential force-refresh and rotation for authStorage-backed providers (OpenAI-hosted, antigravity, xai-oauth) while env-only providers (openrouter, gemini) stay single-attempt
+- Changed web-search providers using `authStorage.getApiKey` (anthropic, exa, tavily, parallel, synthetic, zai, kimi) to wrap HTTP calls in `withAuth` for automatic credential rotation on 401 / usage-limit errors
+- Changed the directory grouping for `find`, `search`, `ast_grep`, `ast_edit`, and `lsp` diagnostics from a single flat `# dir/` heading per immediate directory to a multi-level tree that folds the common path prefix into one heading. Previously every group repeated the full directory path — so results rooted outside cwd printed the absolute prefix (e.g. `/Users/me/proj/`) on every heading and nested directories were never collapsed. Now a single-child directory chain folds into one heading (`# packages/pkg/src/`, including an absolute root for out-of-cwd results), subdirectories nest one `#` deeper (`## nested/` → `### child.ts`), and each directory's own files are listed before its subdirectories. TUI hyperlink reconstruction tracks the nested directory stack across the whole output so file and code-frame links keep resolving to the correct absolute paths.
+- Changed the plan-mode approval surface from an inline transcript block plus a separate bottom selector into a single fullscreen overlay (like `/copy`) and overhauled its navigation. The overlay now renders the plan per-section through `ScrollView` (line-level ↑/↓ scroll, Shift+↑/↓ to scroll faster, PgUp/PgDn, g/G) with no stray per-line `…`, and — when the terminal is wide enough and the plan has ≥2 headings — shows a compact VS Code-style section sidebar (the redundant plan-title heading and any "Contents" label are omitted). Focus moves between regions with Tab/Shift+Tab (and flows at the edges: Down past the last section or the bottom of the body drops into the approval options; Up steps back), while the sidebar glows to track the scrolled section. The sidebar can fast-jump between sections, delete a section (with `u` undo), and annotate sections with feedback (`a`); deletions and annotations are collected into refinement feedback that is submitted back to the model when the operator picks "Refine plan". Mouse works too: clicking an approval option activates it, clicking a sidebar section jumps to it, and the wheel scrolls the plan. ←/→ always drive the model-tier slider, Enter confirms, the external-editor key opens the plan, and Esc cancels. The overlay borrows the terminal's alternate screen buffer for its lifetime (`fullscreen` overlay), so the transcript stays put on the normal screen instead of bleeding through scrollback behind the modal.
+- Changed the interactive controllers (command, MCP, selector, extension-UI, event), debug panels, and the status/error/warning helpers to render chat output through `ctx.present(...)` instead of appending to `chatContainer` and calling `ui.requestRender()` directly; transcript rebuilds dispose live blocks via `ctx.resetTranscript()` so animated blocks' timers stop on reset.
+- Changed tool-execution block rendering so the container (`ToolExecutionComponent`) is a transparent passthrough — it no longer inserts a top/bottom blank line, adds left/right padding, or paints a state-colored background behind tool output. Tools with substantial body now self-frame with a muted outline and the tool title in the frame's top bar (`edit`/`apply_patch`, `write`, `ask`, `todo`, `github`, `goal`, `inspect_image`, `search_tool_bm25`, `task`), matching the already-framed `bash`/`read`/`eval`/`debug`/`web_search`/`lsp` blocks, while streaming/in-progress and trivial results collapse to a clean status line. The search-family list tools (`find`, `search`, `ast_grep`) and `job` render frameless/minimal; `find`/`search`/`ast_grep` show a magnifier on success instead of a checkmark, and `job` drops its `Job:` label prefix (the per-job rows are self-describing). The `search_tool_bm25`, `github`, and `inspect_image` frames draw with no background fill, and `inspect_image`'s label was shortened to `Inspect`.
+- Changed the plan-mode active prompt (`prompts/system/plan-mode-active.md`) to make plans decision-complete and cut filler. Added an Objective framing ("another engineer can execute end-to-end without making a single design decision"), a shared "Resolving Unknowns" section (explore discoverable facts before asking; reserve `ask` for non-derivable preferences/tradeoffs with 2–4 options + a recommended default), and a single shared "The Plan" structure (Context / Approach grouped by behavior not file-by-file / ≤5 Critical files / Verification / Assumptions) that replaces the per-branch structure guidance previously duplicated across the iterative and parallel workflows. Added explicit prohibitions on sections that decide nothing (Non-Goals, Out of Scope, Alternatives Considered, Risks/Mitigations boilerplate, Future Work), on enumerating every file/line, and on inventing schema/validation/precedence policy the request never established.
+- Changed completion notifications (`completion.notify`) to fire whenever the agent yields its turn, including in the foreground. The `agent_end` notification was previously gated behind background mode (`isBackgrounded`), so an ordinary foreground turn never emitted one; the gate is gone and the desktop toast now fires on every normal turn completion (still skipped for aborted/error turns and when `completion.notify` is `off`).
+- Changed the in-progress `task` tool block to keep the shared `context` brief (`# Goal` / `# Constraints` background) visible after the first progress snapshot arrives, instead of dropping it the moment the streaming call view was replaced by the result frame, and to stop animating a spinner/clock next to the `Task` frame header while running — the per-agent body lines already carry their own running spinner, so the header now shows a static state icon (matching the completed/failed header icons). The context is rendered through a shared `buildContextSection` helper that also undoes per-field double-encoding, so the brief reads cleanly in the result frame even though `renderResult` receives the raw (un-repaired) tool args.
+- Changed the messaging shown when you press Esc to interrupt a streaming turn from the ambiguous `Operation aborted` / `Tool execution was aborted: Request was aborted` to `Interrupted by user`, so a deliberate user interrupt no longer reads like an internal failure. Every Esc/flush interrupt path (`onEscape` while streaming, the queued-message restore-and-abort path, and the empty-submit queue flush) threads the reason through `AgentSession.abort({ reason })` → `Agent.abort(reason)` so it rides the `AbortController` onto the aborted assistant message's `errorMessage`; the turn label renders it verbatim on both the live and replay paths, and the synthetic placeholder results paired with in-flight tool calls now read `Tool execution was aborted: Interrupted by user`. Aborts that carry no reason still fall back to the retry-aware `Operation aborted` generic. Transcript label resolution is centralized in `resolveAbortLabel` (`session/messages.ts`).
+
+### Removed
+
+- Removed the `/background` (and `/bg`) slash command and the background-mode subsystem it was the sole entry point for — `InteractiveMode.isBackgrounded`, `createBackgroundUiContext`, `handleBackgroundEvent`, and every `isBackgrounded` guard across the input/event/extension-UI controllers and UI helpers. The command suspended the whole process group via `SIGTSTP` (a leftover testing shortcut) instead of detaching the running agent, which is not the expected workflow — use terminal panes or a multiplexer instead.
+
+### Fixed
+
+- Fixed inline `find` and `search` result blocks to align with grouped `read` output and render their success headers with the normal tool-title color instead of accent blue.
+
+- Fixed the working-status shimmer to opt into the loader's 30fps animated-message repaint path while keeping both the status spinner and pending bash/eval tool spinners on their normal 80 ms glyph cadence.
+- Fixed consecutive `read` tool calls failing to collapse into a single grouped block when a reasoning model emits one read per completion (`[thinking, read]`). The read group was reset on every assistant `message_start`, so each read rendered as its own one-entry `Read …` line; now a read run accretes across completions and is broken only by a rendered non-empty text/thinking block, a non-read tool, or a user/IRC message — matching the transcript-rebuild path. `ReadToolGroupComponent` now reports its live/finalized state so the growing `Read (N)` header repaints correctly on native-scrollback (risk) terminals.
+- Fixed the `task` tool shared-context brief rendering raw Markdown headings (`# Goal`, `# Constraints`) inside framed call/result blocks instead of using the normal Markdown renderer.
+- Fixed the animated pending border on `bash`/`eval` blocks leaving a frozen dark "bar" segment behind after a backgrounded command finalized through the async update path. Once a command is auto-backgrounded (`details.async.state === "running"`) the block stays "partial" in the TUI until the async job-manager delivers the final result, but it also gets committed to native scrollback — so a mid-sweep shimmer frame baked a stray darkened border segment into the committed copy. The border now stops animating (and the 60fps redraw loop stops) the moment a block enters the backgrounded state, so the committed frame is a clean static border.
+- Fixed cold `omp` launch to clear native terminal history on the first paint, avoiding a once-per-launch duplicate welcome/transcript copy before the normal session replay.
+- Fixed plan approval resolution so `resolve` with `action: "apply"` can still find the plan file when `extra.title` is missing or stale by falling back to the current plan path and most-recent local plan artifacts
+- Fixed the search-family tool magnifier glyph (`find`, `search`, `ast_grep`, `search_tool_bm25`) to use the `accent` title color instead of `success` green, so the icon matches the tool title in the status header instead of standing out
+- Fixed TTSR stream interrupts to pass the matched rule name through the abort reason, so aborted in-flight tool placeholders say why they were stopped instead of `Request was aborted`.
+- Fixed URL reads for binary/special payloads to reuse local readers: remote archives list their root entries, SQLite databases show their table overview, notebooks render as editable cells, and unrenderable binary returns a metadata notice instead of decoded byte garbage.
+- Fixed pasted image-file paths that cannot be loaded to fall back to normal text paste with status feedback instead of disappearing.
+- Fixed tool-output file paths not being clickable OSC 8 `file://` hyperlinks in several renderers. `read` titles for plain text and image files (the common case) emitted no link at all because the renderer only linked when a `resolvedPath` was recorded — which the ordinary file/image read paths never set, keeping the absolute path only in `meta.source`; the renderer now falls back to that source path. `write` headers were never wrapped in a hyperlink and now link to the absolute path written (file, archive entry, SQLite, and conflict resolutions). `edit`/`apply_patch` headers wrapped the model-supplied (often cwd-relative) argument path, producing a root-anchored `file:///rel/path` URI; they now link the absolute `details.path` instead. Finally, `search`, `ast_grep`, and `ast_edit` produced doubled link targets (`/proj/src/src/file.ts`) for searches scoped to a subdirectory, because the renderer resolved the cwd-relative display paths against the scope directory rather than cwd — the scoped-search base is now the session cwd (with the scoped file's absolute path still seeding single-file body lines).
+- Fixed `omp dry-balance --bench` to recover from 401 token failures by re-minting the failing OAuth credential in place before switching accounts
+- Fixed the bash tool corrupting commands that embed multi-byte UTF-8 (e.g. `✓`/`×` inside a `grep -E` pattern) ahead of a trailing `| head`/`| tail`. The `bash.stripTrailingHeadTail` rewrite cut at char-offset positions reported by `brush-parser` while slicing the command by byte offset, so the trailing-pipe strip landed mid-pattern and dropped the closing quote — turning `… |✓|×|XCTAssert" | tail -80` into `… |✓|×-80` and making execution fail with `pi-natives:command: unterminated double quote`. Fixed in `pi_shell::fixup` (`@oh-my-pi/pi-natives`).
+- Fixed `omp dry-balance --bench` to recover from 401 token failures by re-minting the failing OAuth credential in place before switching accounts
+- Fixed duplicate file entries in grouped outputs for `find`, `search`, `ast_grep`, `ast_edit`, and `lsp` diagnostics when the same path appeared multiple times
+- Fixed search, grep, and edit output rendering so repeated directory group blank-line boundaries no longer break nested path/link reconstruction
+- Fixed `omp dry-balance --bench` flooding the terminal with staircased, duplicated spinner/status lines (and an indented summary) when the tty has ONLCR/OPOST disabled (raw mode). The interactive progress region separated rows with a bare LF and repositioned with a column-preserving `\x1b[<n>A` cursor-up, both of which only land at column 0 when the terminal translates LF→CRLF; with that translation off, every 80 ms redraw cascaded down and to the right into scrollback. The live region now carriage-returns before every cleared row, terminates each row with CRLF, and caps each row to the terminal width so a wrapped line cannot desync the cursor-up from the logical line count.
+- Fixed inconsistent vertical spacing between transcript blocks: some blocks (tool results from `search`/`find` and other renderer-backed tools) rendered with a doubled gap (a leading `Spacer` plus the content box's own `paddingY`), while others (the grouped `read` card, file-mention lists, IRC cards) rendered with no gap at all. Vertical spacing is now owned entirely by the chat renderer: `TranscriptContainer` strips each block's plain-blank top/bottom edges and inserts exactly one blank line between consecutive blocks, so every block is separated by a single consistent gap regardless of which component produced it. Individual components (assistant/user/tool/read-group/bash/eval/skill/custom/hook/compaction/branch/todo-reminder/plan-review messages) no longer emit their own leading `Spacer`/`paddingY` for separation, and multi-row groups (IRC cards, file-mention lists, completed-job batches, and the bordered command/`/changelog`/`/context`/version/OAuth/debug panels) are wrapped as single `TranscriptBlock` children so the renderer spaces them as one unit. Background-colored box padding is preserved as block-internal design.
+- Fixed `resolve` with `action: "discard"` surfacing a hard `isError` "No pending action to resolve" failure to the model when the agent asked to cancel a staged action (e.g. an `ast_edit` preview) but nothing was pending. A discard is a request to reach the "no staged change" end-state, which already holds in that case, so it is now honored as a successful cancellation (`"Nothing to discard; no pending action remains."` with `details.action: "discard"`) instead of an error. `action: "apply"` with no pending action still errors.
+- Fixed the collapsed tool-output expand hint rendering double brackets (e.g. `((Ctrl+O for more))`) — the `EXPAND_HINT` text already carried its own parentheses and then `formatExpandHint` wrapped it again with the theme's bracket glyphs. The hint now resolves the key actually bound to `app.tools.expand` at render time and reads `⟨<key>: Expand⟩` (e.g. `⟨Ctrl+O: Expand⟩`), so a single bracket pair surrounds it and a user remap of the expand keybinding is reflected instead of a hard-coded `Ctrl+O`.
+- Fixed the `edit`/`apply_patch` tool dropping its outlined frame while streaming/in-progress (only the final result was framed); the in-progress diff preview now renders inside the same muted frame as the completed result.
+- Fixed the `todo` and `job` tools rendering a success icon and success styling on a failed/error result; error results now show the error icon and a red frame border.
+- Fixed `debug` tool refusing every `dlv` launch on Go modules. The launch handler ran `validateLaunchProgram` before adapter selection and rejected any directory program with `launch program resolves to a directory`, while dlv's default `mode=debug` requires a Go package path (a directory or `.go` source file). Adapter resolution now precedes validation, directory programs prefer adapters that advertise `acceptsDirectoryProgram` before falling back to native extensionless debuggers, the rejection only fires when the resolved adapter does not advertise that flag (set on `dlv` in `dap/defaults.json`), and dlv's `mode` is derived from the program shape — directories and `.go` files launch as `mode=debug`, other files as `mode=exec` — so `omp` can debug both Go packages and pre-built binaries ([#2020](https://github.com/can1357/oh-my-pi/issues/2020)).
+
+## [15.10.0] - 2026-06-06
+
+### Breaking Changes
+
+- Replaced the `providers.parallelFetch` boolean setting with the `providers.fetch` enum (`auto` / `native` / `trafilatura` / `lynx` / `parallel` / `jina`) that selects the URL reader-backend priority for the `read`/`fetch` tool, mirroring `providers.image`/`providers.webSearch`. Existing configs are migrated automatically: the legacy key is dropped and the new `auto` default applies.
+
+### Added
+
+- Added a GitHub Actions read handler to the `read`/web-fetch GitHub scraper. Fetching `github.com/{owner}/{repo}/actions/runs/{id}` renders the run metadata plus a per-job breakdown (steps listed for any job that did not succeed), and `…/actions/runs/{id}/job/{id}` (also the API-style `…/jobs/{id}`) renders a single job's metadata, step table, and full plain-text logs. Logs are fetched via the `actions/jobs/{id}/logs` redirect using `GITHUB_TOKEN`/`GH_TOKEN` when present, with the per-line ISO timestamp prefix and leading BOM stripped; the section degrades to an explicit notice when logs are unavailable (no token, private repo, or expired/unfinalized run).
+
+### Changed
+
+- Changed eval `agent()` subagents so they are never subject to the `task.maxRuntimeMs` wall-clock cap. The parent cell's idle watchdog is already suspended for the entire bridge call (`withBridgeTimeoutPause`), so a long-running fan-out/recovery workflow must not be killed by a per-subagent runtime limit. `runEvalAgent` now passes `maxRuntimeMs: 0` to `runSubprocess`, which honors an explicit `ExecutorOptions.maxRuntimeMs` override over the inherited setting.
+- Changed interactive timing behavior so `PI_TIMING=x pi` preloads the module timer before the CLI graph loads and includes the `(modules)` report. `PI_TIMING=full` now also exits after printing, matching `PI_TIMING=x`, so full module reports are usable for cold-start measurement without launching the TUI. Added the root `dev:timing` script for the same profiled startup path.
+- Changed coding-agent startup imports so normal TUI launch imports `InteractiveMode` directly, keeps print/RPC/ACP runners on their branch-only paths, and moves marketplace auto-update work behind a lightweight deferred starter.
+- Changed cold-launch setup gating so the full setup wizard (every scene plus the overlay and their TUI/OAuth/web-search/theme dependencies) is no longer statically imported by `main.ts`. The current setup version now lives in a tiny dependency-free `modes/setup-version` module, and the wizard barrel is lazy-loaded only when the stored setup version is stale or the wizard is forced — the common up-to-date launch skips loading it entirely.
+- Changed cold-launch startup imports so the hot-path CLI files no longer pull the full `@oh-my-pi/pi-ai` barrel: `commands/launch.ts` and `cli/args.ts` import `THINKING_EFFORTS`/`Effort` from the tiny `@oh-my-pi/pi-ai/effort` module, and `config/model-registry.ts` now imports its ~20 symbols from narrow subpaths (`api-registry`, `model-cache`, `model-manager`, `model-thinking`, `models`, `provider-models`, `types`, `utils/event-stream`) instead of the barrel — so launching no longer eagerly loads every provider, auth, OAuth, and usage module re-exported by the barrel.
+- Changed the `read`/`fetch` HTML reader-backend priority to `native > trafilatura > lynx > parallel > jina` (was `parallel > jina > trafilatura > lynx > native`). The in-process native `htmlToMarkdown` runs first — instant, no network, full-fidelity — so the common case no longer depends on a remote service, and a stalled remote backend can no longer mask it. Selecting a specific backend via `providers.fetch` tries it first, then the rest fall back. The low-quality gate (`>100` chars and not `isLowQualityOutput`) now applies uniformly to every backend; when none clears it, the highest-priority substantial-but-low-quality output is still surfaced so the `llms.txt` / document-extraction fallbacks keep running.
+
+### Fixed
+
+- Fixed eval `agent()` failures surfacing as an opaque `RuntimeError: bridge call '__agent__' failed` with no reason. When a subagent aborted, `runEvalAgent` built its failure message with `result.error ?? result.stderr ?? result.abortReason ?? …`, but `result.stderr` is the empty string on a clean abort (and `result.error` is gated on a non-empty `stderr`), so the nullish chain stopped at `""` and never reached `abortReason`. The empty string propagated through the loopback bridge and the Python prelude's `RuntimeError(msg or "bridge call … failed")`, discarding the real reason. The chain now uses `||` so an empty `stderr` falls through to `abortReason`.
+- Fixed subagent aborts being mislabeled as the generic "Cancelled by caller" when the abort originated inside the subagent's own turn (`stopReason: "aborted"` with no caller signal and no runtime-limit timer). `runSubprocess` now prefers the aborted assistant message's `errorMessage` (e.g. "Request was aborted" or a specific stream error) for that case, while a real caller signal or wall-clock abort still reports its precise reason.
+- Fixed a long streaming tool preview that alone overflows the viewport dropping its scrolled-off head on ED3-risk terminals (ghostty/kitty/iTerm2/…). When expanded with `Ctrl+O`, a streaming `write` (content streaming in) and a streaming `eval` (stdout streaming below its fixed code cell) render top-anchored and grow append-only, but the tool block never reported itself append-only to the transcript, so the renderer's commit-as-you-go boundary stopped at the block start and the earlier rows that scrolled above the viewport were committed nowhere — they vanished, leaving the preview looking like a viewport-tall circular buffer. `ToolExecutionComponent` now implements `isTranscriptBlockAppendOnly()` (gated on `isTranscriptBlockFinalized()`, so it also covers partial-result streams like `eval`), delegating to a renderer-declared `isStreamingPreviewAppendOnly` predicate so the expanded stream commits its head exactly like a streamed assistant reply. Collapsed previews (bounded sliding tail windows) and finalized/result previews (which can collapse to a capped view) stay deferred.
+- Fixed `read`/`fetch` silently dropping whole list sections on pages with malformed list markup — stray `<img>`, text, or `<video>` nodes as direct children of `<ul>` (e.g. the Alacritty changelog). The Jina reader (previously tried before the local renderers) mis-extracts such lists, returning empty `Added`/`Changed` sections; the native in-process renderer now runs first and preserves the full content.
+- Fixed `/usage` aggregate amount fallback using raw `limits.length` as account count — now counts unique `accountId` values from limit scopes, so N limits from a single account no longer display as "N accts".
+- Fixed `/usage` account labeling falling back to "account N" for providers that use `projectId` as their primary identity (e.g. Google Antigravity, Gemini CLI) — `projectId` from report metadata is now considered before the generic fallback.
+- Fixed the `todo` tool's TUI renderer crashing with `TypeError: args?.ops?.map is not a function` when a streaming tool-call delta surfaced a non-array `ops` field (mid-stream `parseStreamingJson` shapes like `{ ops: "[{" }`, or `[null]` entries before fields arrive). The renderer now treats non-array `ops`, non-object entries, and non-array `items` as missing structure instead of crashing, which also stops the spam-warn cascade that followed each malformed delta. Paired with the Anthropic-side reasoning-replay fix in `packages/ai` ([#2005](https://github.com/can1357/oh-my-pi/issues/2005)).
+- Fixed Python eval `agent()` collapsing subagent runtime-limit aborts (and other empty-stderr aborts) into a generic `RuntimeError: bridge call '__agent__' failed`. `runEvalAgent` coalesced the failure message with `??`, which stopped at the empty `stderr` and never reached `abortReason`, shipping an empty error through the loopback bridge. The bridge now prefers `abortReason` for aborts and trims empty `stderr`/`error` out of the fallback chain, so Python surfaces the actionable reason (e.g. `Subagent runtime limit exceeded (task.maxRuntimeMs=900000)`) ([#2006](https://github.com/can1357/oh-my-pi/issues/2006)).
+
+## [15.9.69] - 2026-06-06
+
+### Added
+
+- Added anonymous fallback for Perplexity web search, allowing `web_search` and explicit Perplexity provider usage when no Perplexity credentials are configured
+- Added `gallery` CLI command to render built-in tool renderer output across streaming, in-progress, success, and failure states
+- Added `omp gallery` filtering and rendering options (`--tool`, `--state`, `--width`, `--expanded`, and `--plain`) for focused renderer previews and plain-text output
+- Added `omp gallery` fidelity for tools whose renderers are attached on the tool instance (`lsp`, `task`): the gallery now drives them through the same custom-tool render branch production uses, so regressions in that path surface in the gallery rather than only in a live session.
+- Added `omp gallery --screenshot`, which renders the gallery through a real virtual terminal (VHS) and writes PNG screenshot(s) instead of ANSI, so agents (and anything that can only read raw bytes) can actually see the rendered output. The capture forces truecolor and matches the active theme/symbol preset; tall galleries split across multiple images (whole renderers are never cut). Tune with `--out`, `--font`, and `--font-size`; requires `vhs` on `PATH` and fails with install guidance when absent.
+- Added `app.display.reset`, bound to `Ctrl+L` by default, to force an immediate terminal display reset/redraw without resizing the window.
+
+### Changed
+
+- Changed authenticated Perplexity ask requests to default to the `experimental` model preference, matching the anonymous fallback.
+- Changed Perplexity explicit provider availability checks so the setup wizard can mark `perplexity` as available for manual selection without credentials, while auto provider discovery still requires auth
+- Changed the default persistent model selector shortcut from `Ctrl+L` to `Alt+M`, leaving `Ctrl+L` for display reset. Existing user remaps in `keybindings.yml` are preserved.
+- Changed the edit tool result header to carry the diff change stats (`+N / -M / K hunks`) inline next to the file path, and removed the redundant lone language-icon metadata row and the blank line between the header and the diff body, so a single-hunk edit renders as `✔ Edit: <icon> path:LINE ⟨+3 / 1 hunk⟩` immediately followed by the diff.
+- Changed the `web_search` tool result rendering: the answer text now shows in full instead of being truncated to a "… N more lines" preview (the `omp q` CLI still caps its compact output), each source renders as a single `title (domain) · age` line with the URL linked on the title (dropping the snippet and bare-URL rows), and the metadata block collapses to one `Provider: <model> @ <provider> (<auth>)` line plus `Usage:` (removing the redundant Sources/Citations/Request/Queries rows).
+
+### Fixed
+
+- Fixed Perplexity `perplexity_ask` response parsing so OAuth, cookie, and anonymous searches correctly extract answer text and sources from JSON payloads
+- Fixed framed read results rendering with an extra blank row above and below the output block.
+- Fixed collapsed search result previews that could show only "… N more matches" when the first grouped section exceeded the preview budget. Collapsed search output now compacts to match rows, fills the budget with visible hits before the summary, and keeps truncation details out of the bottom user-visible notice.
+- Fixed the expanded search result view dumping every match when all hits live in one file. A single file's matches collapse into one blank-line group, and the expanded tree list ignored the line budget, so a hot file whose matches span its whole length rendered every row (e.g. lines 374–2858). Expanded search output is now bounded by a larger-than-collapsed budget (`EXPANDED_LINES × 2`), keeps surrounding context rows, and appends a `… N more matches` summary when truncated.
+- Fixed boolean environment flag overrides that were ORed with settings, so `PI_INTENT_TRACING=0`, `PI_AUTO_QA=0`, and per-backend eval flags now take precedence when present while falling back to config when unset.
+- Fixed custom-rendered tools that set `mergeCallAndResult` (e.g. `lsp`) rendering a redundant tool-name line above the framed result once a result arrived. `ToolExecutionComponent`'s custom-tool branch now emits the fallback label only when the tool has no `renderCall` and the call is not suppressed by an existing result, matching the built-in renderer branch.
+- Fixed the `write` tool result rendering with a green success checkmark even when the write failed. `writeToolRenderer.renderResult` now branches on `result.isError`, rendering the error status icon plus the failure message instead of the success header and content preview.
+- Fixed Perplexity OAuth/cookie web search returning a refusal answer ("I don't currently have access to the web-search tools in this turn") despite returning real sources. `callPerplexityOAuth` was prepending the API-style `web-search` system prompt to the query (`query_str = systemPrompt + "\n\n" + query`), but the consumer `www.perplexity.ai/rest/sse/perplexity_ask` endpoint has no system-message slot and reads the prepended instruction as a meta-prompt, making the model decline. The OAuth/cookie path now sends the bare query; the API-key path still passes the system prompt as a proper `system` message.
+- Fixed Perplexity OAuth web search always returning the free `turbo` model instead of the account's Pro model (e.g. `pplx_pro_upgraded`/Sonar). The `www.perplexity.ai/rest/sse/perplexity_ask` endpoint authenticates via the `__Secure-next-auth.session-token` cookie and ignores the `Authorization: Bearer` header entirely — so sending the OAuth session token as a bearer was treated as an anonymous request, which silently downgrades to `turbo` regardless of `model_preference`. The stored Perplexity OAuth token is itself the next-auth session JWT (the macOS app injects the same value as that cookie), so `callPerplexityAsk` now sends it as the `__Secure-next-auth.session-token` cookie, unlocking Pro model selection.
+
+## [15.9.67] - 2026-06-06
+
+### Added
+
+- Added `timeout-pause` and `timeout-resume` eval bridge status events emitted around `agent()`/`llm()` operations
+- Added a `/copy` picker: `/copy` now opens a fullscreen, outlined tree of recent assistant messages with their code blocks nested beneath (like `/tree`). Navigate with ↑↓, and Enter copies the highlighted node — a whole message, an individual code block, "All N blocks", or a bash/eval command interleaved with the assistant turn that issued it. A live preview pane shows the selected target, wrapping prose and syntax-highlighting code/commands.
+
+### Changed
+
+- Changed eval timeout accounting so delegated bridge calls now suspend the cell watchdog and start a fresh timeout window when runtime control returns
+- Changed `IdleTimeout` to support reference-counted pauses so overlapping delegated bridge calls keep timeout paused until all calls complete
+- Changed the default `app.message.followUp` binding from `Ctrl+Enter` alone to `[Ctrl+Q, Ctrl+Enter]` so the follow-up shortcut works in Windows Terminal, which does not deliver a distinct `Ctrl+Enter` event to console apps. `Ctrl+Q` mirrors the GitHub Copilot CLI default for the same action; existing remaps in `~/.omp/agent/keybindings.yml` are untouched, and if another user-remapped action already claims `Ctrl+Q`, that user binding wins while follow-up keeps `Ctrl+Enter`. `Ctrl+Q` is also reserved by `ExtensionRunner` so an extension cannot register that chord and be silently overwritten by the built-in follow-up handler ([#1903](https://github.com/can1357/oh-my-pi/issues/1903)).
+- Changed all scrollable TUI pickers and viewports to render through the shared `ScrollView` right-edge scrollbar for a uniform look, replacing their ad-hoc `(N/M)` / `[a-b/total]` text indicators (search hints and the tree filter-mode label are preserved). Covers the session/resume picker, model selector, OAuth provider selector, history search, session tree selector, agent dashboard list, extension list, user-message selector, the raw SSE debug viewer, the autoresearch dashboard overlay, and the session observer overlay.
+- Changed the `/model` and `/switch` selectors to dim and skip models whose context windows are smaller than the current chat context.
+- Changed `/copy` command targets to appear inline with recent assistant messages instead of as a separate "Last bash command" row at the end of the picker.
+
+### Fixed
+
+- Fixed the idle `Working...` loader freezing on ED3-risk terminals with unobservable native scrollback by keeping foreground live-region rendering enabled from `agent_start` until `agent_end`, before the first assistant or tool event arrives.
+- Fixed framed tool output blocks rendering one column inset inside tool boxes; modern bordered blocks now span the same width as legacy background-filled tool boxes.
+- Fixed potential `TimeoutError` aborts for short `timeout` eval cells during long bridged `agent()`/`llm()` work where no progress events are emitted until completion
+- Fixed retry recovery to allow automatic retries without switching models when `retry.modelFallback` is disabled.
+- Fixed `ttsr.enabled: false` being ignored at runtime. TTSR rules were still being registered with `TtsrManager.addRule` and matched against stream deltas even when the global toggle was off, so disabling TTSR did not suppress rule injection or stream abort. The manager now gates `addRule`, `hasRules`, and `#matchBuffer` on the enabled flag, so disabling fully short-circuits the TTSR path. Condition rules fall through to the rulebook bucket instead of being silently swallowed. ([#1767](https://github.com/can1357/oh-my-pi/issues/1767))
+- Fixed the Python eval kernel hanging on Windows during `import pandas` / `import numpy`, with SIGINT unable to recover the cell. `PythonKernel.start()` spawned the runner with `windowsHide: true`, which in Bun maps to the Win32 `CREATE_NO_WINDOW` flag and detaches the long-lived child from any inherited console — so native extensions like `numpy/_core/_multiarray_umath.pyd` (and its bundled OpenBLAS/SLEEF thread-pool init) could deadlock inside `LoadLibraryExW`, and `GenerateConsoleCtrlEvent`-based SIGINT delivery silently became a no-op. The kernel now hides its window only when the host itself has no console to share (service / piped launch); an interactive TUI launch lets the kernel inherit the parent's console, matching the behavior of `python.exe` invoked from `cmd.exe` ([#1960](https://github.com/can1357/oh-my-pi/issues/1960)).
+- Fixed `task` renderer crashing the TUI with `TypeError: completeData?.map is not a function` when a subagent's `extractedToolData.yield` slot held a non-array value. `renderAgentResult` (and the live-progress sibling) cast the slot to `Array<{ data }>` and called `?.map`, but optional chaining short-circuits only on `null`/`undefined`, so a plain object made `.map` `undefined` and threw — taking down every `review` task render. Both sites now go through `normalizeYieldData`, which wraps a single object as a 1-element array and drops primitives ([#1987](https://github.com/can1357/oh-my-pi/issues/1987))
+- Fixed `sdk-async-job-manager-singleton` tests flaking under the full parallel suite. The four `createAgentSession`-based cases ran on the default 5000ms per-test timeout, which two real session startups can exceed when `test:ts` saturates the machine across packages; on timeout the still-running test body and `afterEach` reset raced, surfacing a spurious "Unhandled error between tests" on the `AsyncJobManager.instance()` assertion. They now carry an explicit 60000ms timeout, matching the convention used by the other session-creating tests in this suite.
+- Fixed streaming `eval`, `bash`, `ssh`, and `task` call previews overflowing the live transcript viewport and cutting off their top while pending. A volatile tool block taller than the viewport could strand its scrolled-off head out of native scrollback on ED3-risk terminals (committed nowhere, repainted nowhere) until the result landed. The pending `eval` source preview now follows the streaming edge in a bounded 12-line tail window (newest lines pinned to the bottom, "… N earlier lines" on top) so you can watch the code being written without the box overflowing; `bash`/`ssh` commands and `task` context use a bounded head+tail window. `Ctrl+O` still lifts the cap for a full view.
+- Fixed the streaming `write` call preview ignoring `Ctrl+O` so the expand toggle was a no-op while a file was being written. Unlike the `eval`/`bash`/`ssh`/`task` streaming previews, `formatStreamingContent` never received the `expanded` flag, leaving the preview pinned to a bounded 12-line tail window even after pressing `Ctrl+O` — so on a large write you could not widen past the streaming edge until the tool result landed. The preview now lifts the cap to the full file (head through tail) when expanded, matching the documented streaming-preview behavior of the other tools.
+- Fixed turn-ending provider errors rendering twice — once as the transcript's inline `Error: …` line and again in the pinned banner above the editor (added in 15.9.5). The inline line is now suppressed while the same error is mirrored in the banner and restored to the transcript when the banner clears at the next turn, so the error stays in history without the duplicate render at the error moment.
+
+### Removed
+
+- Removed the `/copy last|code|all|cmd` subcommands; every copy target is now reachable by picking it in the `/copy` tree.
+
+## [15.9.5] - 2026-06-05
+
 ### Added
 
 - Added a persistent error banner pinned above the editor when an assistant turn ends on a provider error (e.g. Anthropic's "Output blocked by content filtering policy"). The transcript `Error: …` line scrolls away as the conversation grows, so terminal turns that ended on a stream error could pass unnoticed; the banner stays in the fixed region above the input and is cleared when the next turn starts.
@@ -10,6 +182,10 @@
 - Added `OLLAMA_HOST` support for implicit local Ollama discovery when `OLLAMA_BASE_URL` is unset, so OMP picks up the same host setting used by Ollama.
 - Added `OLLAMA_CONTEXT_LENGTH` as a positive-integer context-window override for implicit local Ollama discovery, so users can correct OMP context budgeting without writing per-model overrides.
 
+### Changed
+
+- Changed `tools.discoveryMode` to default to `auto`, which keeps discovery off for small tool sets and automatically switches to MCP-only tool discovery when more than 40 tools are registered.
+
 ### Fixed
 
 - Fixed user-message rendering to materialize image links from embedded image blocks when rebuilding chat output, so image placeholders remain clickable after replayed or restored messages
@@ -25,6 +201,7 @@
 - Blocked OSC 8 hyperlink wrapping for URI targets containing terminal control bytes to avoid rendering malformed control-sequence links
 
 ## [15.9.4] - 2026-06-05
+
 ### Fixed
 
 - Fixed chat transcript updates after submitting input so frozen scrollback is only thawed when native scrollback replay succeeds, preventing misplaced or duplicated rows when the viewport is not at the tail
diff --git a/packages/coding-agent/DEVELOPMENT.md b/packages/coding-agent/DEVELOPMENT.md
index 731434b48..3f16ac804 100644
--- a/packages/coding-agent/DEVELOPMENT.md
+++ b/packages/coding-agent/DEVELOPMENT.md
@@ -1017,7 +1017,7 @@ Core pipeline in `renderUrl(url, timeout, raw, signal)`:
    - HTML-to-text conversion (`renderHtmlToText`)
 7. If HTML conversion is low-quality (`isLowQualityOutput`), try document-link extraction (`extractDocumentLinks`) + markit.
 
-`renderHtmlToText(...)` fallback order is explicit in code: Jina reader endpoint (`https://r.jina.ai/<url>`), then `trafilatura` (via `ensureTool`), then `lynx`, then native `htmlToMarkdown`.
+`renderHtmlToText(...)` tries reader backends in priority order: native in-process `htmlToMarkdown`, then `trafilatura` (via `ensureTool`), then `lynx`, then Parallel extract, then the Jina reader endpoint (`https://r.jina.ai/<url>`). The `providers.fetch` setting picks the order — `auto` uses the default above; selecting a specific backend tries it first, then the rest as fallbacks. Every backend's output must clear the same quality gate (`>100` chars and not `isLowQualityOutput`); if none clears it, the highest-priority substantial-but-low-quality output is surfaced so step 7's targeted fallbacks still run. native operates on already-fetched HTML (instant, no network), so it both wins the common case and survives a stalled remote backend.
 
 ### Scraper registry role (`web/scrapers/index.ts`)
 
diff --git a/packages/coding-agent/package.json b/packages/coding-agent/package.json
index 774dff017..68682aa71 100644
--- a/packages/coding-agent/package.json
+++ b/packages/coding-agent/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-coding-agent",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
diff --git a/packages/coding-agent/scripts/dev-launch b/packages/coding-agent/scripts/dev-launch
new file mode 100755
index 000000000..c187c6161
--- /dev/null
+++ b/packages/coding-agent/scripts/dev-launch
@@ -0,0 +1,42 @@
+#!/bin/sh
+# Dev launcher for the omp CLI, installed by `bun run install:dev`.
+#
+# Problem it solves: Bun reads `bunfig.toml` from the *current working
+# directory* at startup and evaluates its `preload` entries before running the
+# script. A bun-shebang bin (what `bun link` creates for `src/cli.ts`)
+# therefore inherits whatever `preload` the directory you happen to be in
+# declares. Running `omp`/`pi` inside an unrelated Bun project can execute — and
+# crash on — that project's preload, e.g.
+#   error: Cannot find module '@v12sh/utils/frontmatter' from '.../loader.ts'
+#
+# Bun only reads the *exact* cwd (it does not walk parents) and ignores
+# `--config`/`BUN_BE_BUN` for this, so the fix is to launch Bun from an empty,
+# bunfig-free directory and restore the real cwd inside the process via the
+# preload shim alongside this file.
+set -e
+
+# Resolve this script's real location even when invoked through a symlink
+# (`$HOME/.bun/bin/omp` -> this file).
+self=$0
+while [ -L "$self" ]; do
+	link=$(readlink "$self")
+	case $link in
+	/*) self=$link ;;
+	*) self=$(dirname "$self")/$link ;;
+	esac
+done
+scripts_dir=$(CDPATH= cd -- "$(dirname -- "$self")" && pwd -P)
+cli=$scripts_dir/../src/cli.ts
+preload=$scripts_dir/dev-launch-preload.ts
+timing_preload=$scripts_dir/../../utils/src/module-timer.ts
+
+launch_dir=${OMP_DEV_LAUNCH_DIR:-${HOME}/.omp/.dev-cwd}
+mkdir -p "$launch_dir"
+
+OMP_LAUNCH_CWD=$PWD
+export OMP_LAUNCH_CWD
+cd "$launch_dir"
+if [ -n "${PI_TIMING:-}" ]; then
+	exec bun --preload "$preload" --preload "$timing_preload" "$cli" "$@"
+fi
+exec bun --preload "$preload" "$cli" "$@"
diff --git a/packages/coding-agent/scripts/dev-launch-preload.ts b/packages/coding-agent/scripts/dev-launch-preload.ts
new file mode 100644
index 000000000..5cafa09ad
--- /dev/null
+++ b/packages/coding-agent/scripts/dev-launch-preload.ts
@@ -0,0 +1,19 @@
+/**
+ * Bun `--preload` shim for the omp dev launcher (`scripts/dev-launch`).
+ *
+ * The launcher starts Bun from an empty, bunfig-free directory so a foreign
+ * project's `bunfig.toml` `preload` cannot run inside the omp CLI: Bun reads
+ * `bunfig.toml` from the *current working directory* on startup and evaluates
+ * its `preload` entries before the entrypoint, so a bun-shebang bin inherits
+ * whatever `preload` the directory you launched from declares (and crashes if
+ * that preload can't resolve). This shim is loaded before the entrypoint's
+ * imports run, so it restores the user's real working directory in time for
+ * import-time snapshots (e.g. `getProjectDir()` in `@oh-my-pi/pi-utils/dirs`).
+ */
+const launchCwd = process.env.OMP_LAUNCH_CWD;
+if (launchCwd) {
+	delete process.env.OMP_LAUNCH_CWD;
+	try {
+		process.chdir(launchCwd);
+	} catch {}
+}
diff --git a/packages/coding-agent/src/auto-thinking/classifier.ts b/packages/coding-agent/src/auto-thinking/classifier.ts
index db2048b18..ccae82f73 100644
--- a/packages/coding-agent/src/auto-thinking/classifier.ts
+++ b/packages/coding-agent/src/auto-thinking/classifier.ts
@@ -15,6 +15,7 @@
  */
 import { type AssistantMessage, completeSimple, Effort, type Model } from "@oh-my-pi/pi-ai";
 import { prompt } from "@oh-my-pi/pi-utils";
+
 import type { ModelRegistry } from "../config/model-registry";
 import { resolveRoleSelection } from "../config/model-resolver";
 import type { Settings } from "../config/settings";
@@ -82,7 +83,10 @@ async function classifyOnline(input: string, deps: ClassifyDifficultyDeps): Prom
 			messages: [{ role: "user", content: input, timestamp: Date.now() }],
 		},
 		{
-			apiKey,
+			apiKey: deps.registry.resolver(model.provider, {
+				sessionId: deps.sessionId,
+				baseUrl: model.baseUrl,
+			}),
 			maxTokens,
 			disableReasoning: true,
 			metadata,
diff --git a/packages/coding-agent/src/autoresearch/dashboard.ts b/packages/coding-agent/src/autoresearch/dashboard.ts
index b9544dace..4ea76e0a7 100644
--- a/packages/coding-agent/src/autoresearch/dashboard.ts
+++ b/packages/coding-agent/src/autoresearch/dashboard.ts
@@ -1,4 +1,4 @@
-import { matchesKey, replaceTabs, Text, truncateToWidth, visibleWidth } from "@oh-my-pi/pi-tui";
+import { matchesKey, replaceTabs, ScrollView, Text, truncateToWidth, visibleWidth } from "@oh-my-pi/pi-tui";
 import type { Theme } from "../modes/theme/theme";
 import { formatElapsed, formatNum, isBetter } from "./helpers";
 import { currentResults, findBaselineMetric, findBaselineRunNumber, findBaselineSecondary } from "./state";
@@ -76,14 +76,14 @@ export function createDashboardController(): DashboardController {
 							const viewportRows = Math.max(4, terminalRows - 4);
 							const maxScroll = Math.max(0, body.length - viewportRows);
 							if (scrollOffset > maxScroll) scrollOffset = maxScroll;
-							const visible = body.slice(scrollOffset, scrollOffset + viewportRows);
-							const footer = renderOverlayFooter(width, scrollOffset, viewportRows, body.length, theme);
-							return [
-								header,
-								...visible,
-								...Array.from({ length: Math.max(0, viewportRows - visible.length) }, () => ""),
-								footer,
-							];
+							const sv = new ScrollView(body.slice(scrollOffset, scrollOffset + viewportRows), {
+								height: viewportRows,
+								scrollbar: "auto",
+								totalRows: body.length,
+								theme: { track: t => theme.fg("dim", t), thumb: t => theme.fg("accent", t) },
+							});
+							sv.setScrollOffset(scrollOffset);
+							return [header, ...sv.render(width), renderOverlayFooter(width, theme)];
 						},
 						handleInput(data: string): void {
 							const totalRows =
@@ -406,18 +406,8 @@ function renderOverlayRunningLine(
 	);
 }
 
-function renderOverlayFooter(
-	width: number,
-	scrollOffset: number,
-	viewportRows: number,
-	totalRows: number,
-	theme: Theme,
-): string {
-	const position =
-		totalRows > viewportRows
-			? ` ${scrollOffset + 1}-${Math.min(totalRows, scrollOffset + viewportRows)}/${totalRows}`
-			: "";
-	const hint = theme.fg("dim", ` up/down j/k pageup pagedown g G esc${position} `);
+function renderOverlayFooter(width: number, theme: Theme): string {
+	const hint = theme.fg("dim", " up/down j/k pageup pagedown g G esc ");
 	const fill = Math.max(0, width - visibleWidth(hint));
 	return theme.fg("borderMuted", "-".repeat(fill)) + hint;
 }
diff --git a/packages/coding-agent/src/cli-commands.ts b/packages/coding-agent/src/cli-commands.ts
index c9ac00741..efa68d4fa 100644
--- a/packages/coding-agent/src/cli-commands.ts
+++ b/packages/coding-agent/src/cli-commands.ts
@@ -22,6 +22,7 @@ export const commands: CommandEntry[] = [
 	{ name: "config", load: () => import("./commands/config").then(m => m.default) },
 	{ name: "dry-balance", load: () => import("./commands/dry-balance").then(m => m.default) },
 	{ name: "grep", load: () => import("./commands/grep").then(m => m.default) },
+	{ name: "gallery", load: () => import("./commands/gallery").then(m => m.default) },
 	{ name: "grievances", load: () => import("./commands/grievances").then(m => m.default) },
 	{ name: "install", load: () => import("./commands/install").then(m => m.default) },
 	{ name: "plugin", load: () => import("./commands/plugin").then(m => m.default) },
diff --git a/packages/coding-agent/src/cli/args.ts b/packages/coding-agent/src/cli/args.ts
index cefb84942..707d87c54 100644
--- a/packages/coding-agent/src/cli/args.ts
+++ b/packages/coding-agent/src/cli/args.ts
@@ -1,7 +1,7 @@
 /**
  * CLI argument parsing and help display
  */
-import { type Effort, THINKING_EFFORTS } from "@oh-my-pi/pi-ai";
+import { type Effort, THINKING_EFFORTS } from "@oh-my-pi/pi-ai/effort";
 import { APP_NAME, CONFIG_DIR_NAME, logger } from "@oh-my-pi/pi-utils";
 import chalk from "chalk";
 import { parseEffort } from "../thinking";
@@ -284,7 +284,7 @@ export function getExtraHelpText(): string {
   ${chalk.dim("# Search & Tools")}
   EXA_API_KEY                - Exa web search
   BRAVE_API_KEY              - Brave web search
-  PERPLEXITY_API_KEY         - Perplexity web search (API)
+  PERPLEXITY_API_KEY         - Perplexity web search API key (optional; anonymous fallback)
   PERPLEXITY_COOKIES         - Perplexity web search (session cookie)
   TAVILY_API_KEY             - Tavily web search
   ANTHROPIC_SEARCH_API_KEY   - Anthropic web search (override; isolates search from main ANTHROPIC_API_KEY)
diff --git a/packages/coding-agent/src/cli/claude-trace-cli.ts b/packages/coding-agent/src/cli/claude-trace-cli.ts
index f776b1dbc..50365a45c 100644
--- a/packages/coding-agent/src/cli/claude-trace-cli.ts
+++ b/packages/coding-agent/src/cli/claude-trace-cli.ts
@@ -380,6 +380,18 @@ function isMessagesRequest(message: ParsedHttpMessage): boolean {
 	return pathNameFromRequestTarget(message.path ?? "") === "/v1/messages";
 }
 
+// Claude Code fires a background warmup/classification call on its small fast
+// model (a haiku variant, ANTHROPIC_SMALL_FAST_MODEL) before sending the user's
+// real message. Skip it so the capture lands on the actual prompt.
+function isBackgroundModelRequest(message: ParsedHttpMessage): boolean {
+	try {
+		const parsed = JSON.parse(decodeBody(message.headers, message.body)) as { model?: unknown };
+		return typeof parsed.model === "string" && parsed.model.toLowerCase().includes("haiku");
+	} catch {
+		return false;
+	}
+}
+
 function decodeBody(headers: readonly HeaderEntry[], body: Buffer): string {
 	const encoding = headerValue(headers, "content-encoding")?.toLowerCase().trim();
 	try {
@@ -636,7 +648,7 @@ export class ClaudeMessagesProxy {
 			upstreamTls.write(data);
 			const messages = requestParser.push(data);
 			for (const message of messages) {
-				if (!isMessagesRequest(message)) {
+				if (!isMessagesRequest(message) || isBackgroundModelRequest(message)) {
 					responseQueue.push(null);
 					continue;
 				}
diff --git a/packages/coding-agent/src/cli/dry-balance-cli.ts b/packages/coding-agent/src/cli/dry-balance-cli.ts
index dc35d254f..85e9388c5 100644
--- a/packages/coding-agent/src/cli/dry-balance-cli.ts
+++ b/packages/coding-agent/src/cli/dry-balance-cli.ts
@@ -1,8 +1,10 @@
 import type {
 	Api,
+	ApiKeyResolver,
 	AssistantMessage,
 	AssistantMessageEvent,
 	AssistantMessageEventStream,
+	AuthCredentialSnapshotEntry,
 	Context,
 	Model,
 	OAuthAccess,
@@ -59,6 +61,12 @@ export interface DryBalanceAuthStorage {
 		options?: DryBalanceAuthOptions,
 	): Promise<OAuthAccess | undefined>;
 	getOAuthAccesses?(provider: string, options?: DryBalanceAuthOptions): Promise<OAuthAccessResolution[]>;
+	/**
+	 * Force-refresh a single credential by id (step (b) of the auth-retry
+	 * policy). The bench re-mints the failing account's token in place on a
+	 * 401 rather than rotating accounts — it is measuring each account.
+	 */
+	forceRefreshCredentialById?(id: number, signal?: AbortSignal): Promise<AuthCredentialSnapshotEntry>;
 }
 
 export interface DryBalanceModelRegistry {
@@ -152,6 +160,8 @@ export interface DryBalanceDependencies {
 	now?: () => number;
 	stdoutIsTTY?: boolean;
 	stderrIsTTY?: boolean;
+	stdoutColumns?: number;
+	stderrColumns?: number;
 }
 
 type DryBalanceAttemptResult =
@@ -181,6 +191,7 @@ type DryBalanceBenchTarget =
 			ok: true;
 			account: string;
 			accessToken: string;
+			credentialId?: number;
 	  }
 	| {
 			ok: false;
@@ -310,10 +321,11 @@ function renderBenchStatusLine(
 	}
 }
 
-function createBenchProgressSink(
+export function createBenchProgressSink(
 	total: number,
 	write: (text: string) => void,
 	interactive: boolean,
+	columns: number,
 ): DryBalanceBenchProgressSink {
 	const statuses: DryBalanceBenchProgressStatus[] = Array.from({ length: total }, () => ({ state: "waiting" }));
 	if (!interactive) {
@@ -333,13 +345,21 @@ function createBenchProgressSink(
 	let frame = 0;
 	let lineCount = 0;
 	let timer: NodeJS.Timeout | undefined;
+	const width = Number.isFinite(columns) && columns > 0 ? Math.trunc(columns) : 80;
 	const render = (): void => {
 		const lines = [
 			chalk.bold("bench requests"),
 			...statuses.map((status, index) => renderBenchStatusLine(status, index, total, frame)),
 		];
-		if (lineCount > 0) write(`\x1b[${lineCount}A`);
-		write(`${lines.map(line => `\x1b[2K${line}`).join("\n")}\n`);
+		// Anchor every redraw at column 0 and terminate each row with CRLF: a
+		// bare `\n` only returns to column 0 when the tty performs ONLCR
+		// translation, which is off whenever the terminal is in raw mode — there
+		// the old column-preserving cursor-up staircased each frame into
+		// scrollback. Cap each line to the terminal width so a wrapped row never
+		// desyncs the `\x1b[<n>A` cursor-up from the logical line count.
+		const move = lineCount > 0 ? `\x1b[${lineCount}A` : "";
+		const body = lines.map(line => `\x1b[2K${truncateToWidth(line, width)}`).join("\r\n");
+		write(`${move}\r${body}\r\n`);
 		lineCount = lines.length;
 	};
 	render();
@@ -370,13 +390,23 @@ function createBenchProgressSink(
 async function runBenchRequest(
 	model: Model<Api>,
 	sessionId: string,
-	account: string,
-	accessToken: string,
+	target: Extract<DryBalanceBenchTarget, { ok: true }>,
+	authStorage: DryBalanceAuthStorage,
 	streamFn: DryBalanceStreamSimple,
 	now: () => number,
 ): Promise<DryBalanceBenchResult> {
+	const { account, accessToken, credentialId } = target;
 	const startedAt = now();
 	let firstTokenAt: number | undefined;
+	// Re-mint the cached token on a 401: a peer/broker may have rotated it out
+	// from under our snapshot (Anthropic rotates refresh tokens on every use).
+	// The bench measures one account, so the switch step intentionally declines.
+	const apiKey: ApiKeyResolver = async ({ lastChance, error }) => {
+		if (error === undefined) return accessToken;
+		if (lastChance || credentialId === undefined || !authStorage.forceRefreshCredentialById) return undefined;
+		const refreshed = await authStorage.forceRefreshCredentialById(credentialId);
+		return refreshed.credential.type === "oauth" ? refreshed.credential.access : undefined;
+	};
 	try {
 		const context: Context = {
 			messages: [
@@ -389,7 +419,7 @@ async function runBenchRequest(
 			],
 		};
 		const stream = streamFn(model, context, {
-			apiKey: accessToken,
+			apiKey,
 			sessionId,
 			maxTokens: resolveBenchMaxTokens(model),
 			temperature: 0.2,
@@ -454,7 +484,7 @@ async function resolveBenchTargets(
 		seen.add(key);
 		const account = extractAccount(entry);
 		if (entry.ok) {
-			targets.push({ ok: true, account, accessToken: entry.accessToken });
+			targets.push({ ok: true, account, accessToken: entry.accessToken, credentialId: entry.credentialId });
 		} else {
 			targets.push({ ok: false, account, error: entry.error });
 		}
@@ -465,6 +495,7 @@ async function resolveBenchTargets(
 async function runBenchTargets(
 	model: Model<Api>,
 	targets: DryBalanceBenchTarget[],
+	authStorage: DryBalanceAuthStorage,
 	randomSessionId: () => string,
 	progress: DryBalanceBenchProgressSink | undefined,
 	streamFn: DryBalanceStreamSimple,
@@ -482,14 +513,7 @@ async function runBenchTargets(
 				return result;
 			}
 			progress?.markRunning(index, target.account);
-			const result = await runBenchRequest(
-				model,
-				randomSessionId(),
-				target.account,
-				target.accessToken,
-				streamFn,
-				now,
-			);
+			const result = await runBenchRequest(model, randomSessionId(), target, authStorage, streamFn, now);
 			progress?.complete(index, result);
 			return result;
 		}),
@@ -792,8 +816,19 @@ export async function runDryBalanceCommand(
 			const progressInteractive = command.flags.json
 				? (deps.stderrIsTTY ?? process.stderr.isTTY === true)
 				: (deps.stdoutIsTTY ?? process.stdout.isTTY === true);
-			progress = createBenchProgressSink(targets.length, progressWrite, progressInteractive);
-			benchResults = await runBenchTargets(model, targets, randomSessionId, progress, streamFn, now);
+			const progressColumns = command.flags.json
+				? (deps.stderrColumns ?? process.stderr.columns ?? 80)
+				: (deps.stdoutColumns ?? process.stdout.columns ?? 80);
+			progress = createBenchProgressSink(targets.length, progressWrite, progressInteractive, progressColumns);
+			benchResults = await runBenchTargets(
+				model,
+				targets,
+				runtime.modelRegistry.authStorage,
+				randomSessionId,
+				progress,
+				streamFn,
+				now,
+			);
 			results = targets.map(target =>
 				target.ok ? { ok: true, account: target.account } : { ok: false, reason: target.error },
 			);
diff --git a/packages/coding-agent/src/cli/gallery-cli.ts b/packages/coding-agent/src/cli/gallery-cli.ts
new file mode 100644
index 000000000..b145f04f7
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-cli.ts
@@ -0,0 +1,226 @@
+/**
+ * `omp gallery` — render every built-in tool's renderer across its lifecycle.
+ *
+ * For each tool with a registered renderer, the gallery drives a real
+ * {@link ToolExecutionComponent} through four states — streaming arguments,
+ * arguments complete (in progress), success, and failure — and prints the
+ * rendered output to stdout. It exists for visual QA of tool renderers without
+ * having to provoke each state through a live agent session.
+ */
+import type { AgentTool } from "@oh-my-pi/pi-agent-core";
+import type { TUI } from "@oh-my-pi/pi-tui";
+import { getProjectDir } from "@oh-my-pi/pi-utils";
+import { Settings } from "../config/settings";
+import { ToolExecutionComponent } from "../modes/components/tool-execution";
+import { initTheme, theme } from "../modes/theme/theme";
+import { toolRenderers } from "../tools/renderers";
+import { type GalleryFixture, type GalleryResult, galleryFixtures } from "./gallery-fixtures";
+import { captureGalleryScreenshots } from "./gallery-screenshot";
+
+/** Lifecycle states the gallery renders, in display order. */
+export const GALLERY_STATES = ["streaming", "progress", "success", "error"] as const;
+export type GalleryState = (typeof GALLERY_STATES)[number];
+
+const STATE_LABELS: Record<GalleryState, string> = {
+	streaming: "streaming args",
+	progress: "in progress",
+	success: "done",
+	error: "failed",
+};
+
+export interface GalleryCommandArgs {
+	/** Render width in columns (defaults to terminal width, clamped). */
+	width?: number;
+	/** Restrict to a single tool name. */
+	tool?: string;
+	/** Restrict to specific lifecycle states. */
+	states?: GalleryState[];
+	/** Render the expanded variant of each renderer. */
+	expanded?: boolean;
+	/** Strip ANSI styling from the output (useful when redirecting to a file). */
+	plain?: boolean;
+	/** Capture the rendered gallery as PNG screenshot(s) via VHS instead of printing ANSI. */
+	screenshot?: boolean;
+	/** Screenshot output path (single image) or base path (suffixed when split across images). */
+	out?: string;
+	/** Font family for screenshots (must be installed; Nerd Font recommended for icon glyphs). */
+	font?: string;
+	/** Font size in points for screenshots. */
+	fontSize?: number;
+}
+
+/** One tool's rendered lifecycle, as ANSI lines: a leading blank, the section rule, then each state. */
+export interface GallerySection {
+	heading: string;
+	lines: string[];
+}
+
+const GENERIC_ERROR: GalleryResult = {
+	content: [{ type: "text", text: "Error: operation failed" }],
+	isError: true,
+};
+
+/**
+ * Build the fake `AgentTool` the component needs for its label, edit mode, and —
+ * for `customRendered` fixtures — the renderer functions that route it through
+ * the same custom-tool branch production uses (see {@link GalleryFixture}).
+ */
+function fakeToolFor(name: string, fixture: GalleryFixture | undefined): AgentTool | undefined {
+	if (!fixture?.label && !fixture?.editMode && !fixture?.customRendered) return undefined;
+	const tool: Record<string, unknown> = { name, label: fixture.label ?? name, mode: fixture.editMode };
+	if (fixture.customRendered) {
+		const renderer = toolRenderers[name] as
+			| { renderCall?: unknown; renderResult?: unknown; mergeCallAndResult?: unknown; inline?: unknown }
+			| undefined;
+		if (renderer) {
+			tool.renderCall = renderer.renderCall;
+			tool.renderResult = renderer.renderResult;
+			tool.mergeCallAndResult = renderer.mergeCallAndResult;
+			tool.inline = renderer.inline;
+		}
+	}
+	return tool as unknown as AgentTool;
+}
+
+/** The curated fixture for a tool, or a generic one for registry tools lacking sample data. */
+export function resolveFixture(name: string): GalleryFixture {
+	return (
+		galleryFixtures[name] ??
+		({
+			args: { note: `sample ${name} call` },
+			result: { content: [{ type: "text", text: `${name} completed` }] },
+		} satisfies GalleryFixture)
+	);
+}
+
+/**
+ * Render a single tool/state pair to lines. Builds a fresh component, drives it
+ * to the requested state, settles any async edit preview, then snapshots the
+ * render and stops all animation timers.
+ */
+export async function renderGalleryState(
+	name: string,
+	fixture: GalleryFixture,
+	state: GalleryState,
+	width: number,
+	expanded = false,
+): Promise<string[]> {
+	const tool = fakeToolFor(name, fixture);
+	const streamingArgs = state === "streaming" ? (fixture.streamingArgs ?? fixture.args) : fixture.args;
+	// The component only calls `requestRender` during a static render;
+	// `imageBudget` is consulted solely when images render, which the gallery
+	// disables. A cast avoids constructing a real terminal.
+	const ui = { requestRender() {} } as unknown as TUI;
+	const component = new ToolExecutionComponent(name, streamingArgs, { showImages: false }, tool, ui, getProjectDir());
+	component.setExpanded(expanded);
+
+	if (state !== "streaming") {
+		component.setArgsComplete();
+	}
+	if (state === "success") {
+		component.updateResult(fixture.result, false);
+	} else if (state === "error") {
+		component.updateResult(fixture.errorResult ?? GENERIC_ERROR, false);
+	}
+
+	// Edit-like renderers compute their diff preview off the render path; wait
+	// for it to settle so the snapshot is deterministic instead of racing a tick.
+	await component.whenPreviewSettled();
+
+	const lines = component.render(width);
+	component.stopAnimation();
+	return lines;
+}
+
+function resolveWidth(requested: number | undefined): number {
+	const fallback = process.stdout.columns ?? 100;
+	const width = requested ?? fallback;
+	return Math.max(40, Math.min(200, width));
+}
+
+function sectionRule(label: string, width: number): string {
+	const prefix = `── ${label} `;
+	const fill = Math.max(0, width - prefix.length);
+	return theme.fg("accent", theme.bold(`${prefix}${"─".repeat(fill)}`));
+}
+
+/**
+ * Render each requested tool's lifecycle into ANSI section blocks. The block
+ * layout (leading blank, section rule, then a blank + dim label + body per
+ * state) is shared by the stdout and screenshot paths so both stay identical.
+ */
+async function renderGallerySections(
+	names: string[],
+	states: GalleryState[],
+	width: number,
+	expanded: boolean,
+): Promise<GallerySection[]> {
+	const sections: GallerySection[] = [];
+	for (const name of names) {
+		const fixture = resolveFixture(name);
+		const heading = fixture.label && fixture.label !== name ? `${name} — ${fixture.label}` : name;
+		const lines: string[] = ["", sectionRule(heading, width)];
+		for (const state of states) {
+			lines.push("", theme.fg("dim", `  · ${STATE_LABELS[state]}`));
+			try {
+				for (const line of await renderGalleryState(name, fixture, state, width, expanded)) lines.push(line);
+			} catch (err) {
+				lines.push(theme.fg("error", `  render failed: ${String(err)}`));
+			}
+		}
+		sections.push({ heading, lines });
+	}
+	return sections;
+}
+
+/**
+ * Render the gallery. Iterates the renderer registry (or a single tool),
+ * printing each requested lifecycle state under a labeled section — or, with
+ * `screenshot`, capturing the rendered output as PNG(s) via VHS.
+ */
+export async function runGalleryCommand(args: GalleryCommandArgs): Promise<void> {
+	const settingsInstance = await Settings.init();
+	// Screenshots must carry exact theme RGB regardless of how the invoking
+	// terminal advertises its color support, so force truecolor before the theme
+	// (and therefore every SGR escape it emits) is built.
+	if (args.screenshot) process.env.COLORTERM = "truecolor";
+	await initTheme(
+		false,
+		settingsInstance.get("symbolPreset"),
+		settingsInstance.get("colorBlindMode"),
+		settingsInstance.get("theme.dark"),
+		settingsInstance.get("theme.light"),
+	);
+
+	const width = resolveWidth(args.width);
+	const expanded = args.expanded ?? false;
+	const states = args.states && args.states.length > 0 ? args.states : [...GALLERY_STATES];
+
+	// Renderer-registry tools plus fixture-only tools (no dedicated renderer,
+	// e.g. `report_tool_issue` / custom extension tools) so the gallery covers
+	// the generic fallback + custom-tool branches too.
+	const allNames = Array.from(new Set([...Object.keys(toolRenderers), ...Object.keys(galleryFixtures)])).sort();
+	const names = args.tool ? allNames.filter(name => name === args.tool) : allNames;
+	if (args.tool && names.length === 0) {
+		process.stdout.write(`Unknown tool '${args.tool}'. Known tools: ${allNames.join(", ")}\n`);
+		return;
+	}
+
+	const sections = await renderGallerySections(names, states, width, expanded);
+
+	if (args.screenshot) {
+		const paths = await captureGalleryScreenshots(sections, {
+			width,
+			font: args.font,
+			fontSize: args.fontSize,
+			out: args.out,
+		});
+		process.stdout.write(`${paths.join("\n")}\n`);
+		return;
+	}
+
+	const lines = sections.flatMap(section => section.lines);
+	lines.push("");
+	const text = lines.map(line => (args.plain ? Bun.stripANSI(line) : line)).join("\n");
+	process.stdout.write(`${text}\n`);
+}
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/agentic.ts b/packages/coding-agent/src/cli/gallery-fixtures/agentic.ts
new file mode 100644
index 000000000..d1c262abe
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/agentic.ts
@@ -0,0 +1,292 @@
+// Gallery fixtures for the agentic orchestration tools (task, goal, job).
+import type { GalleryFixture } from "./types";
+
+export const agenticFixtures: Record<string, GalleryFixture> = {
+	task: {
+		label: "Task",
+		customRendered: true,
+		// Streaming: agent chosen, first task fully arrived, second still landing.
+		streamingArgs: {
+			agent: "task",
+			tasks: [
+				{
+					id: "AuthLoader",
+					description: "Load auth middleware",
+					assignment: "Read packages/server/src/auth/*.ts and summarize the session-cookie flow.",
+				},
+				{ id: "RateLimiter", description: "Audit rate limiter" },
+			],
+		},
+		args: {
+			agent: "task",
+			context: [
+				"# Goal",
+				"Harden the HTTP auth stack before the release cut.",
+				"# Constraints",
+				"Touch only files under packages/server/src/auth/. Do not run gates.",
+			].join("\n"),
+			tasks: [
+				{
+					id: "AuthLoader",
+					description: "Load auth middleware",
+					assignment:
+						"Read packages/server/src/auth/session.ts and middleware.ts, then document the session-cookie validation flow and any TODOs.",
+				},
+				{
+					id: "RateLimiter",
+					description: "Audit rate limiter",
+					assignment:
+						"Inspect packages/server/src/auth/rate-limit.ts. Confirm the 429 path sets Retry-After and report gaps.",
+				},
+				{
+					id: "TokenRotation",
+					description: "Check token rotation",
+					assignment:
+						"Trace refresh-token rotation in packages/server/src/auth/tokens.ts and flag any reuse window.",
+				},
+			],
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: "3 agents completed: AuthLoader, RateLimiter, TokenRotation.",
+				},
+			],
+			details: {
+				projectAgentsDir: null,
+				totalDurationMs: 48_200,
+				usage: { cost: { total: 0.34 } },
+				results: [
+					{
+						index: 0,
+						id: "AuthLoader",
+						agent: "task",
+						agentSource: "bundled",
+						description: "Load auth middleware",
+						task: "Read packages/server/src/auth/session.ts and middleware.ts",
+						assignment:
+							"Read packages/server/src/auth/session.ts and middleware.ts, then document the session-cookie validation flow and any TODOs.",
+						exitCode: 0,
+						output: [
+							"Session validation runs in middleware.ts:42 via verifySessionCookie().",
+							"Cookies are HMAC-signed (SHA-256) and checked against the session store.",
+							"TODO at session.ts:88 — sliding-expiration refresh is stubbed.",
+						].join("\n"),
+						stderr: "",
+						truncated: false,
+						durationMs: 41_900,
+						tokens: 61_400,
+						contextTokens: 23_100,
+						contextWindow: 200_000,
+						resolvedModel: "anthropic/claude-sonnet",
+						usage: { cost: { total: 0.12 } },
+						outputMeta: { lineCount: 3, charCount: 214 },
+					},
+					{
+						index: 1,
+						id: "RateLimiter",
+						agent: "task",
+						agentSource: "bundled",
+						description: "Audit rate limiter",
+						task: "Inspect packages/server/src/auth/rate-limit.ts",
+						assignment:
+							"Inspect packages/server/src/auth/rate-limit.ts. Confirm the 429 path sets Retry-After and report gaps.",
+						exitCode: 0,
+						output: [
+							"rate-limit.ts uses a fixed-window counter keyed by client IP.",
+							"429 responses set Retry-After (rate-limit.ts:57).",
+							"Gap: no per-account limit, so a botnet across IPs bypasses the cap.",
+						].join("\n"),
+						stderr: "",
+						truncated: false,
+						durationMs: 38_500,
+						tokens: 54_800,
+						contextTokens: 19_700,
+						contextWindow: 200_000,
+						resolvedModel: "anthropic/claude-sonnet",
+						usage: { cost: { total: 0.1 } },
+						outputMeta: { lineCount: 3, charCount: 198 },
+					},
+					{
+						index: 2,
+						id: "TokenRotation",
+						agent: "task",
+						agentSource: "bundled",
+						description: "Check token rotation",
+						task: "Trace refresh-token rotation in packages/server/src/auth/tokens.ts",
+						assignment:
+							"Trace refresh-token rotation in packages/server/src/auth/tokens.ts and flag any reuse window.",
+						exitCode: 0,
+						output: [
+							"Refresh tokens rotate on every use (tokens.ts:120) and the old jti is revoked.",
+							"Reuse of a rotated token triggers full-family revocation — no reuse window found.",
+						].join("\n"),
+						stderr: "",
+						truncated: false,
+						durationMs: 48_200,
+						tokens: 49_200,
+						contextTokens: 17_500,
+						contextWindow: 200_000,
+						resolvedModel: "anthropic/claude-sonnet",
+						usage: { cost: { total: 0.12 } },
+						outputMeta: { lineCount: 2, charCount: 160 },
+					},
+				],
+			},
+		},
+		errorResult: {
+			isError: true,
+			content: [
+				{
+					type: "text",
+					text: "1 of 3 agents failed: RateLimiter.",
+				},
+			],
+			details: {
+				projectAgentsDir: null,
+				totalDurationMs: 39_400,
+				usage: { cost: { total: 0.21 } },
+				results: [
+					{
+						index: 0,
+						id: "AuthLoader",
+						agent: "task",
+						agentSource: "bundled",
+						description: "Load auth middleware",
+						task: "Read packages/server/src/auth/session.ts and middleware.ts",
+						assignment:
+							"Read packages/server/src/auth/session.ts and middleware.ts, then document the session-cookie validation flow and any TODOs.",
+						exitCode: 0,
+						output: "Session validation runs in middleware.ts:42 via verifySessionCookie().",
+						stderr: "",
+						truncated: false,
+						durationMs: 31_200,
+						tokens: 58_100,
+						contextTokens: 21_900,
+						contextWindow: 200_000,
+						resolvedModel: "anthropic/claude-sonnet",
+						usage: { cost: { total: 0.11 } },
+						outputMeta: { lineCount: 1, charCount: 70 },
+					},
+					{
+						index: 1,
+						id: "RateLimiter",
+						agent: "task",
+						agentSource: "bundled",
+						description: "Audit rate limiter",
+						task: "Inspect packages/server/src/auth/rate-limit.ts",
+						assignment:
+							"Inspect packages/server/src/auth/rate-limit.ts. Confirm the 429 path sets Retry-After and report gaps.",
+						exitCode: 1,
+						output: "",
+						stderr: "ENOENT: packages/server/src/auth/rate-limit.ts",
+						truncated: false,
+						durationMs: 9_800,
+						tokens: 12_300,
+						contextTokens: 6_400,
+						contextWindow: 200_000,
+						resolvedModel: "anthropic/claude-sonnet",
+						usage: { cost: { total: 0.1 } },
+						error: "Subagent exited 1: target file packages/server/src/auth/rate-limit.ts does not exist.",
+						outputMeta: { lineCount: 0, charCount: 0 },
+					},
+				],
+			},
+		},
+	},
+
+	goal: {
+		label: "Goal",
+		// Streaming: op is "create"; objective text still being typed.
+		streamingArgs: { op: "create", objective: "Ship the auth hardening" },
+		args: {
+			op: "create",
+			objective: "Ship the auth hardening pass: per-account rate limits and sliding session expiry.",
+			token_budget: 500_000,
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: "Goal set. Working toward: Ship the auth hardening pass.",
+				},
+			],
+			details: {
+				op: "create",
+				remainingTokens: 451_800,
+				completionBudgetReport: null,
+				goal: {
+					id: "goal_8f2a",
+					objective: "Ship the auth hardening pass: per-account rate limits and sliding session expiry.",
+					status: "active",
+					tokenBudget: 500_000,
+					tokensUsed: 48_200,
+					timeUsedSeconds: 312,
+					createdAt: 1_749_200_000_000,
+					updatedAt: 1_749_200_312_000,
+				},
+			},
+		},
+		errorResult: {
+			isError: true,
+			content: [{ type: "text", text: "Goal tool failed: objective is required when op=create." }],
+			details: { op: "create" },
+		},
+	},
+
+	job: {
+		label: "Job",
+		// Streaming: polling a single job id; the second id is still arriving.
+		streamingArgs: { poll: ["job_a1"] },
+		args: { poll: ["job_a1", "job_b2", "job_c3"] },
+		result: {
+			content: [{ type: "text", text: "3 jobs settled." }],
+			details: {
+				jobs: [
+					{
+						id: "job_a1",
+						type: "bash",
+						status: "completed",
+						label: "bun test packages/server/test/auth.test.ts",
+						durationMs: 18_400,
+						resultText: "42 pass, 0 fail (18.4s)",
+					},
+					{
+						id: "job_b2",
+						type: "task",
+						status: "completed",
+						label: "Migrate rate limiter to a sliding window",
+						durationMs: 96_700,
+						resultText: "Rewrote rate-limit.ts to a token-bucket; added per-account keys.",
+					},
+					{
+						id: "job_c3",
+						type: "bash",
+						status: "failed",
+						label: "bunx biome check packages/server/src/auth",
+						durationMs: 4_100,
+						errorText: "biome: 2 errors in tokens.ts — noUnusedVariables, useConst",
+					},
+				],
+			},
+		},
+		errorResult: {
+			isError: true,
+			content: [{ type: "text", text: "Job cancelled by user." }],
+			details: {
+				jobs: [
+					{
+						id: "job_d4",
+						type: "task",
+						status: "cancelled",
+						label: "Refactor the session store to Redis",
+						durationMs: 52_300,
+						errorText: "Aborted: superseded by goal re-scope.",
+					},
+				],
+				cancelled: [{ id: "job_d4", status: "cancelled" }],
+			},
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/codeintel.ts b/packages/coding-agent/src/cli/gallery-fixtures/codeintel.ts
new file mode 100644
index 000000000..0d9faca65
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/codeintel.ts
@@ -0,0 +1,188 @@
+/** Gallery fixtures for the code-intelligence tools (lsp, debug). */
+import type { GalleryFixture } from "./types";
+
+export const codeintelFixtures: Record<string, GalleryFixture> = {
+	lsp: {
+		label: "LSP",
+		customRendered: true,
+		streamingArgs: {
+			action: "references",
+			file: "src/server/auth.ts",
+		},
+		args: {
+			action: "references",
+			file: "src/server/auth.ts",
+			line: 42,
+			symbol: "validateToken",
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"Found 6 reference(s):",
+						"  src/server/auth.ts:42:14",
+						"    41: ",
+						"    42: export function validateToken(token: string): Claims {",
+						"    43:   const claims = verifyJwt(token);",
+						"  src/server/auth.ts:118:21",
+						'    117:   if (!header) throw new HttpError(401, "missing token");',
+						"    118:   const claims = validateToken(stripBearer(header));",
+						"    119:   return claims.sub;",
+						"  src/server/middleware/session.ts:57:18",
+						"    56:   const token = req.cookies.session;",
+						"    57:   const claims = validateToken(token);",
+						"    58:   req.userId = claims.sub;",
+						"  src/server/router.ts:153:20",
+						"    152: router.use(async (req, res, next) => {",
+						"    153:   req.claims = await validateToken(req.token);",
+						"    154:   next();",
+						"  test/auth.test.ts:24:9",
+						'    23: it("rejects expired tokens", () => {',
+						"    24:   expect(() => validateToken(expired)).toThrow(/expired/);",
+						"    25: });",
+						"  test/auth.test.ts:41:9",
+						'    40: it("accepts valid tokens", () => {',
+						"    41:   const claims = validateToken(signed);",
+						'    42:   expect(claims.sub).toBe("u_123");',
+					].join("\n"),
+				},
+			],
+			details: {
+				serverName: "typescript-language-server",
+				action: "references",
+				success: true,
+				request: {
+					action: "references",
+					file: "src/server/auth.ts",
+					line: 42,
+					symbol: "validateToken",
+				},
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "No language server found for this file",
+				},
+			],
+			isError: true,
+			details: {
+				serverName: "typescript-language-server",
+				action: "references",
+				success: false,
+				request: {
+					action: "references",
+					file: "src/server/auth.ts",
+					line: 42,
+					symbol: "validateToken",
+				},
+			},
+		},
+	},
+
+	debug: {
+		label: "Debug",
+		streamingArgs: {
+			action: "stack_trace",
+		},
+		args: {
+			action: "stack_trace",
+			levels: 20,
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"Stack trace:",
+						"- #1000 validate_token @ app/server.py:42:14",
+						"- #1001 authenticate @ app/server.py:88:9",
+						"- #1002 handle_request @ app/router.py:153:20",
+						"- #1003 dispatch @ app/router.py:97:5",
+						"- #1004 <module> @ app/server.py:212:1",
+					].join("\n"),
+				},
+			],
+			details: {
+				action: "stack_trace",
+				success: true,
+				snapshot: {
+					id: "dbg-1",
+					adapter: "debugpy",
+					cwd: "/Users/dev/project",
+					program: "./app/server.py",
+					status: "stopped",
+					launchedAt: "2026-06-06T14:21:08.412Z",
+					lastUsedAt: "2026-06-06T14:22:55.901Z",
+					threadId: 1,
+					frameId: 1000,
+					stopReason: "breakpoint",
+					stopDescription: "breakpoint 2",
+					frameName: "validate_token",
+					instructionPointerReference: "0x00000001000034a8",
+					source: { name: "server.py", path: "app/server.py" },
+					line: 42,
+					column: 14,
+					breakpointFiles: 1,
+					breakpointCount: 2,
+					functionBreakpointCount: 0,
+					outputBytes: 248,
+					outputTruncated: false,
+					needsConfigurationDone: false,
+				},
+				stackFrames: [
+					{
+						id: 1000,
+						name: "validate_token",
+						source: { name: "server.py", path: "app/server.py" },
+						line: 42,
+						column: 14,
+					},
+					{
+						id: 1001,
+						name: "authenticate",
+						source: { name: "server.py", path: "app/server.py" },
+						line: 88,
+						column: 9,
+					},
+					{
+						id: 1002,
+						name: "handle_request",
+						source: { name: "router.py", path: "app/router.py" },
+						line: 153,
+						column: 20,
+					},
+					{
+						id: 1003,
+						name: "dispatch",
+						source: { name: "router.py", path: "app/router.py" },
+						line: 97,
+						column: 5,
+					},
+					{
+						id: 1004,
+						name: "<module>",
+						source: { name: "server.py", path: "app/server.py" },
+						line: 212,
+						column: 1,
+					},
+				],
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "No active debug session. Launch or attach first.",
+				},
+			],
+			isError: true,
+			details: {
+				action: "stack_trace",
+				success: false,
+			},
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/edit.ts b/packages/coding-agent/src/cli/gallery-fixtures/edit.ts
new file mode 100644
index 000000000..b1da61d4b
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/edit.ts
@@ -0,0 +1,194 @@
+/** Gallery fixtures for the edit tools (edit, apply_patch, ast_edit). */
+import type { GalleryFixture } from "./types";
+
+export const editFixtures: Record<string, GalleryFixture> = {
+	edit: {
+		label: "Edit",
+		editMode: "replace",
+		// `previewDiff` is surfaced verbatim by the renderer's call preview, and the
+		// harness diff strategy skips `{ file_path, previewDiff }` (no `path`/`edits`),
+		// so the canned diff survives the streaming and progress states.
+		streamingArgs: {
+			file_path: "packages/coding-agent/src/tools/read.ts",
+			previewDiff: [
+				"@@ -88,3 +88,4 @@",
+				" 	const offset = args.offset ?? 1;",
+				"-	const limit = args.limit ?? 2000;",
+				"+	const limit = args.limit ?? 4000;",
+			].join("\n"),
+		},
+		args: {
+			file_path: "packages/coding-agent/src/tools/read.ts",
+			previewDiff: [
+				"@@ -88,5 +88,6 @@",
+				" 	const offset = args.offset ?? 1;",
+				"-	const limit = args.limit ?? 2000;",
+				"+	const limit = args.limit ?? 4000;",
+				" 	const raw = await Bun.file(path).text();",
+				"-	return raw.slice(offset, offset + limit);",
+				'+	return raw.split("\\n").slice(offset - 1, offset - 1 + limit).join("\\n");',
+			].join("\n"),
+		},
+		result: {
+			content: [{ type: "text", text: "Edited packages/coding-agent/src/tools/read.ts (1 hunk, +3 -2)" }],
+			details: {
+				path: "packages/coding-agent/src/tools/read.ts",
+				firstChangedLine: 89,
+				diff: [
+					"@@ -88,5 +88,6 @@",
+					" 	const offset = args.offset ?? 1;",
+					"-	const limit = args.limit ?? 2000;",
+					"+	const limit = args.limit ?? 4000;",
+					" 	const raw = await Bun.file(path).text();",
+					"-	return raw.slice(offset, offset + limit);",
+					'+	return raw.split("\\n").slice(offset - 1, offset - 1 + limit).join("\\n");',
+				].join("\n"),
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "Edit failed: the search text was not found in packages/coding-agent/src/tools/read.ts",
+				},
+			],
+			isError: true,
+			details: {
+				path: "packages/coding-agent/src/tools/read.ts",
+				diff: "",
+				errorText:
+					"No match for the search text. Expected `const limit = args.limit ?? 2000;` near line 89, but the file has `const limit = args.limit ?? 1000;`. Re-read the file and retry with the current contents.",
+			},
+		},
+	},
+
+	apply_patch: {
+		label: "Apply Patch",
+		editMode: "apply_patch",
+		streamingArgs: {
+			file_path: "packages/coding-agent/src/edit/renderer.ts",
+			previewDiff: [
+				"@@ -464,2 +464,2 @@",
+				"-		fileCount = countEditFiles(editArgs.edits);",
+				"+		fileCount = countDistinctFiles(editArgs.edits);",
+			].join("\n"),
+		},
+		args: {
+			file_path: "packages/coding-agent/src/edit/renderer.ts",
+			previewDiff: [
+				"@@ -177,4 +177,4 @@",
+				" /** Count distinct file paths in an edits array. */",
+				"-function countEditFiles(edits: EditRenderEntry[]): number {",
+				"+function countDistinctFiles(edits: EditRenderEntry[]): number {",
+				" 	return new Set(edits.map(edit => filePathFromEditEntry(edit.path)).filter(Boolean)).size;",
+				" }",
+				"@@ -467,2 +467,2 @@",
+				"-		fileCount = countEditFiles(editArgs.edits);",
+				"+		fileCount = countDistinctFiles(editArgs.edits);",
+			].join("\n"),
+		},
+		result: {
+			content: [
+				{ type: "text", text: "Applied patch to packages/coding-agent/src/edit/renderer.ts (2 hunks, +2 -2)" },
+			],
+			details: {
+				op: "update",
+				path: "packages/coding-agent/src/edit/renderer.ts",
+				firstChangedLine: 178,
+				diff: [
+					"@@ -177,4 +177,4 @@",
+					" /** Count distinct file paths in an edits array. */",
+					"-function countEditFiles(edits: EditRenderEntry[]): number {",
+					"+function countDistinctFiles(edits: EditRenderEntry[]): number {",
+					" 	return new Set(edits.map(edit => filePathFromEditEntry(edit.path)).filter(Boolean)).size;",
+					" }",
+					"@@ -467,2 +467,2 @@",
+					"-		fileCount = countEditFiles(editArgs.edits);",
+					"+		fileCount = countDistinctFiles(editArgs.edits);",
+				].join("\n"),
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "Apply patch failed: context does not match at line 177 of packages/coding-agent/src/edit/renderer.ts",
+				},
+			],
+			isError: true,
+			details: {
+				op: "update",
+				path: "packages/coding-agent/src/edit/renderer.ts",
+				diff: "",
+				errorText:
+					"Hunk @@ -177,4 +177,4 @@ failed to apply: the context line `function countEditFiles(edits: EditRenderEntry[]): number {` does not match the file. The file may have changed since it was read.",
+			},
+		},
+	},
+
+	ast_edit: {
+		label: "AST Edit",
+		streamingArgs: {
+			ops: [{ pat: "countEditFiles($$$ARGS)" }],
+			paths: ["packages/coding-agent/src/**/*.ts"],
+		},
+		args: {
+			ops: [{ pat: "countEditFiles($$$ARGS)", out: "countDistinctFiles($$$ARGS)" }],
+			paths: ["packages/coding-agent/src/**/*.ts"],
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"# edit/renderer.ts (2 replacements)",
+						"-468:		fileCount = countEditFiles(editArgs.edits);",
+						"+468:		fileCount = countDistinctFiles(editArgs.edits);",
+						"-488:		const totalFiles = args?.edits ? countEditFiles(args.edits) : 0;",
+						"+488:		const totalFiles = args?.edits ? countDistinctFiles(args.edits) : 0;",
+						"",
+						"# tools/tool-result.ts (1 replacement)",
+						"-42:	return countEditFiles(files);",
+						"+42:	return countDistinctFiles(files);",
+					].join("\n"),
+				},
+			],
+			details: {
+				totalReplacements: 3,
+				filesTouched: 2,
+				filesSearched: 214,
+				applied: false,
+				limitReached: false,
+				scopePath: "packages/coding-agent/src",
+				searchPath: "/Users/dev/Projects/pi/packages/coding-agent/src",
+				files: ["edit/renderer.ts", "tools/tool-result.ts"],
+				fileReplacements: [
+					{ path: "edit/renderer.ts", count: 2 },
+					{ path: "tools/tool-result.ts", count: 1 },
+				],
+				displayContent: [
+					"# edit/",
+					"## renderer.ts (2 replacements)",
+					"-468│		fileCount = countEditFiles(editArgs.edits);",
+					"+468│		fileCount = countDistinctFiles(editArgs.edits);",
+					"-488│		const totalFiles = args?.edits ? countEditFiles(args.edits) : 0;",
+					"+488│		const totalFiles = args?.edits ? countDistinctFiles(args.edits) : 0;",
+					"",
+					"# tools/",
+					"## tool-result.ts (1 replacement)",
+					"-42│	return countEditFiles(files);",
+					"+42│	return countDistinctFiles(files);",
+				].join("\n"),
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "Pattern parse error in ops[0].pat: unbalanced parenthesis in `countEditFiles($$$ARGS`",
+				},
+			],
+			isError: true,
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/fs.ts b/packages/coding-agent/src/cli/gallery-fixtures/fs.ts
new file mode 100644
index 000000000..290571544
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/fs.ts
@@ -0,0 +1,153 @@
+// biome-ignore-all lint/suspicious/noTemplateCurlyInString: sample source-code strings (read fixtures) intentionally contain literal ${...}.
+// Gallery fixtures for the filesystem tools (read, write, find).
+import type { GalleryFixture } from "./types";
+
+const readSnippet = [
+	"export const findToolRenderer = {",
+	"\tinline: true,",
+	"\trenderCall(args: FindRenderArgs, _options: RenderResultOptions, uiTheme: Theme): Component {",
+	"\t\tconst meta: string[] = [];",
+	"\t\tif (args.limit !== undefined) meta.push(`limit:${args.limit}`);",
+	"",
+	"\t\tconst text = renderStatusLine(",
+	'\t\t\t{ icon: "pending", title: "Find", description: formatFindRenderPaths(args.paths) || "*", meta },',
+	"\t\t\tuiTheme,",
+	"\t\t);",
+	"\t\treturn new Text(text, 0, 0);",
+	"\t},",
+].join("\n");
+
+const writtenContent = [
+	'import { describe, expect, it } from "bun:test";',
+	'import { parseSel } from "../src/tools/read";',
+	"",
+	'describe("parseSel", () => {',
+	'\tit("parses a single line range", () => {',
+	'\t\texpect(parseSel("42-58")).toEqual({',
+	'\t\t\tkind: "lines",',
+	"\t\t\tranges: [{ startLine: 42, endLine: 58 }],",
+	"\t\t});",
+	"\t});",
+	"",
+	'\tit("treats raw as a verbatim selector", () => {',
+	'\t\texpect(parseSel("raw")).toEqual({ kind: "raw" });',
+	"\t});",
+	"});",
+	"",
+].join("\n");
+
+export const fsFixtures: Record<string, GalleryFixture> = {
+	read: {
+		label: "Read",
+		// Streaming: path still being typed, selector not yet appended.
+		streamingArgs: { path: "packages/coding-agent/src/tools/find" },
+		args: { path: "packages/coding-agent/src/tools/find.ts:437-448" },
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"[packages/coding-agent/src/tools/find.ts#E48E]",
+						"437:export const findToolRenderer = {",
+						"438:\tinline: true,",
+						"439:\trenderCall(args: FindRenderArgs, _options: RenderResultOptions, uiTheme: Theme): Component {",
+						"440:\t\tconst meta: string[] = [];",
+						"441:\t\tif (args.limit !== undefined) meta.push(`limit:${args.limit}`);",
+						"442:",
+						"443:\t\tconst text = renderStatusLine(",
+						'444:\t\t\t{ icon: "pending", title: "Find", description: formatFindRenderPaths(args.paths) || "*", meta },',
+						"445:\t\t\tuiTheme,",
+						"446:\t\t);",
+						"447:\t\treturn new Text(text, 0, 0);",
+						"448:\t},",
+					].join("\n"),
+				},
+			],
+			details: {
+				kind: "file",
+				resolvedPath: "/Users/dev/Projects/pi/packages/coding-agent/src/tools/find.ts",
+				contentType: "text/typescript",
+				displayContent: { text: readSnippet, startLine: 437 },
+			},
+		},
+		errorResult: {
+			isError: true,
+			content: [
+				{
+					type: "text",
+					text: "Error: ENOENT: no such file or directory, open 'packages/coding-agent/src/tools/find.ts'",
+				},
+			],
+		},
+	},
+
+	write: {
+		label: "Write",
+		// Streaming: path known, content still arriving (only the imports so far).
+		streamingArgs: {
+			path: "packages/coding-agent/test/parse-sel.test.ts",
+			content: 'import { describe, expect, it } from "bun:test";\nimport { parseSel } from "../src/tools/read";\n',
+		},
+		args: {
+			path: "packages/coding-agent/test/parse-sel.test.ts",
+			content: writtenContent,
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: "Created packages/coding-agent/test/parse-sel.test.ts (17 lines, 412 bytes).",
+				},
+			],
+			details: {},
+		},
+		errorResult: {
+			isError: true,
+			content: [
+				{
+					type: "text",
+					text: "Error: EACCES: permission denied, open 'packages/coding-agent/test/parse-sel.test.ts'",
+				},
+			],
+		},
+	},
+
+	find: {
+		label: "Find",
+		// Streaming: glob half-typed, no limit yet.
+		streamingArgs: { paths: ["packages/coding-agent/src/tools/*-render"] },
+		args: { paths: ["packages/coding-agent/src/**/*.test.ts"], limit: 50 },
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"packages/coding-agent/src/tools/read.test.ts",
+						"packages/coding-agent/src/tools/write.test.ts",
+						"packages/coding-agent/src/tools/find.test.ts",
+						"packages/coding-agent/src/cli/gallery-cli.test.ts",
+						"packages/coding-agent/src/edit/edit.test.ts",
+					].join("\n"),
+				},
+			],
+			details: {
+				scopePath: "packages/coding-agent/src",
+				cwd: "/Users/dev/Projects/pi",
+				fileCount: 5,
+				truncated: false,
+				files: [
+					"packages/coding-agent/src/cli/gallery-cli.test.ts",
+					"packages/coding-agent/src/edit/edit.test.ts",
+					"packages/coding-agent/src/tools/find.test.ts",
+					"packages/coding-agent/src/tools/read.test.ts",
+					"packages/coding-agent/src/tools/write.test.ts",
+				],
+			},
+		},
+		errorResult: {
+			isError: true,
+			content: [{ type: "text", text: "Find failed: invalid glob pattern '[unclosed'." }],
+			details: { error: "invalid glob pattern '[unclosed'" },
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/index.ts b/packages/coding-agent/src/cli/gallery-fixtures/index.ts
new file mode 100644
index 000000000..404082263
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/index.ts
@@ -0,0 +1,40 @@
+/**
+ * Aggregated sample data for the `omp gallery` command.
+ *
+ * Each fixture drives one tool's renderer through the four lifecycle states the
+ * gallery showcases: arguments streaming in, arguments complete but awaiting a
+ * result, a successful result, and a failed result. The data is intentionally
+ * hand-written (rather than schema-derived) so the gallery reflects what a real
+ * tool call looks like — the whole point is visual QA of the renderers.
+ *
+ * Fixtures are grouped by subsystem into sibling modules and merged here.
+ * Adding a tool to one of those groups is enough for the gallery to render it.
+ * Tools present in the renderer registry but missing here fall back to a
+ * generic fixture (see `gallery-cli.ts`), so the gallery never crashes on a
+ * newly added tool — it just looks plain until a fixture is supplied.
+ */
+import { agenticFixtures } from "./agentic";
+import { codeintelFixtures } from "./codeintel";
+import { editFixtures } from "./edit";
+import { fsFixtures } from "./fs";
+import { interactionFixtures } from "./interaction";
+import { memoryFixtures } from "./memory";
+import { miscFixtures } from "./misc";
+import { searchFixtures } from "./search";
+import { shellFixtures } from "./shell";
+import { webFixtures } from "./web";
+
+export * from "./types";
+
+export const galleryFixtures = {
+	...interactionFixtures,
+	...shellFixtures,
+	...fsFixtures,
+	...searchFixtures,
+	...editFixtures,
+	...agenticFixtures,
+	...memoryFixtures,
+	...webFixtures,
+	...codeintelFixtures,
+	...miscFixtures,
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/interaction.ts b/packages/coding-agent/src/cli/gallery-fixtures/interaction.ts
new file mode 100644
index 000000000..34da85a14
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/interaction.ts
@@ -0,0 +1,49 @@
+/** Gallery fixtures for the todo / ask / resolve interaction tools. */
+import type { GalleryFixture } from "./types";
+
+export const interactionFixtures: Record<string, GalleryFixture> = {
+	todo: {
+		label: "Todo",
+		streamingArgs: {
+			ops: [{ op: "init", list: [{ phase: "Foundation", items: ["Scaffold crate"] }] }],
+		},
+		args: {
+			ops: [
+				{
+					op: "init",
+					list: [
+						{ phase: "Foundation", items: ["Scaffold crate", "Wire workspace"] },
+						{ phase: "Auth", items: ["Port credential store", "Wire OAuth providers"] },
+					],
+				},
+			],
+		},
+		result: {
+			content: [{ type: "text", text: "Initialized 4 tasks across 2 phases" }],
+			details: {
+				storage: "session",
+				phases: [
+					{
+						name: "Foundation",
+						tasks: [
+							{ content: "Scaffold crate", status: "done" },
+							{ content: "Wire workspace", status: "in_progress" },
+						],
+					},
+					{
+						name: "Auth",
+						tasks: [
+							{ content: "Port credential store", status: "pending" },
+							{ content: "Wire OAuth providers", status: "pending" },
+						],
+					},
+				],
+				completedTasks: [{ phase: "Foundation", content: "Scaffold crate" }],
+			},
+		},
+		errorResult: {
+			content: [{ type: "text", text: "Unknown phase 'Auth' — initialize the list first" }],
+			isError: true,
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/memory.ts b/packages/coding-agent/src/cli/gallery-fixtures/memory.ts
new file mode 100644
index 000000000..01b70e7e4
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/memory.ts
@@ -0,0 +1,81 @@
+// Gallery fixtures for the long-term memory tools (retain, recall, reflect).
+import type { GalleryFixture } from "./types";
+
+export const memoryFixtures: Record<string, GalleryFixture> = {
+	retain: {
+		label: "Retain",
+		// Streaming: first item complete, second still arriving without a context.
+		streamingArgs: {
+			items: [{ content: "User prefers Bun over Node for all new scripts in this repo." }],
+		},
+		args: {
+			items: [
+				{
+					content: "User prefers Bun over Node for all new scripts in this repo.",
+					context: "Established while wiring up the gallery command tooling.",
+				},
+				{
+					content: "The TUI renderers live in packages/coding-agent/src/tools/*-render.ts.",
+					context: "Discovered during the gallery-fixtures task.",
+				},
+			],
+		},
+		result: {
+			content: [{ type: "text", text: "2 memories stored." }],
+			details: { count: 2 },
+		},
+		errorResult: {
+			isError: true,
+			content: [{ type: "text", text: "Retain failed: memory store is not initialized." }],
+		},
+	},
+
+	recall: {
+		label: "Recall",
+		// Streaming: query partially typed.
+		streamingArgs: { query: "bun vs node" },
+		args: { query: "Which runtime does the user prefer for scripts?" },
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"Found 2 relevant memories:",
+						"",
+						"1. [0.92] User prefers Bun over Node for all new scripts in this repo.",
+						"   (Established while wiring up the gallery command tooling.)",
+						"2. [0.78] The TUI renderers live in packages/coding-agent/src/tools/*-render.ts.",
+						"   (Discovered during the gallery-fixtures task.)",
+					].join("\n"),
+				},
+			],
+		},
+		errorResult: {
+			isError: true,
+			content: [{ type: "text", text: "Recall failed: vector index unavailable." }],
+		},
+	},
+
+	reflect: {
+		label: "Reflect",
+		streamingArgs: { query: "what have we learned about the user's" },
+		args: { query: "What have we learned about the user's tooling preferences?" },
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"The user consistently favors Bun as the runtime for scripts in this",
+						"repository, avoiding Node where possible. They also track the location",
+						"of TUI renderers under packages/coding-agent/src/tools, suggesting an",
+						"interest in keeping rendering logic discoverable and well-organized.",
+					].join("\n"),
+				},
+			],
+		},
+		errorResult: {
+			isError: true,
+			content: [{ type: "text", text: "Reflect failed: no memories matched the query." }],
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/misc.ts b/packages/coding-agent/src/cli/gallery-fixtures/misc.ts
new file mode 100644
index 000000000..4860d01f2
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/misc.ts
@@ -0,0 +1,250 @@
+/** Gallery fixtures for the ask / resolve / ssh / github / inspect_image tools. */
+import type { GalleryFixture } from "./types";
+
+export const miscFixtures: Record<string, GalleryFixture> = {
+	ask: {
+		label: "Ask",
+		streamingArgs: {
+			questions: [
+				{
+					id: "db",
+					question: "Which database should the new service use?",
+					options: [{ label: "Postgres" }],
+				},
+			],
+		},
+		args: {
+			questions: [
+				{
+					id: "db",
+					question: "Which database should the new service use?",
+					options: [
+						{ label: "Postgres", description: "Relational, strong consistency, JSONB support" },
+						{ label: "SQLite", description: "Embedded, zero-ops, great for single-node" },
+						{ label: "MongoDB", description: "Document store, flexible schema" },
+					],
+					recommended: 0,
+				},
+				{
+					id: "features",
+					question: "Which auth flows should ship in v1?",
+					options: [
+						{ label: "Email + password" },
+						{ label: "OAuth (Google, GitHub)" },
+						{ label: "Magic links" },
+						{ label: "SAML SSO", description: "Enterprise; can be deferred" },
+					],
+					multi: true,
+				},
+			],
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: "db: Postgres\nfeatures: Email + password, OAuth (Google, GitHub)",
+				},
+			],
+			details: {
+				results: [
+					{
+						id: "db",
+						question: "Which database should the new service use?",
+						options: ["Postgres", "SQLite", "MongoDB"],
+						multi: false,
+						selectedOptions: ["Postgres"],
+					},
+					{
+						id: "features",
+						question: "Which auth flows should ship in v1?",
+						options: ["Email + password", "OAuth (Google, GitHub)", "Magic links", "SAML SSO"],
+						multi: true,
+						selectedOptions: ["Email + password", "OAuth (Google, GitHub)"],
+					},
+				],
+			},
+		},
+		errorResult: {
+			content: [{ type: "text", text: "Prompt cancelled by user before any answer was given" }],
+			isError: true,
+		},
+	},
+
+	resolve: {
+		label: "Resolve",
+		streamingArgs: {
+			action: "apply",
+		},
+		args: {
+			action: "apply",
+			reason: "Rename is mechanical and the staged diff matches the intended refactor.",
+			extra: { title: "rename-usecredentials-hook" },
+		},
+		result: {
+			content: [{ type: "text", text: "Applied pending ast_edit: 7 replacements across 3 files" }],
+			details: {
+				action: "apply",
+				reason: "Rename is mechanical and the staged diff matches the intended refactor.",
+				extra: { title: "rename-usecredentials-hook" },
+				sourceToolName: "ast_edit",
+				label: "ast_edit: 7 replacements across 3 files",
+			},
+		},
+		errorResult: {
+			content: [{ type: "text", text: "No pending action to resolve" }],
+			isError: true,
+			details: {
+				action: "apply",
+				reason: "Rename is mechanical and the staged diff matches the intended refactor.",
+				sourceToolName: "ast_edit",
+				label: "ast_edit: 7 replacements across 3 files",
+			},
+		},
+	},
+
+	ssh: {
+		label: "SSH",
+		streamingArgs: {
+			host: "deploy@web-01",
+			command: "systemctl status",
+		},
+		args: {
+			host: "deploy@web-01",
+			command: "systemctl status omp-api --no-pager | head -n 12",
+			cwd: "/srv/omp",
+			timeout: 60,
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"● omp-api.service - Oh My Pi API",
+						"     Loaded: loaded (/etc/systemd/system/omp-api.service; enabled)",
+						"     Active: active (running) since Sat 2026-06-06 09:14:02 UTC; 3h 21min ago",
+						"   Main PID: 4812 (bun)",
+						"      Tasks: 17 (limit: 4915)",
+						"     Memory: 142.6M",
+						"        CPU: 38.214s",
+						"     CGroup: /system.slice/omp-api.service",
+						"             └─4812 /usr/local/bin/bun run dist/server.js",
+					].join("\n"),
+				},
+			],
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "ssh: connect to host web-01 port 22: Connection timed out",
+				},
+			],
+			isError: true,
+		},
+	},
+
+	github: {
+		label: "GitHub",
+		streamingArgs: {
+			op: "search_prs",
+			query: "is:open author:@me",
+		},
+		args: {
+			op: "search_prs",
+			query: "is:open review-requested:@me sort:updated",
+			repo: "oh-my-pi/pi",
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"#1842  feat(tui): virtualized scrollback for tool output     openyou · 2h ago   +312 -47",
+						"#1839  fix(agent): retry stream on transient 529             dvir   · 5h ago   +18 -4",
+						"#1830  refactor(edit): unify hashline + ast_edit previews    mira   · 1d ago   +540 -210",
+						"#1817  docs: document gallery fixtures contract             leo    · 2d ago   +96 -0",
+						"",
+						"4 open pull requests requesting your review",
+					].join("\n"),
+				},
+			],
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "gh: Could not resolve to a Repository with the name 'oh-my-pi/pi'. (HTTP 404)",
+				},
+			],
+			isError: true,
+		},
+	},
+
+	inspect_image: {
+		label: "Inspect Image",
+		streamingArgs: {
+			path: "docs/assets/dashboard-mock.png",
+		},
+		args: {
+			path: "docs/assets/dashboard-mock.png",
+			question: "What chart types are shown and roughly what layout does the dashboard use?",
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"The dashboard uses a two-column layout on a dark background.",
+						"Top row: four KPI cards (Revenue, Active Users, Churn, MRR) with sparklines.",
+						"Left column: a stacked area chart of weekly sessions over ~3 months.",
+						"Right column: a horizontal bar chart ranking the top 6 referrers.",
+						"Bottom: a paginated table of recent transactions with status pills.",
+					].join("\n"),
+				},
+			],
+			details: {
+				model: "claude-opus-4",
+				imagePath: "docs/assets/dashboard-mock.png",
+				mimeType: "image/png",
+			},
+		},
+		errorResult: {
+			content: [{ type: "text", text: "Image not found: docs/assets/dashboard-mock.png" }],
+			isError: true,
+			details: {
+				model: "claude-opus-4",
+				imagePath: "docs/assets/dashboard-mock.png",
+				mimeType: "image/png",
+			},
+		},
+	},
+
+	// Built-in tool with no dedicated renderer — exercises the generic fallback
+	// (`#formatToolExecution`) path so its padded, state-tinted block is QA'd.
+	report_tool_issue: {
+		label: "Report Tool Issue",
+		streamingArgs: { tool: "lsp" },
+		args: {
+			tool: "lsp",
+			report: "Rename returned no edit for an exported symbol that has 12 references",
+		},
+		result: { content: [{ type: "text", text: "Noted, thanks!" }] },
+		errorResult: {
+			content: [{ type: "text", text: "Could not record the report: issue tracker unreachable" }],
+			isError: true,
+		},
+	},
+
+	// Stand-in for a custom/extension tool that ships no renderer — same generic
+	// fallback path most MCP/extension tools take.
+	custom: {
+		label: "Custom Tool",
+		streamingArgs: { query: "weather" },
+		args: { query: "weather in Tokyo", units: "metric" },
+		result: { content: [{ type: "text", text: "Tokyo: 22°C, partly cloudy, humidity 64%." }] },
+		errorResult: {
+			content: [{ type: "text", text: "Upstream provider returned 503 Service Unavailable" }],
+			isError: true,
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/search.ts b/packages/coding-agent/src/cli/gallery-fixtures/search.ts
new file mode 100644
index 000000000..2a0962070
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/search.ts
@@ -0,0 +1,213 @@
+/** Gallery fixtures for the search tools (search, search_tool_bm25, ast_grep). */
+import type { GalleryFixture } from "./types";
+
+export const searchFixtures: Record<string, GalleryFixture> = {
+	search: {
+		label: "Search",
+		streamingArgs: {
+			pattern: "useState",
+		},
+		args: {
+			pattern: "useState",
+			paths: ["packages/tui/src"],
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"# packages/tui/src/components/",
+						"## SearchBox.tsx",
+						'18:  const [query, setQuery] = useState("");',
+						"19:  const [results, setResults] = useState<Match[]>([]);",
+						"## StatusBar.tsx",
+						"27:  const [expanded, setExpanded] = useState(false);",
+						"",
+						"# packages/tui/src/hooks/",
+						"## useDebounced.ts",
+						"9:  const [value, setValue] = useState(initial);",
+						"10:  const [pending, setPending] = useState(false);",
+					].join("\n"),
+				},
+			],
+			details: {
+				scopePath: "packages/tui/src",
+				searchPath: "/Users/dev/Projects/pi/packages/tui/src",
+				matchCount: 5,
+				fileCount: 3,
+				files: [
+					"packages/tui/src/components/SearchBox.tsx",
+					"packages/tui/src/components/StatusBar.tsx",
+					"packages/tui/src/hooks/useDebounced.ts",
+				],
+				fileMatches: [
+					{ path: "packages/tui/src/components/SearchBox.tsx", count: 2 },
+					{ path: "packages/tui/src/components/StatusBar.tsx", count: 1 },
+					{ path: "packages/tui/src/hooks/useDebounced.ts", count: 2 },
+				],
+				truncated: false,
+				displayContent: [
+					"# packages/tui/src/components/",
+					"## SearchBox.tsx",
+					'*18│  const [query, setQuery] = useState("");',
+					"*19│  const [results, setResults] = useState<Match[]>([]);",
+					"## StatusBar.tsx",
+					"*27│  const [expanded, setExpanded] = useState(false);",
+					"",
+					"# packages/tui/src/hooks/",
+					"## useDebounced.ts",
+					" *9│  const [value, setValue] = useState(initial);",
+					"*10│  const [pending, setPending] = useState(false);",
+				].join("\n"),
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "Invalid regex pattern: unclosed group near index 8",
+				},
+			],
+			isError: true,
+			details: {
+				error: "Invalid regex pattern: unclosed group near index 8",
+			},
+		},
+	},
+
+	search_tool_bm25: {
+		label: "SearchTools",
+		streamingArgs: {
+			query: "read pdf and ext",
+		},
+		args: {
+			query: "read pdf and extract tables",
+			limit: 5,
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: JSON.stringify({
+						query: "read pdf and extract tables",
+						activated_tools: ["docling_extract_tables", "docling_convert", "pdf_read_text"],
+						match_count: 4,
+						total_tools: 142,
+					}),
+				},
+			],
+			details: {
+				query: "read pdf and extract tables",
+				limit: 5,
+				total_tools: 142,
+				activated_tools: ["docling_extract_tables", "docling_convert", "pdf_read_text"],
+				active_selected_tools: ["read", "search", "edit", "bash"],
+				tools: [
+					{
+						name: "docling_extract_tables",
+						label: "Extract Tables",
+						description: "Extract tabular data from PDF documents into CSV or JSON rows.",
+						server_name: "docling",
+						mcp_tool_name: "extract_tables",
+						schema_keys: ["path", "pages", "format"],
+						score: 9.412037,
+					},
+					{
+						name: "docling_convert",
+						label: "Convert Document",
+						description: "Convert PDF, DOCX, or PPTX into structured Markdown with layout preserved.",
+						server_name: "docling",
+						mcp_tool_name: "convert",
+						schema_keys: ["path", "target", "ocr"],
+						score: 6.83102,
+					},
+					{
+						name: "pdf_read_text",
+						label: "Read PDF Text",
+						description: "Read raw text from a PDF, optionally scoped to a page range.",
+						server_name: "pdf-tools",
+						mcp_tool_name: "read_text",
+						schema_keys: ["path", "page_start", "page_end"],
+						score: 5.207884,
+					},
+					{
+						name: "tabula_scan",
+						label: "Scan Tables",
+						description: "Detect table bounding boxes on scanned PDF pages before extraction.",
+						server_name: "pdf-tools",
+						mcp_tool_name: "scan",
+						schema_keys: ["path", "dpi"],
+						score: 3.119556,
+					},
+				],
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "Tool discovery is disabled. Enable tools.discoveryMode or mcp.discoveryMode to use search_tool_bm25.",
+				},
+			],
+			isError: true,
+		},
+	},
+
+	ast_grep: {
+		label: "AST Grep",
+		streamingArgs: {
+			pat: "useState(",
+		},
+		args: {
+			pat: "useState($A)",
+			paths: ["packages/tui/src/components"],
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"# packages/tui/src/components/",
+						"## SearchBox.tsx",
+						'18:  const [query, setQuery] = useState("");',
+						'  meta: $A=""',
+						"## StatusBar.tsx",
+						"27:  const [expanded, setExpanded] = useState(false);",
+						"  meta: $A=false",
+					].join("\n"),
+				},
+			],
+			details: {
+				matchCount: 2,
+				fileCount: 2,
+				filesSearched: 14,
+				limitReached: false,
+				scopePath: "packages/tui/src/components",
+				searchPath: "/Users/dev/Projects/pi/packages/tui/src/components",
+				files: ["packages/tui/src/components/SearchBox.tsx", "packages/tui/src/components/StatusBar.tsx"],
+				fileMatches: [
+					{ path: "packages/tui/src/components/SearchBox.tsx", count: 1 },
+					{ path: "packages/tui/src/components/StatusBar.tsx", count: 1 },
+				],
+				displayContent: [
+					"# packages/tui/src/components/",
+					"## SearchBox.tsx",
+					'*18│  const [query, setQuery] = useState("");',
+					'  meta: $A=""',
+					"## StatusBar.tsx",
+					"*27│  const [expanded, setExpanded] = useState(false);",
+					"  meta: $A=false",
+				].join("\n"),
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: "Pattern parse error: incomplete node `useState(` — expected a closing `)`",
+				},
+			],
+			isError: true,
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/shell.ts b/packages/coding-agent/src/cli/gallery-fixtures/shell.ts
new file mode 100644
index 000000000..81bb0739d
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/shell.ts
@@ -0,0 +1,167 @@
+/** Gallery fixtures for the shell tools (bash, eval). */
+import type { GalleryFixture } from "./types";
+
+export const shellFixtures: Record<string, GalleryFixture> = {
+	bash: {
+		label: "Bash",
+		streamingArgs: {
+			command: "git status --short && git log --on",
+		},
+		args: {
+			command: "git status --short && git log --oneline -5",
+			cwd: "packages/coding-agent",
+			timeout: 30,
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						" M src/cli/gallery-cli.ts",
+						" M src/tools/bash.ts",
+						"?? src/cli/gallery-fixtures/shell.ts",
+						"a1b2c3d Wire gallery command into CLI dispatch",
+						"9f8e7d6 Add ToolExecutionComponent lifecycle states",
+						"4c5b6a7 Extract createShellRenderer from bashToolRenderer",
+						"2d3e4f5 Strip LLM-facing notices before TUI render",
+						"7a8b9c0 Cap preview lines in pending command block",
+					].join("\n"),
+				},
+			],
+			details: {
+				exitCode: 0,
+				wallTimeMs: 184,
+				timeoutSeconds: 30,
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"src/tools/bash.ts:1142:34 - error TS2339: Property 'requestedTimeoutSeconds' does not exist on type 'BashToolDetails'.",
+						"",
+						"1142   const requestedTimeoutSeconds = details?.requestedTimeoutSeconds;",
+						"                                            ~~~~~~~~~~~~~~~~~~~~~~~~",
+						"Found 1 error in src/tools/bash.ts:1142",
+					].join("\n"),
+				},
+			],
+			isError: true,
+			details: {
+				exitCode: 2,
+				wallTimeMs: 5120,
+				timeoutSeconds: 30,
+			},
+		},
+	},
+
+	eval: {
+		label: "Eval",
+		streamingArgs: {
+			cells: [
+				{
+					language: "py",
+					code: 'import json\nfrom pathlib import Path\n\ndata = json.loads(Path("package.js',
+					title: "load config",
+				},
+			],
+		},
+		args: {
+			cells: [
+				{
+					language: "py",
+					title: "load config",
+					code: [
+						"import json",
+						"from pathlib import Path",
+						"",
+						'data = json.loads(Path("package.json").read_text())',
+						'deps = data.get("dependencies", {})',
+						'print(f"{data[\\"name\\"]} v{data[\\"version\\"]}")',
+						'print(f"{len(deps)} dependencies")',
+						"display(sorted(deps)[:3])",
+					].join("\n"),
+				},
+			],
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: ["@oh-my-pi/coding-agent v0.42.0", "37 dependencies"].join("\n"),
+				},
+			],
+			details: {
+				language: "python",
+				languages: ["python"],
+				jsonOutputs: [["@ai-sdk/anthropic", "@oh-my-pi/pi-ai", "@oh-my-pi/pi-tui"]],
+				cells: [
+					{
+						index: 0,
+						title: "load config",
+						language: "python",
+						code: [
+							"import json",
+							"from pathlib import Path",
+							"",
+							'data = json.loads(Path("package.json").read_text())',
+							'deps = data.get("dependencies", {})',
+							'print(f"{data[\\"name\\"]} v{data[\\"version\\"]}")',
+							'print(f"{len(deps)} dependencies")',
+							"display(sorted(deps)[:3])",
+						].join("\n"),
+						output: ["@oh-my-pi/coding-agent v0.42.0", "37 dependencies"].join("\n"),
+						status: "complete",
+						durationMs: 64,
+						exitCode: 0,
+					},
+				],
+			},
+		},
+		errorResult: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"Traceback (most recent call last):",
+						'  File "<cell 0>", line 4, in <module>',
+						'    data = json.loads(Path("package.json").read_text())',
+						"          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^",
+						"json.decoder.JSONDecodeError: Expecting ',' delimiter: line 12 column 3 (char 318)",
+					].join("\n"),
+				},
+			],
+			isError: true,
+			details: {
+				language: "python",
+				languages: ["python"],
+				isError: true,
+				cells: [
+					{
+						index: 0,
+						title: "load config",
+						language: "python",
+						code: [
+							"import json",
+							"from pathlib import Path",
+							"",
+							'data = json.loads(Path("package.json").read_text())',
+							'deps = data.get("dependencies", {})',
+							'print(f"{data[\\"name\\"]} v{data[\\"version\\"]}")',
+						].join("\n"),
+						output: [
+							"Traceback (most recent call last):",
+							'  File "<cell 0>", line 4, in <module>',
+							'    data = json.loads(Path("package.json").read_text())',
+							"json.decoder.JSONDecodeError: Expecting ',' delimiter: line 12 column 3 (char 318)",
+						].join("\n"),
+						status: "error",
+						durationMs: 41,
+						exitCode: 1,
+					},
+				],
+			},
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/types.ts b/packages/coding-agent/src/cli/gallery-fixtures/types.ts
new file mode 100644
index 000000000..97b7da510
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/types.ts
@@ -0,0 +1,41 @@
+/**
+ * Types for `omp gallery` sample data. See {@link ./index} for the aggregated
+ * fixture registry and the contract each fixture must satisfy.
+ */
+import type { EditMode } from "../../edit";
+
+/** A tool result snapshot, matching the shape `ToolExecutionComponent` consumes. */
+export interface GalleryResult {
+	content: Array<{ type: string; text?: string; data?: string; mimeType?: string }>;
+	details?: unknown;
+	isError?: boolean;
+}
+
+export interface GalleryFixture {
+	/** Display label for the tool header (defaults to the tool name). */
+	label?: string;
+	/** Edit mode for edit-like tools so the streaming preview dispatches correctly. */
+	editMode?: EditMode;
+	/**
+	 * Set for tools whose real `AgentTool` attaches `renderCall`/`renderResult`
+	 * directly on the instance (e.g. `lsp`, `task`). The harness then attaches
+	 * the registry renderer onto the fake tool so the component routes through
+	 * the custom-tool branch — the same path production takes — instead of the
+	 * built-in registry branch. The two branches can diverge, so exercising the
+	 * real one keeps the gallery honest for these tools.
+	 */
+	customRendered?: boolean;
+	/**
+	 * Arguments shown during the streaming state — a partial view of {@link args}
+	 * as if the tool-call JSON were still arriving. May include `__partialJson`
+	 * for renderers (bash, edit) that surface fields before the object closes.
+	 * Defaults to {@link args} when omitted.
+	 */
+	streamingArgs?: unknown;
+	/** Complete arguments shown for the in-progress, success, and error states. */
+	args: unknown;
+	/** Successful result. */
+	result: GalleryResult;
+	/** Failed result. Falls back to a generic error when omitted. */
+	errorResult?: GalleryResult;
+}
diff --git a/packages/coding-agent/src/cli/gallery-fixtures/web.ts b/packages/coding-agent/src/cli/gallery-fixtures/web.ts
new file mode 100644
index 000000000..bec37debe
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-fixtures/web.ts
@@ -0,0 +1,158 @@
+// Gallery fixtures for the web tools (web_search, browser).
+import type { GalleryFixture } from "./types";
+
+export const webFixtures: Record<string, GalleryFixture> = {
+	web_search: {
+		label: "Web Search",
+		// Streaming: query still being typed, no recency/limit yet.
+		streamingArgs: { query: "bun vs node performance" },
+		args: {
+			query: "Bun vs Node.js performance benchmarks 2026",
+			recency: "month",
+			limit: 4,
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						"Bun continues to outperform Node.js on raw HTTP throughput and cold-start",
+						"time thanks to its JavaScriptCore engine and native-Zig runtime, while",
+						"Node.js retains an edge in ecosystem maturity and long-term stability.",
+						"For script-heavy workflows Bun's faster startup is the decisive factor.",
+					].join("\n"),
+				},
+			],
+			details: {
+				response: {
+					provider: "perplexity",
+					model: "sonar-pro",
+					authMode: "api_key",
+					requestId: "req_a1b2c3d4e5f6",
+					answer: [
+						"Bun continues to outperform Node.js on raw HTTP throughput and cold-start",
+						"time thanks to its JavaScriptCore engine and native-Zig runtime, while",
+						"Node.js retains an edge in ecosystem maturity and long-term stability.",
+						"For script-heavy workflows Bun's faster startup is the decisive factor.",
+					].join("\n"),
+					searchQueries: ["bun vs node.js performance benchmarks 2026", "bun http throughput vs node"],
+					sources: [
+						{
+							title: "Bun 1.2 Benchmarks: HTTP, SQLite, and Startup Time",
+							url: "https://bun.sh/blog/bun-v1.2-benchmarks",
+							snippet:
+								"Bun serves roughly 2.5x the requests per second of Node.js on a simple HTTP server and starts in under 10ms.",
+							ageSeconds: 86400 * 12,
+							author: "The Bun Team",
+						},
+						{
+							title: "Node.js vs Bun: A 2026 Performance Deep Dive",
+							url: "https://blog.platformatic.dev/nodejs-vs-bun-2026",
+							snippet:
+								"Across CPU-bound workloads the gap narrows, but Bun's faster module resolution keeps cold starts ahead.",
+							ageSeconds: 86400 * 3,
+							author: "Matteo Collina",
+						},
+						{
+							title: "Real-world API latency: Bun, Deno, and Node compared",
+							url: "https://www.theregister.com/2026/05/18/js_runtime_latency/",
+							snippet:
+								"Under sustained load p99 latencies converge, suggesting runtime choice matters less for steady-state services.",
+							ageSeconds: 86400 * 19,
+						},
+						{
+							title: "Why we migrated our CLI tooling from Node to Bun",
+							url: "https://engineering.example.com/posts/bun-cli-migration",
+							snippet:
+								"Startup dropped from 180ms to 22ms, shaving seconds off every developer command invocation.",
+							ageSeconds: 86400 * 27,
+							author: "Dana Whitfield",
+						},
+					],
+					citations: [
+						{
+							url: "https://bun.sh/blog/bun-v1.2-benchmarks",
+							title: "Bun 1.2 Benchmarks",
+							citedText: "Bun serves roughly 2.5x the requests per second of Node.js",
+						},
+					],
+					usage: {
+						inputTokens: 312,
+						outputTokens: 248,
+						totalTokens: 560,
+						searchRequests: 2,
+					},
+				},
+			},
+		},
+		errorResult: {
+			isError: true,
+			content: [{ type: "text", text: "Web search failed: provider returned HTTP 429 (rate limited)." }],
+			details: {
+				response: {
+					provider: "perplexity",
+					sources: [],
+				},
+				error: "Provider returned HTTP 429 (rate limited). Retry after 30s.",
+			},
+		},
+	},
+
+	browser: {
+		label: "Browser",
+		// Streaming: code body still arriving for a `run` action.
+		streamingArgs: {
+			action: "run",
+			name: "docs",
+			code: "const obs = await tab.observe();\n",
+		},
+		args: {
+			action: "run",
+			name: "docs",
+			code: [
+				"const obs = await tab.observe();",
+				"const heading = obs.elements.find(e => e.role === 'heading');",
+				"display({ url: obs.url, title: obs.title, headings: obs.elements.filter(e => e.role === 'heading').length });",
+				"return heading?.name ?? 'no heading found';",
+			].join("\n"),
+		},
+		result: {
+			content: [
+				{
+					type: "text",
+					text: [
+						'{ url: "https://bun.sh/docs", title: "Bun Documentation", headings: 14 }',
+						'"Get started with Bun"',
+					].join("\n"),
+				},
+			],
+			details: {
+				action: "run",
+				name: "docs",
+				url: "https://bun.sh/docs",
+				browser: "headless",
+				viewport: { width: 1280, height: 800, deviceScaleFactor: 1 },
+				result: '"Get started with Bun"',
+			},
+		},
+		errorResult: {
+			isError: true,
+			content: [
+				{
+					type: "text",
+					text: [
+						"TimeoutError: waiting for selector `aria/Sign in` failed: timeout 30000ms exceeded",
+						"    at Tab.waitFor (browser/tab.ts:212:13)",
+						"    at run (eval:3:7)",
+					].join("\n"),
+				},
+			],
+			details: {
+				action: "run",
+				name: "docs",
+				url: "https://bun.sh/docs",
+				browser: "headless",
+			},
+		},
+	},
+};
diff --git a/packages/coding-agent/src/cli/gallery-screenshot.ts b/packages/coding-agent/src/cli/gallery-screenshot.ts
new file mode 100644
index 000000000..05ba3eafa
--- /dev/null
+++ b/packages/coding-agent/src/cli/gallery-screenshot.ts
@@ -0,0 +1,279 @@
+/**
+ * Render `omp gallery` output to PNG screenshots via VHS.
+ *
+ * ANSI escapes are invisible to anything that can only read raw bytes (e.g.
+ * agents), so `--screenshot` drives the rendered gallery through a real virtual
+ * terminal (VHS + ttyd + ffmpeg) and writes the captured frame to disk. The
+ * gallery is pre-rendered to truecolor ANSI in this process — where the user's
+ * theme and symbol preset are correct — then `cat`'d inside VHS so the captured
+ * pixels match exactly what the live TUI would draw.
+ *
+ * VHS is a hard dependency of this path: if it is not installed we fail loudly
+ * rather than degrade to a lossy fallback.
+ */
+import * as fs from "node:fs";
+import * as os from "node:os";
+import * as path from "node:path";
+import { $which } from "@oh-my-pi/pi-utils";
+import { theme } from "../modes/theme/theme";
+import type { GallerySection } from "./gallery-cli";
+
+/** Nerd Font family so the gallery's icon glyphs (PUA) render instead of tofu. */
+export const DEFAULT_SCREENSHOT_FONT = "JetBrainsMono Nerd Font";
+export const DEFAULT_SCREENSHOT_FONT_SIZE = 18;
+
+/** Inner padding (px) VHS leaves around the terminal grid. */
+const PADDING = 14;
+const LINE_HEIGHT = 1.0;
+/**
+ * Upper-bound cell metrics relative to font size. Real monospace cells are
+ * smaller, so over-provisioning the canvas guarantees the gallery never
+ * soft-wraps (too few columns) or scrolls off the top (too few rows). The slack
+ * shows up only as a modest background margin, which is harmless for review.
+ */
+const CELL_WIDTH_RATIO = 0.65;
+const CELL_HEIGHT_RATIO = 1.5;
+/** Keep each image well under headless-Chromium's tall-canvas limits. */
+const MAX_IMAGE_HEIGHT_PX = 8000;
+
+export interface GalleryScreenshotOptions {
+	/** Gallery render width in columns (matches the ANSI line width). */
+	width: number;
+	/** VHS `FontFamily`. */
+	font?: string;
+	/** VHS `FontSize`. */
+	fontSize?: number;
+	/**
+	 * Output destination. When omitted, PNGs land in a fresh temp directory.
+	 * With multiple images the path is suffixed (`name-01.png`, `name-02.png`).
+	 */
+	out?: string;
+}
+
+/**
+ * Capture the gallery sections as one or more PNGs and return their absolute
+ * paths. Tall galleries are split across images so no single capture exceeds
+ * the terminal-canvas height limit.
+ */
+export async function captureGalleryScreenshots(
+	sections: GallerySection[],
+	options: GalleryScreenshotOptions,
+): Promise<string[]> {
+	const vhs = $which("vhs");
+	if (!vhs) {
+		throw new Error(
+			"`omp gallery --screenshot` requires VHS, which is not installed. " +
+				"Install it (e.g. `brew install vhs`, or see https://github.com/charmbracelet/vhs) and retry.",
+		);
+	}
+
+	const font = options.font ?? DEFAULT_SCREENSHOT_FONT;
+	const fontSize = options.fontSize ?? DEFAULT_SCREENSHOT_FONT_SIZE;
+	const cellHeight = fontSize * Math.max(LINE_HEIGHT, 1) * CELL_HEIGHT_RATIO;
+	const cellWidth = fontSize * CELL_WIDTH_RATIO;
+	const rowBudget = Math.max(40, Math.floor((MAX_IMAGE_HEIGHT_PX - 2 * PADDING) / cellHeight) - 2);
+	const chunks = chunkGallerySections(sections, rowBudget);
+	const themeJson = buildVhsTheme();
+
+	const baseDir = options.out
+		? path.dirname(path.resolve(options.out))
+		: fs.mkdtempSync(path.join(os.tmpdir(), "omp-gallery-"));
+	await fs.promises.mkdir(baseDir, { recursive: true });
+
+	const outPaths: string[] = [];
+	for (let i = 0; i < chunks.length; i++) {
+		if (chunks.length > 1) {
+			process.stderr.write(`Rendering gallery screenshot ${i + 1}/${chunks.length}…\n`);
+		}
+		const outPng = resolveScreenshotOutputPath(options.out, baseDir, i, chunks.length);
+		const lines = chunks[i].flatMap(section => section.lines);
+		await renderChunk({ vhs, lines, outPng, font, fontSize, cellWidth, cellHeight, width: options.width, themeJson });
+		outPaths.push(outPng);
+	}
+	return outPaths;
+}
+
+interface RenderChunkArgs {
+	vhs: string;
+	lines: string[];
+	outPng: string;
+	font: string;
+	fontSize: number;
+	cellWidth: number;
+	cellHeight: number;
+	width: number;
+	themeJson: string;
+}
+
+async function renderChunk(args: RenderChunkArgs): Promise<void> {
+	const rows = args.lines.length;
+	const widthPx = Math.ceil(args.width * args.cellWidth) + 2 * PADDING;
+	const heightPx = Math.ceil((rows + 2) * args.cellHeight) + 2 * PADDING;
+
+	const dir = path.dirname(args.outPng);
+	const stem = path.basename(args.outPng, path.extname(args.outPng));
+	const ansiPath = path.join(dir, `.${stem}.ansi`);
+	const tapePath = path.join(dir, `.${stem}.tape`);
+	const gifPath = path.join(dir, `.${stem}.gif`);
+
+	// CRLF so each gallery line is its own terminal row regardless of how the
+	// captured shell handles bare LF.
+	await Bun.write(ansiPath, `${args.lines.join("\r\n")}\r\n`);
+	await Bun.write(
+		tapePath,
+		buildTape({
+			gifPath,
+			outPng: args.outPng,
+			ansiPath,
+			widthPx,
+			heightPx,
+			font: args.font,
+			fontSize: args.fontSize,
+			themeJson: args.themeJson,
+		}),
+	);
+
+	try {
+		const result = await Bun.$`${args.vhs} ${tapePath}`.quiet().nothrow();
+		if (result.exitCode !== 0 || !(await Bun.file(args.outPng).exists())) {
+			const detail = result.stderr.toString().trim() || result.stdout.toString().trim();
+			throw new Error(`VHS failed to render the gallery screenshot${detail ? `: ${detail.slice(-600)}` : ""}`);
+		}
+	} finally {
+		await Promise.all([
+			fs.promises.rm(ansiPath, { force: true }),
+			fs.promises.rm(tapePath, { force: true }),
+			fs.promises.rm(gifPath, { force: true }),
+		]);
+	}
+}
+
+interface TapeArgs {
+	gifPath: string;
+	outPng: string;
+	ansiPath: string;
+	widthPx: number;
+	heightPx: number;
+	font: string;
+	fontSize: number;
+	themeJson: string;
+}
+
+function buildTape(args: TapeArgs): string {
+	// `Output` (a throwaway GIF) is mandatory for VHS to record; the screenshot
+	// is captured from the final visible frame. Setup is hidden so the typed
+	// `cat` command and shell prompt never appear in the capture, and a trailing
+	// `sleep` keeps the shell from drawing a fresh prompt under the output.
+	const shellCommand = `clear; cat ${shellSingleQuote(args.ansiPath)}; sleep 120`;
+	return `${[
+		`Output ${JSON.stringify(args.gifPath)}`,
+		`Set Width ${args.widthPx}`,
+		`Set Height ${args.heightPx}`,
+		`Set FontFamily ${JSON.stringify(args.font)}`,
+		`Set FontSize ${args.fontSize}`,
+		`Set Padding ${PADDING}`,
+		`Set LineHeight ${LINE_HEIGHT}`,
+		`Set Theme ${args.themeJson}`,
+		"Hide",
+		`Type ${JSON.stringify(shellCommand)}`,
+		"Enter",
+		"Sleep 1.2s",
+		"Show",
+		"Sleep 400ms",
+		`Screenshot ${JSON.stringify(args.outPng)}`,
+	].join("\n")}\n`;
+}
+
+/**
+ * Build the VHS terminal theme. Only background/foreground/cursor matter: the
+ * gallery emits truecolor (`38;2`/`48;2`) escapes, so the 16-color palette is
+ * never consulted — it is filler to satisfy VHS's theme schema.
+ */
+function buildVhsTheme(): string {
+	const background = parseAnsiRgb(theme.getBgAnsi("statusLineBg")) ?? (theme.isLight ? "#ffffff" : "#1a1a1a");
+	const foreground = theme.isLight ? "#1a1a1a" : "#d4d4d4";
+	const selection = theme.isLight ? "#c8d6ff" : "#404862";
+	return JSON.stringify({
+		name: "omp-gallery",
+		background,
+		foreground,
+		cursor: foreground,
+		selection,
+		black: "#000000",
+		red: "#ff5555",
+		green: "#50fa7b",
+		yellow: "#f1fa8c",
+		blue: "#6272ff",
+		magenta: "#ff79c6",
+		cyan: "#8be9fd",
+		white: "#bfbfbf",
+		brightBlack: "#4d4d4d",
+		brightRed: "#ff6e6e",
+		brightGreen: "#69ff94",
+		brightYellow: "#ffffa5",
+		brightBlue: "#8aa0ff",
+		brightMagenta: "#ff92df",
+		brightCyan: "#a4ffff",
+		brightWhite: "#ffffff",
+	});
+}
+
+/** Extract `#rrggbb` from a truecolor SGR escape (`…38;2;r;g;b…` / `…48;2;…`). */
+function parseAnsiRgb(ansi: string): string | undefined {
+	const match = /[34]8;2;(\d+);(\d+);(\d+)/.exec(ansi);
+	if (!match) return undefined;
+	const hex = (value: string) => Number(value).toString(16).padStart(2, "0");
+	return `#${hex(match[1])}${hex(match[2])}${hex(match[3])}`;
+}
+
+/** POSIX single-quote a path for embedding in the VHS shell command. */
+function shellSingleQuote(value: string): string {
+	return `'${value.replace(/'/g, `'\\''`)}'`;
+}
+
+/**
+ * Resolve a chunk's PNG path. A single image keeps the bare name (or the exact
+ * `out`); multiple images gain a zero-padded `-NN` suffix so they sort and never
+ * collide.
+ */
+export function resolveScreenshotOutputPath(
+	out: string | undefined,
+	baseDir: string,
+	index: number,
+	total: number,
+): string {
+	if (total === 1) {
+		return out ? path.resolve(out) : path.join(baseDir, "gallery.png");
+	}
+	const suffix = String(index + 1).padStart(2, "0");
+	if (out) {
+		const resolved = path.resolve(out);
+		const ext = path.extname(resolved) || ".png";
+		const stem = path.basename(resolved, ext);
+		return path.join(path.dirname(resolved), `${stem}-${suffix}${ext}`);
+	}
+	return path.join(baseDir, `gallery-${suffix}.png`);
+}
+
+/**
+ * Group whole tool sections into chunks that stay under `rowBudget` rows. A
+ * single section larger than the budget gets its own (taller) image rather than
+ * being split mid-renderer.
+ */
+export function chunkGallerySections(sections: GallerySection[], rowBudget: number): GallerySection[][] {
+	const chunks: GallerySection[][] = [];
+	let current: GallerySection[] = [];
+	let currentRows = 0;
+	for (const section of sections) {
+		const rows = section.lines.length;
+		if (current.length > 0 && currentRows + rows > rowBudget) {
+			chunks.push(current);
+			current = [];
+			currentRows = 0;
+		}
+		current.push(section);
+		currentRows += rows;
+	}
+	if (current.length > 0) chunks.push(current);
+	return chunks.length > 0 ? chunks : [[]];
+}
diff --git a/packages/coding-agent/src/commands/gallery.ts b/packages/coding-agent/src/commands/gallery.ts
new file mode 100644
index 000000000..d0b878fac
--- /dev/null
+++ b/packages/coding-agent/src/commands/gallery.ts
@@ -0,0 +1,52 @@
+/**
+ * Render every built-in tool's renderer across its lifecycle states.
+ */
+import { Command, Flags } from "@oh-my-pi/pi-utils/cli";
+import { GALLERY_STATES, type GalleryState, runGalleryCommand } from "../cli/gallery-cli";
+
+export default class Gallery extends Command {
+	static description = "Preview tool renderers across streaming, in-progress, success, and failure states";
+
+	static flags = {
+		tool: Flags.string({ char: "t", description: "Render a single tool by name" }),
+		state: Flags.string({
+			char: "s",
+			description: "Render only the given lifecycle state(s)",
+			options: [...GALLERY_STATES],
+			multiple: true,
+		}),
+		width: Flags.integer({ char: "w", description: "Render width in columns" }),
+		expanded: Flags.boolean({
+			char: "e",
+			description: "Render the expanded variant of each renderer",
+			default: false,
+		}),
+		plain: Flags.boolean({ description: "Strip ANSI styling from the output", default: false }),
+		screenshot: Flags.boolean({
+			description:
+				"Capture the rendered output as PNG screenshot(s) via VHS instead of printing ANSI (requires vhs)",
+			default: false,
+		}),
+		out: Flags.string({
+			char: "o",
+			description: "Screenshot output path (with --screenshot); suffixed per image when split across multiple",
+		}),
+		font: Flags.string({ description: "Screenshot font family (default: JetBrainsMono Nerd Font)" }),
+		"font-size": Flags.integer({ description: "Screenshot font size in points (default: 18)" }),
+	};
+
+	async run(): Promise<void> {
+		const { flags } = await this.parse(Gallery);
+		await runGalleryCommand({
+			tool: flags.tool,
+			states: flags.state as GalleryState[] | undefined,
+			width: flags.width,
+			expanded: flags.expanded,
+			plain: flags.plain,
+			screenshot: flags.screenshot,
+			out: flags.out,
+			font: flags.font,
+			fontSize: flags["font-size"],
+		});
+	}
+}
diff --git a/packages/coding-agent/src/commands/launch.ts b/packages/coding-agent/src/commands/launch.ts
index a752cdf82..d26dc543a 100644
--- a/packages/coding-agent/src/commands/launch.ts
+++ b/packages/coding-agent/src/commands/launch.ts
@@ -2,7 +2,7 @@
  * Root command for the coding agent CLI.
  */
 
-import { THINKING_EFFORTS } from "@oh-my-pi/pi-ai";
+import { THINKING_EFFORTS } from "@oh-my-pi/pi-ai/effort";
 import { APP_NAME } from "@oh-my-pi/pi-utils";
 import { Args, Command, Flags } from "@oh-my-pi/pi-utils/cli";
 import { parseArgs } from "../cli/args";
diff --git a/packages/coding-agent/src/commit/analysis/conventional.ts b/packages/coding-agent/src/commit/analysis/conventional.ts
index 5b52bb5c4..4f8c4ee75 100644
--- a/packages/coding-agent/src/commit/analysis/conventional.ts
+++ b/packages/coding-agent/src/commit/analysis/conventional.ts
@@ -1,5 +1,5 @@
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, Model } from "@oh-my-pi/pi-ai";
 import { completeSimple } from "@oh-my-pi/pi-ai";
 import { prompt } from "@oh-my-pi/pi-utils";
 import analysisSystemPrompt from "../../commit/prompts/analysis-system.md" with { type: "text" };
@@ -14,7 +14,7 @@ const ConventionalAnalysisTool = createConventionalAnalysisTool(
 
 export interface ConventionalAnalysisInput {
 	model: Model<Api>;
-	apiKey: string;
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 	contextFiles?: Array<{ path: string; content: string }>;
 	userContext?: string;
diff --git a/packages/coding-agent/src/commit/analysis/summary.ts b/packages/coding-agent/src/commit/analysis/summary.ts
index 5dd23ff1f..80bc3f096 100644
--- a/packages/coding-agent/src/commit/analysis/summary.ts
+++ b/packages/coding-agent/src/commit/analysis/summary.ts
@@ -1,5 +1,5 @@
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, AssistantMessage, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, AssistantMessage, Model } from "@oh-my-pi/pi-ai";
 import { completeSimple, validateToolCall } from "@oh-my-pi/pi-ai";
 import { prompt } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
@@ -19,7 +19,7 @@ const SummaryTool = {
 
 export interface SummaryInput {
 	model: Model<Api>;
-	apiKey: string;
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 	commitType: string;
 	scope: string | null;
diff --git a/packages/coding-agent/src/commit/changelog/generate.ts b/packages/coding-agent/src/commit/changelog/generate.ts
index bc43cd1e7..c59e17b90 100644
--- a/packages/coding-agent/src/commit/changelog/generate.ts
+++ b/packages/coding-agent/src/commit/changelog/generate.ts
@@ -1,5 +1,5 @@
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, AssistantMessage, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, AssistantMessage, Model } from "@oh-my-pi/pi-ai";
 import { completeSimple, validateToolCall } from "@oh-my-pi/pi-ai";
 import { prompt } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
@@ -25,7 +25,7 @@ export const changelogTool = {
 
 export interface ChangelogPromptInput {
 	model: Model<Api>;
-	apiKey: string;
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 	changelogPath: string;
 	isPackageChangelog: boolean;
diff --git a/packages/coding-agent/src/commit/changelog/index.ts b/packages/coding-agent/src/commit/changelog/index.ts
index f80f79401..5260db181 100644
--- a/packages/coding-agent/src/commit/changelog/index.ts
+++ b/packages/coding-agent/src/commit/changelog/index.ts
@@ -1,6 +1,6 @@
 import * as path from "node:path";
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, Model } from "@oh-my-pi/pi-ai";
 import { logger } from "@oh-my-pi/pi-utils";
 import { CHANGELOG_CATEGORIES } from "../../commit/types";
 import * as git from "../../utils/git";
@@ -15,7 +15,7 @@ const DEFAULT_MAX_DIFF_CHARS = 120_000;
 export interface ChangelogFlowInput {
 	cwd: string;
 	model: Model<Api>;
-	apiKey: string;
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 	stagedFiles: string[];
 	dryRun: boolean;
diff --git a/packages/coding-agent/src/commit/map-reduce/index.ts b/packages/coding-agent/src/commit/map-reduce/index.ts
index c2e9f3332..a11b417c5 100644
--- a/packages/coding-agent/src/commit/map-reduce/index.ts
+++ b/packages/coding-agent/src/commit/map-reduce/index.ts
@@ -1,5 +1,5 @@
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, Model } from "@oh-my-pi/pi-ai";
 import { $env } from "@oh-my-pi/pi-utils";
 import { parseFileDiffs } from "../../commit/git/diff";
 import type { ConventionalAnalysis } from "../../commit/types";
@@ -21,10 +21,10 @@ export interface MapReduceSettings {
 
 export interface MapReduceInput {
 	model: Model<Api>;
-	apiKey: string;
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 	smolModel: Model<Api>;
-	smolApiKey: string;
+	smolApiKey: ApiKey;
 	smolThinkingLevel?: ThinkingLevel;
 	diff: string;
 	stat: string;
diff --git a/packages/coding-agent/src/commit/map-reduce/map-phase.ts b/packages/coding-agent/src/commit/map-reduce/map-phase.ts
index 3f376cb87..a263bfead 100644
--- a/packages/coding-agent/src/commit/map-reduce/map-phase.ts
+++ b/packages/coding-agent/src/commit/map-reduce/map-phase.ts
@@ -1,5 +1,5 @@
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, AssistantMessage, Message, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, AssistantMessage, Message, Model } from "@oh-my-pi/pi-ai";
 import { completeSimple } from "@oh-my-pi/pi-ai";
 import { prompt } from "@oh-my-pi/pi-utils";
 import fileObserverSystemPrompt from "../../commit/prompts/file-observer-system.md" with { type: "text" };
@@ -18,7 +18,7 @@ const RETRY_BACKOFF_MS = 1000;
 
 export interface MapPhaseInput {
 	model: Model<Api>;
-	apiKey: string;
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 	files: FileDiff[];
 	config?: {
diff --git a/packages/coding-agent/src/commit/map-reduce/reduce-phase.ts b/packages/coding-agent/src/commit/map-reduce/reduce-phase.ts
index 1ea5278e9..b0caccd0d 100644
--- a/packages/coding-agent/src/commit/map-reduce/reduce-phase.ts
+++ b/packages/coding-agent/src/commit/map-reduce/reduce-phase.ts
@@ -1,5 +1,5 @@
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, Model } from "@oh-my-pi/pi-ai";
 import { completeSimple } from "@oh-my-pi/pi-ai";
 import { prompt } from "@oh-my-pi/pi-utils";
 import reduceSystemPrompt from "../../commit/prompts/reduce-system.md" with { type: "text" };
@@ -12,7 +12,7 @@ const ReduceTool = createConventionalAnalysisTool("Synthesize file observations
 
 export interface ReducePhaseInput {
 	model: Model<Api>;
-	apiKey: string;
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 	observations: FileObservation[];
 	stat: string;
diff --git a/packages/coding-agent/src/commit/model-selection.ts b/packages/coding-agent/src/commit/model-selection.ts
index 3045dfdf6..135bb6be7 100644
--- a/packages/coding-agent/src/commit/model-selection.ts
+++ b/packages/coding-agent/src/commit/model-selection.ts
@@ -1,5 +1,6 @@
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, Model } from "@oh-my-pi/pi-ai";
+import type { ApiKeyResolverRegistry } from "../config/api-key-resolver";
 import { MODEL_ROLE_IDS } from "../config/model-registry";
 import {
 	type ModelLookupRegistry,
@@ -12,13 +13,19 @@ import MODEL_PRIO from "../priority.json" with { type: "json" };
 
 export interface ResolvedCommitModel {
 	model: Model<Api>;
-	apiKey: string;
+	/**
+	 * Resolver for the model's bearer: re-resolves on 401 / usage-limit so the
+	 * whole commit pipeline (analysis, map/reduce, changelog) inherits the
+	 * central force-refresh + account-rotation policy.
+	 */
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 }
 
-type CommitModelRegistry = ModelLookupRegistry & {
-	getApiKey: (model: Model<Api>) => Promise<string | undefined>;
-};
+type CommitModelRegistry = ModelLookupRegistry &
+	ApiKeyResolverRegistry & {
+		getApiKey: (model: Model<Api>) => Promise<string | undefined>;
+	};
 
 export async function resolvePrimaryModel(
 	override: string | undefined,
@@ -38,20 +45,32 @@ export async function resolvePrimaryModel(
 	if (!apiKey) {
 		throw new Error(`No API key available for model ${model.provider}/${model.id}`);
 	}
-	return { model, apiKey, thinkingLevel: resolved?.thinkingLevel };
+	return {
+		model,
+		apiKey: modelRegistry.resolver(model.provider, { baseUrl: model.baseUrl }),
+		thinkingLevel: resolved?.thinkingLevel,
+	};
 }
 
 export async function resolveSmolModel(
 	settings: Settings,
 	modelRegistry: CommitModelRegistry,
 	fallbackModel: Model<Api>,
-	fallbackApiKey: string,
+	fallbackApiKey: ApiKey,
 ): Promise<ResolvedCommitModel> {
 	const available = modelRegistry.getAvailable();
 	const resolvedSmol = resolveRoleSelection(["smol"], settings, available, modelRegistry);
 	if (resolvedSmol?.model) {
 		const apiKey = await modelRegistry.getApiKey(resolvedSmol.model);
-		if (apiKey) return { model: resolvedSmol.model, apiKey, thinkingLevel: resolvedSmol.thinkingLevel };
+		if (apiKey) {
+			return {
+				model: resolvedSmol.model,
+				apiKey: modelRegistry.resolver(resolvedSmol.model.provider, {
+					baseUrl: resolvedSmol.model.baseUrl,
+				}),
+				thinkingLevel: resolvedSmol.thinkingLevel,
+			};
+		}
 	}
 
 	const matchPreferences = { usageOrder: settings.getStorage()?.getModelUsageOrder() };
@@ -59,7 +78,12 @@ export async function resolveSmolModel(
 		const candidate = parseModelPattern(pattern, available, matchPreferences, { modelRegistry }).model;
 		if (!candidate) continue;
 		const apiKey = await modelRegistry.getApiKey(candidate);
-		if (apiKey) return { model: candidate, apiKey };
+		if (apiKey) {
+			return {
+				model: candidate,
+				apiKey: modelRegistry.resolver(candidate.provider, { baseUrl: candidate.baseUrl }),
+			};
+		}
 	}
 
 	return { model: fallbackModel, apiKey: fallbackApiKey };
diff --git a/packages/coding-agent/src/commit/pipeline.ts b/packages/coding-agent/src/commit/pipeline.ts
index bc53e54fe..86d626d9c 100644
--- a/packages/coding-agent/src/commit/pipeline.ts
+++ b/packages/coding-agent/src/commit/pipeline.ts
@@ -1,6 +1,6 @@
 import * as path from "node:path";
 import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
-import type { Api, Model } from "@oh-my-pi/pi-ai";
+import type { Api, ApiKey, Model } from "@oh-my-pi/pi-ai";
 import { getProjectDir, logger, prompt } from "@oh-my-pi/pi-utils";
 import { ModelRegistry } from "../config/model-registry";
 import { Settings } from "../config/settings";
@@ -145,10 +145,10 @@ async function generateAnalysis(input: {
 	contextFiles: Array<{ path: string; content: string }>;
 	userContext?: string;
 	primaryModel: Model<Api>;
-	primaryApiKey: string;
+	primaryApiKey: ApiKey;
 	primaryThinkingLevel?: ThinkingLevel;
 	smolModel: Model<Api>;
-	smolApiKey: string;
+	smolApiKey: ApiKey;
 	smolThinkingLevel?: ThinkingLevel;
 	commitSettings: {
 		mapReduceEnabled: boolean;
@@ -206,7 +206,7 @@ async function generateSummaryWithRetry(input: {
 	analysis: ConventionalAnalysis;
 	stat: string;
 	model: Model<Api>;
-	apiKey: string;
+	apiKey: ApiKey;
 	thinkingLevel?: ThinkingLevel;
 	userContext?: string;
 }): Promise<{ summary: string }> {
diff --git a/packages/coding-agent/src/config/api-key-resolver.ts b/packages/coding-agent/src/config/api-key-resolver.ts
new file mode 100644
index 000000000..5204c5496
--- /dev/null
+++ b/packages/coding-agent/src/config/api-key-resolver.ts
@@ -0,0 +1,58 @@
+import type { ApiKeyResolver, AuthStorage } from "@oh-my-pi/pi-ai";
+
+export interface ApiKeyResolverOptions {
+	/** Session id for credential stickiness; read at resolve time by the caller. */
+	sessionId?: string;
+	/** Provider base URL hint forwarded to the auth-storage cascade. */
+	baseUrl?: string;
+}
+
+/**
+ * Minimal slice of `ModelRegistry` the resolver needs. Typed structurally so
+ * narrower registry shells (e.g. the commit pipeline's `CommitModelRegistry`)
+ * can build resolvers without depending on the full class.
+ */
+export interface ApiKeyResolverRegistry {
+	getApiKeyForProvider(
+		provider: string,
+		sessionId?: string,
+		options?: { baseUrl?: string; forceRefresh?: boolean; signal?: AbortSignal },
+	): Promise<string | undefined>;
+	authStorage: Pick<AuthStorage, "rotateSessionCredential">;
+	/**
+	 * Build an {@link ApiKeyResolver} implementing the central a/b/c auth-retry
+	 * policy: initial → resolve; step (b) → force-refresh same account; step (c)
+	 * → rotate to a sibling credential, then re-resolve.
+	 *
+	 * The resolver is stateless (safe to reuse across requests). Callers that
+	 * need the initial key for a guard can call `resolveApiKeyOnce(resolver)`.
+	 */
+	resolver(provider: string, options?: ApiKeyResolverOptions): ApiKeyResolver;
+}
+
+/**
+ * Default implementation of {@link ApiKeyResolverRegistry.resolver}.
+ * Also usable standalone for structural registries that don't carry the method.
+ */
+export function createApiKeyResolver(
+	registry: Pick<ApiKeyResolverRegistry, "getApiKeyForProvider" | "authStorage">,
+	provider: string,
+	options: ApiKeyResolverOptions = {},
+): ApiKeyResolver {
+	const { sessionId, baseUrl } = options;
+	return async ({ lastChance, error, signal }) => {
+		if (error === undefined) {
+			return registry.getApiKeyForProvider(provider, sessionId, { baseUrl });
+		}
+		if (lastChance) {
+			// Account constraint (401 / usage / account-rate-limit): rotate to a
+			// sibling credential. We do NOT honor any retry-after here — if a
+			// sibling exists we switch immediately; the precise no-sibling backoff
+			// is owned by `markUsageLimitReached` (default + server usage-report
+			// reset) and the outer whole-turn retry layer.
+			await registry.authStorage.rotateSessionCredential(provider, sessionId, { error, signal });
+			return registry.getApiKeyForProvider(provider, sessionId, { baseUrl });
+		}
+		return registry.getApiKeyForProvider(provider, sessionId, { baseUrl, forceRefresh: true, signal });
+	};
+}
diff --git a/packages/coding-agent/src/config/keybindings.ts b/packages/coding-agent/src/config/keybindings.ts
index ace7f7860..b652586b3 100644
--- a/packages/coding-agent/src/config/keybindings.ts
+++ b/packages/coding-agent/src/config/keybindings.ts
@@ -21,6 +21,7 @@ interface AppKeybindings {
 	"app.clear": true;
 	"app.exit": true;
 	"app.suspend": true;
+	"app.display.reset": true;
 	"app.thinking.cycle": true;
 	"app.thinking.toggle": true;
 	"app.model.cycleForward": true;
@@ -86,6 +87,10 @@ export const KEYBINDINGS = {
 		defaultKeys: "ctrl+z",
 		description: "Suspend application",
 	},
+	"app.display.reset": {
+		defaultKeys: "ctrl+l",
+		description: "Reset terminal display",
+	},
 	"app.thinking.cycle": {
 		defaultKeys: "shift+tab",
 		description: "Cycle thinking level",
@@ -103,7 +108,7 @@ export const KEYBINDINGS = {
 		description: "Cycle to previous model",
 	},
 	"app.model.select": {
-		defaultKeys: "ctrl+l",
+		defaultKeys: "alt+m",
 		description: "Select model",
 	},
 	"app.model.selectTemporary": {
@@ -119,7 +124,10 @@ export const KEYBINDINGS = {
 		description: "Open external editor",
 	},
 	"app.message.followUp": {
-		defaultKeys: "ctrl+enter",
+		// Ctrl+Enter is preserved for terminals that deliver it (Kitty/iTerm2/WezTerm/Ghostty),
+		// but Windows Terminal does not emit a distinct event for Ctrl+Enter — Ctrl+Q is listed
+		// first so the default binding works there without remapping (#1903).
+		defaultKeys: ["ctrl+q", "ctrl+enter"],
 		description: "Send follow-up message",
 	},
 	"app.message.dequeue": {
@@ -213,6 +221,7 @@ const KEYBINDING_NAME_MIGRATIONS = {
 	clear: "app.clear",
 	exit: "app.exit",
 	suspend: "app.suspend",
+	displayReset: "app.display.reset",
 	cycleThinkingLevel: "app.thinking.cycle",
 	cycleModelForward: "app.model.cycleForward",
 	cycleModelBackward: "app.model.cycleBackward",
@@ -439,16 +448,50 @@ function migrateKeybindingsConfigFile(agentDir: string): void {
 	loadKeybindingsConfig(readPath, writeBackPath);
 }
 
+const FOLLOW_UP_KEYBINDING: AppKeybinding = "app.message.followUp";
+const WINDOWS_FOLLOW_UP_FALLBACK_KEY: KeyId = "ctrl+q";
+function keyListIncludes(keys: KeyId | KeyId[] | undefined, target: KeyId): boolean {
+	if (keys === undefined) return false;
+	const keyList = Array.isArray(keys) ? keys : [keys];
+	for (const key of keyList) {
+		if (key.toLowerCase() === target) return true;
+	}
+	return false;
+}
+
+function userBindingClaimsKey(config: KeybindingsConfig, target: KeyId, except: Keybinding): boolean {
+	for (const [keybinding, keys] of Object.entries(config)) {
+		if (!(keybinding in KEYBINDINGS)) continue;
+		if (keybinding === except) continue;
+		if (keyListIncludes(keys, target)) return true;
+	}
+	return false;
+}
+
+function removeKey(keys: KeyId[], target: KeyId): KeyId[] {
+	return keys.filter(key => key !== target);
+}
+
+function keyConfigValue(keys: KeyId[]): KeyId | KeyId[] {
+	if (keys.length === 1) {
+		const key = keys[0];
+		if (key !== undefined) return key;
+	}
+	return [...keys];
+}
+
 /**
  * Manages all keybindings (app + TUI).
  * Extends the TUI KeybindingsManager with app-specific functionality.
  */
 export class KeybindingsManager extends TuiKeybindingsManager {
 	#configPath: string | undefined;
+	#userBindings: KeybindingsConfig;
 
 	constructor(userBindings: KeybindingsConfig = {}, configPath?: string) {
 		super(KEYBINDINGS, userBindings);
 		this.#configPath = configPath;
+		this.#userBindings = userBindings;
 	}
 
 	/**
@@ -480,6 +523,29 @@ export class KeybindingsManager extends TuiKeybindingsManager {
 		this.setUserBindings(config);
 	}
 
+	setUserBindings(userBindings: KeybindingsConfig): void {
+		this.#userBindings = userBindings;
+		super.setUserBindings(userBindings);
+	}
+
+	getKeys(keybinding: Keybinding): KeyId[] {
+		const keys = super.getKeys(keybinding);
+		if (keybinding === FOLLOW_UP_KEYBINDING) {
+			if (this.#userBindings[FOLLOW_UP_KEYBINDING] !== undefined) return keys;
+			if (!userBindingClaimsKey(this.#userBindings, WINDOWS_FOLLOW_UP_FALLBACK_KEY, FOLLOW_UP_KEYBINDING)) {
+				return keys;
+			}
+			return removeKey(keys, WINDOWS_FOLLOW_UP_FALLBACK_KEY);
+		}
+		return keys;
+	}
+
+	getResolvedBindings(): KeybindingsConfig {
+		const resolved = super.getResolvedBindings();
+		resolved[FOLLOW_UP_KEYBINDING] = keyConfigValue(this.getKeys(FOLLOW_UP_KEYBINDING));
+		return resolved;
+	}
+
 	/**
 	 * Get the effective resolved bindings (defaults + user overrides).
 	 */
diff --git a/packages/coding-agent/src/config/model-equivalence.ts b/packages/coding-agent/src/config/model-equivalence.ts
index 28ae3744a..75fedfc2b 100644
--- a/packages/coding-agent/src/config/model-equivalence.ts
+++ b/packages/coding-agent/src/config/model-equivalence.ts
@@ -58,7 +58,7 @@ const EMPTY_COMPILED_EQUIVALENCE: CompiledEquivalenceConfig = {
 };
 const kModelResolutionCache = Symbol("model-equivalence.resolutionCache");
 interface CompiledEquivalenceConfigWithCache extends CompiledEquivalenceConfig {
-	[kModelResolutionCache]?: WeakMap<Model<Api>, ResolvedCanonicalModel>;
+	[kModelResolutionCache]?: Map<string, ResolvedCanonicalModel>;
 }
 const FAMILY_EXTRACTION_PATTERNS = [
 	/(?:^|[/:._-])((?:claude|gemini|gpt|grok|glm|qwen|minimax|kimi|deepseek|llama|gemma|nova|mistral|ministral|pixtral|codestral|devstral|magistral|ernie|doubao|seed|aion|olmo|molmo|nemotron|palmyra|command|codex|coder|o[1345])[-a-z0-9.]+)(?::|$)/i,
@@ -128,10 +128,18 @@ function normalizeCanonicalIdKey(canonicalId: string): string {
 	return canonicalId.trim().toLowerCase();
 }
 
+function getCanonicalSuffixAliasKey(candidate: string): string {
+	return PENALTY_HAS_UPPERCASE.test(candidate) ? normalizeCanonicalIdKey(candidate) : candidate;
+}
+
 export function formatCanonicalVariantSelector(model: Model<Api>): string {
 	return `${model.provider}/${model.id}`;
 }
 
+function getModelResolutionCacheKey(model: Model<Api>): string {
+	return `${model.provider}\0${model.id}`;
+}
+
 function buildOverrideMap(overrides: Record<string, string> | undefined): Map<string, string> {
 	const result = new Map<string, string>();
 	if (!overrides) {
@@ -159,13 +167,24 @@ function buildExclusionSet(exclusions: readonly string[] | undefined): Set<strin
 	return result;
 }
 
+const compiledEquivalenceCache = new WeakMap<ModelEquivalenceConfig, CompiledEquivalenceConfig>();
 function compileEquivalenceConfig(config: ModelEquivalenceConfig | undefined): CompiledEquivalenceConfig {
+	if (config) {
+		const cached = compiledEquivalenceCache.get(config);
+		if (cached) {
+			return cached;
+		}
+	}
 	const overrides = buildOverrideMap(config?.overrides);
 	const exclude = buildExclusionSet(config?.exclude);
 	if (overrides.size === 0 && exclude.size === 0) {
 		return EMPTY_COMPILED_EQUIVALENCE;
 	}
-	return { overrides, exclude };
+	const compiled: CompiledEquivalenceConfig = { overrides, exclude };
+	if (config) {
+		compiledEquivalenceCache.set(config, compiled);
+	}
+	return compiled;
 }
 
 function addCanonicalCandidate(candidates: Set<string>, candidate: string): void {
@@ -277,7 +296,7 @@ function expandCompactSeriesMinorVersions(candidate: string): string[] {
 // safely return the same instance. Cap keeps memory bounded under adversarial
 // model-id churn.
 const QUALIFIED_NAMESPACE_SUFFIX_CACHE = new Map<string, string[]>();
-const QUALIFIED_NAMESPACE_SUFFIX_CACHE_CAP = 256;
+const QUALIFIED_NAMESPACE_SUFFIX_CACHE_CAP = 4096;
 function getQualifiedNamespaceSuffixes(candidate: string): string[] {
 	const cached = QUALIFIED_NAMESPACE_SUFFIX_CACHE.get(candidate);
 	if (cached !== undefined) {
@@ -670,7 +689,7 @@ function expandHeavyCanonicalCandidates(normalized: string, queue: string[]): vo
 // is unused — kept for signature stability). The returned array is consumed via
 // `.filter` at every callsite, so sharing the cached instance is safe.
 const HEURISTIC_CANDIDATES_CACHE = new Map<string, string[]>();
-const HEURISTIC_CANDIDATES_CACHE_CAP = 256;
+const HEURISTIC_CANDIDATES_CACHE_CAP = 4096;
 function getHeuristicCanonicalCandidates(modelId: string, _officialIds?: ReadonlySet<string>): string[] {
 	const cached = HEURISTIC_CANDIDATES_CACHE.get(modelId);
 	if (cached !== undefined) {
@@ -728,10 +747,10 @@ function getPreferredFallbackCanonicalCandidate(modelId: string, candidates: rea
 
 function resolveCanonicalIdForModel(
 	model: Model<Api>,
+	selector: string,
 	equivalence: CompiledEquivalenceConfig,
 	referenceData: CanonicalReferenceData,
 ): ResolvedCanonicalModel {
-	const selector = formatCanonicalVariantSelector(model);
 	const normalizedSelector = normalizeSelectorKey(selector);
 
 	if (equivalence.overrides.has(normalizedSelector)) {
@@ -753,9 +772,12 @@ function resolveCanonicalIdForModel(
 	}
 
 	const heuristicCandidates = getHeuristicCanonicalCandidates(model.id, referenceData.officialIds);
-	const officialMatches = new Set(heuristicCandidates.filter(candidate => referenceData.officialIds.has(candidate)));
+	const officialMatches = new Set<string>();
 	for (const candidate of heuristicCandidates) {
-		const aliased = referenceData.suffixAliases.get(normalizeCanonicalIdKey(candidate));
+		if (referenceData.officialIds.has(candidate)) {
+			officialMatches.add(candidate);
+		}
+		const aliased = referenceData.suffixAliases.get(getCanonicalSuffixAliasKey(candidate));
 		if (aliased) {
 			officialMatches.add(aliased);
 		}
@@ -814,17 +836,18 @@ export function buildCanonicalModelIndex(
 	const compiledWithCache = compiledEquivalence as CompiledEquivalenceConfigWithCache;
 	let modelCache = compiledWithCache[kModelResolutionCache];
 	if (!modelCache) {
-		modelCache = new WeakMap<Model<Api>, ResolvedCanonicalModel>();
+		modelCache = new Map<string, ResolvedCanonicalModel>();
 		compiledWithCache[kModelResolutionCache] = modelCache;
 	}
 
 	for (const model of models) {
-		let canonical = modelCache.get(model);
-		if (!canonical) {
-			canonical = resolveCanonicalIdForModel(model, compiledEquivalence, referenceData);
-			modelCache.set(model, canonical);
-		}
 		const selector = formatCanonicalVariantSelector(model);
+		const cacheKey = getModelResolutionCacheKey(model);
+		let canonical = modelCache.get(cacheKey);
+		if (!canonical) {
+			canonical = resolveCanonicalIdForModel(model, selector, compiledEquivalence, referenceData);
+			modelCache.set(cacheKey, canonical);
+		}
 		const variant: CanonicalModelVariant = {
 			canonicalId: canonical.id,
 			selector,
diff --git a/packages/coding-agent/src/config/model-id-affixes.ts b/packages/coding-agent/src/config/model-id-affixes.ts
index b4aff136f..7cec31892 100644
--- a/packages/coding-agent/src/config/model-id-affixes.ts
+++ b/packages/coding-agent/src/config/model-id-affixes.ts
@@ -4,34 +4,49 @@ const MODEL_ID_SEGMENT_PATTERN = /[a-z0-9.:-]+/g;
 const MODEL_FAMILY_PREFIX_PATTERN =
 	/^(claude|gemini|gpt|grok|glm|qwen|deepseek|kimi|mimo|doubao|ernie|gpt-oss|gemma|minimax|step|command|jamba|llama|o[1345])/i;
 
-function hasDigit(value: string): boolean {
-	return /\d/.test(value);
+function normalizeModelIdWhitespace(value: string): string {
+	return value.trim().replace(/\s+/g, " ");
 }
 
+/** Ordering for model-like segments: longest first, ties broken lexicographically. */
 function compareSegmentPreference(left: string, right: string): number {
-	if (left.length !== right.length) {
-		return right.length - left.length;
-	}
-	return left.localeCompare(right);
+	return left.length !== right.length ? right.length - left.length : left.localeCompare(right);
 }
 
 export function getModelLikeIdSegments(modelId: string): string[] {
-	const normalized = normalizeModelIdWhitespace(modelId).toLowerCase();
-	if (!normalized) return [];
-	const segments = (normalized.match(MODEL_ID_SEGMENT_PATTERN) ?? []).filter(
-		segment => MODEL_FAMILY_PREFIX_PATTERN.test(segment) && hasDigit(segment),
-	);
-	const unique = [...new Set(segments)];
-	unique.sort(compareSegmentPreference);
-	return unique;
+	const matches = normalizeModelIdWhitespace(modelId).toLowerCase().match(MODEL_ID_SEGMENT_PATTERN);
+	if (!matches) return [];
+	const segments = new Set<string>();
+	for (const segment of matches) {
+		if (MODEL_FAMILY_PREFIX_PATTERN.test(segment) && /\d/.test(segment)) segments.add(segment);
+	}
+	return [...segments].sort(compareSegmentPreference);
 }
 
 export function getLongestModelLikeIdSegment(modelId: string): string | undefined {
-	return getModelLikeIdSegments(modelId)[0];
+	const matches = normalizeModelIdWhitespace(modelId).toLowerCase().match(MODEL_ID_SEGMENT_PATTERN);
+	if (!matches) return undefined;
+	let best: string | undefined;
+	for (const segment of matches) {
+		if (
+			MODEL_FAMILY_PREFIX_PATTERN.test(segment) &&
+			/\d/.test(segment) &&
+			(best === undefined || compareSegmentPreference(segment, best) < 0)
+		) {
+			best = segment;
+		}
+	}
+	return best;
 }
 
-function normalizeModelIdWhitespace(value: string): string {
-	return value.trim().replace(/\s+/g, " ");
+function hasBracketAffixMarker(value: string): boolean {
+	for (let index = 0; index < value.length; index++) {
+		const code = value.charCodeAt(index);
+		if (code === 91 || code === 93 || code === 0x3010 || code === 0x3011) {
+			return true;
+		}
+	}
+	return false;
 }
 
 /**
@@ -39,18 +54,20 @@ function normalizeModelIdWhitespace(value: string): string {
  * upstream model id, e.g.
  *   "[Kiro] claude-opus-4-8"                -> "claude-opus-4-8"
  *   "[gcli转] gemini-3.1-pro-preview [假流]" -> "gemini-3.1-pro-preview"
+ *
+ * Candidates are returned most-stripped first: both ends, then leading-only, then trailing-only.
  */
 export function getBracketStrippedModelIdCandidates(modelId: string): string[] {
+	if (!hasBracketAffixMarker(modelId)) return [];
 	const normalized = normalizeModelIdWhitespace(modelId);
 	if (!normalized) return [];
 
-	const candidates = new Set<string>();
-	const withoutLeading = normalizeModelIdWhitespace(normalized.replace(LEADING_BRACKETED_AFFIX_PATTERN, ""));
+	const strippedLeading = normalized.replace(LEADING_BRACKETED_AFFIX_PATTERN, "");
+	const withoutLeading = normalizeModelIdWhitespace(strippedLeading);
 	const withoutTrailing = normalizeModelIdWhitespace(normalized.replace(TRAILING_BRACKETED_AFFIX_PATTERN, ""));
-	const withoutBoth = normalizeModelIdWhitespace(
-		normalized.replace(LEADING_BRACKETED_AFFIX_PATTERN, "").replace(TRAILING_BRACKETED_AFFIX_PATTERN, ""),
-	);
+	const withoutBoth = normalizeModelIdWhitespace(strippedLeading.replace(TRAILING_BRACKETED_AFFIX_PATTERN, ""));
 
+	const candidates = new Set<string>();
 	for (const candidate of [withoutBoth, withoutLeading, withoutTrailing]) {
 		if (candidate && candidate !== normalized) {
 			candidates.add(candidate);
diff --git a/packages/coding-agent/src/config/model-registry.ts b/packages/coding-agent/src/config/model-registry.ts
index a37fdb8e9..6b0a86a23 100644
--- a/packages/coding-agent/src/config/model-registry.ts
+++ b/packages/coding-agent/src/config/model-registry.ts
@@ -1,27 +1,19 @@
 import * as path from "node:path";
+import { registerCustomApi, unregisterCustomApis } from "@oh-my-pi/pi-ai/api-registry";
+import { readModelCache } from "@oh-my-pi/pi-ai/model-cache";
+import { createModelManager, type ModelManagerOptions, type ModelRefreshStrategy } from "@oh-my-pi/pi-ai/model-manager";
+import { enrichModelThinking } from "@oh-my-pi/pi-ai/model-thinking";
+import { getBundledModels, getBundledProviders } from "@oh-my-pi/pi-ai/models";
 import {
-	type Api,
-	type AssistantMessageEventStream,
-	type Context,
-	createModelManager,
-	enrichModelThinking,
-	getBundledModels,
-	getBundledProviders,
 	googleAntigravityModelManagerOptions,
 	googleGeminiCliModelManagerOptions,
-	type Model,
-	type ModelManagerOptions,
-	type ModelRefreshStrategy,
 	openaiCodexModelManagerOptions,
 	PROVIDER_DESCRIPTORS,
-	readModelCache,
-	registerCustomApi,
-	type SimpleStreamOptions,
-	type ThinkingConfig,
 	UNK_CONTEXT_WINDOW,
 	UNK_MAX_TOKENS,
-	unregisterCustomApis,
-} from "@oh-my-pi/pi-ai";
+} from "@oh-my-pi/pi-ai/provider-models";
+import type { Api, Context, Model, SimpleStreamOptions, ThinkingConfig } from "@oh-my-pi/pi-ai/types";
+import type { AssistantMessageEventStream } from "@oh-my-pi/pi-ai/utils/event-stream";
 
 // Sentinel for local-only OAuth token (LM Studio, vLLM) — declared inline to avoid loading
 // any provider module at startup. Must match `DEFAULT_LOCAL_TOKEN` in oauth/lm-studio.ts.
@@ -103,12 +95,14 @@ const STARTUP_MODEL_CACHE_PROVIDER_IDS: readonly string[] = [
 	...SPECIAL_MODEL_MANAGER_PROVIDER_IDS,
 ];
 
+import type { ApiKeyResolver } from "@oh-my-pi/pi-ai";
 import { registerOAuthProvider, unregisterOAuthProviders } from "@oh-my-pi/pi-ai/utils/oauth";
 import type { OAuthCredentials, OAuthLoginCallbacks } from "@oh-my-pi/pi-ai/utils/oauth/types";
 import { isRecord, logger } from "@oh-my-pi/pi-utils";
 import { parseModelString, resolveProviderModelReference } from "../config/model-resolver";
 import { isValidThemeColor, type ThemeColor } from "../modes/theme/theme";
 import type { AuthStorage, OAuthCredential } from "../session/auth-storage";
+import { type ApiKeyResolverOptions, createApiKeyResolver } from "./api-key-resolver";
 import { type ConfigError, ConfigFile } from "./config-file";
 import {
 	buildCanonicalModelIndex,
@@ -2373,12 +2367,33 @@ export class ModelRegistry {
 
 	/**
 	 * Get API key for a provider (e.g., "openai").
+	 *
+	 * `options.forceRefresh` powers step (b) of the auth-retry policy — it
+	 * re-mints the session-sticky OAuth token even when the cached copy still
+	 * looks valid. `options.signal` is threaded into any broker-bound refresh.
 	 */
-	async getApiKeyForProvider(provider: string, sessionId?: string, baseUrl?: string): Promise<string | undefined> {
+	async getApiKeyForProvider(
+		provider: string,
+		sessionId?: string,
+		options?: { baseUrl?: string; forceRefresh?: boolean; signal?: AbortSignal },
+	): Promise<string | undefined> {
 		if (this.#keylessProviders.has(provider) && !this.authStorage.hasAuth(provider)) {
 			return kNoAuth;
 		}
-		return this.authStorage.getApiKey(provider, sessionId, { baseUrl });
+		return this.authStorage.getApiKey(provider, sessionId, {
+			baseUrl: options?.baseUrl,
+			forceRefresh: options?.forceRefresh,
+			signal: options?.signal,
+		});
+	}
+
+	/**
+	 * Build an {@link ApiKeyResolver} for this provider, implementing the
+	 * central a/b/c auth-retry policy. Callers that need the initial key for
+	 * a guard can call `resolveApiKeyOnce(resolver)`.
+	 */
+	resolver(provider: string, options?: ApiKeyResolverOptions): ApiKeyResolver {
+		return createApiKeyResolver(this, provider, options);
 	}
 
 	async #peekApiKeyForProvider(provider: string): Promise<string | undefined> {
@@ -2609,6 +2624,14 @@ export class ModelRegistry {
 		}
 		return true;
 	}
+
+	/**
+	 * Clear all cooldown suppressions recorded via {@link suppressSelector}.
+	 * Used to reset retry-fallback cooldown state without a full {@link refresh}.
+	 */
+	clearSuppressedSelectors(): void {
+		this.#suppressedSelectors.clear();
+	}
 }
 
 /**
diff --git a/packages/coding-agent/src/config/settings-schema.ts b/packages/coding-agent/src/config/settings-schema.ts
index aa5253727..664488632 100644
--- a/packages/coding-agent/src/config/settings-schema.ts
+++ b/packages/coding-agent/src/config/settings-schema.ts
@@ -660,6 +660,16 @@ export const SETTINGS_SCHEMA = {
 		},
 	},
 
+	"display.smoothStreaming": {
+		type: "boolean",
+		default: true,
+		ui: {
+			tab: "appearance",
+			label: "Smooth Streaming",
+			description: "Reveal assistant text smoothly while streamed chunks arrive",
+		},
+	},
+
 	"display.showTokenUsage": {
 		type: "boolean",
 		default: false,
@@ -902,6 +912,15 @@ export const SETTINGS_SCHEMA = {
 				"Maximum wait between retries, in ms. When the provider asks us to wait longer than this and no credential or model fallback succeeds, the request fails fast instead of sleeping (e.g. 3-hour Anthropic rate-limit windows).",
 		},
 	},
+	"retry.modelFallback": {
+		type: "boolean",
+		default: true,
+		ui: {
+			tab: "model",
+			label: "Retry Model Fallback",
+			description: "Allow retry recovery to switch to configured fallback models",
+		},
+	},
 	"retry.fallbackChains": { type: "record", default: {} as Record<string, string[]> },
 	"retry.fallbackRevertPolicy": {
 		type: "enum",
@@ -1855,7 +1874,7 @@ export const SETTINGS_SCHEMA = {
 			tab: "editing",
 			label: "Hash Lines",
 			description:
-				"Include snapshot-tag headers and line numbers in read output for hashline edit mode (¶PATH#tag plus LINE:content)",
+				"Include snapshot-tag headers and line numbers in read output for hashline edit mode ([PATH#TAG] plus LINE:content)",
 		},
 	},
 
@@ -2493,13 +2512,13 @@ export const SETTINGS_SCHEMA = {
 	// Tool Discovery
 	"tools.discoveryMode": {
 		type: "enum",
-		values: ["off", "mcp-only", "all"] as const,
-		default: "off",
+		values: ["auto", "off", "mcp-only", "all"] as const,
+		default: "auto",
 		ui: {
 			tab: "tools",
 			label: "Tool Discovery",
 			description:
-				"Hide tools behind a search tool to save tokens. 'mcp-only' hides MCP tools; 'all' hides all non-essential built-ins too.",
+				"Hide tools behind a search tool to save tokens. 'auto' hides MCP tools once the tool set has more than 40 tools; 'mcp-only' always hides MCP tools; 'all' hides all non-essential built-ins too.",
 		},
 	},
 
@@ -3050,13 +3069,26 @@ export const SETTINGS_SCHEMA = {
 			],
 		},
 	},
-	"providers.parallelFetch": {
-		type: "boolean",
-		default: true,
+	"providers.fetch": {
+		type: "enum",
+		values: ["auto", "native", "trafilatura", "lynx", "parallel", "jina"] as const,
+		default: "auto",
 		ui: {
 			tab: "providers",
-			label: "Parallel Fetch",
-			description: "Use Parallel extract API for URL fetching when credentials are available",
+			label: "Fetch Provider",
+			description: "Reader backend priority for the fetch/read URL tool",
+			options: [
+				{
+					value: "auto",
+					label: "Auto",
+					description: "Priority: native > trafilatura > lynx > parallel > jina",
+				},
+				{ value: "native", label: "Native", description: "In-process HTML→Markdown converter (always available)" },
+				{ value: "trafilatura", label: "Trafilatura", description: "Auto-installs via uv/pip" },
+				{ value: "lynx", label: "Lynx", description: "Requires lynx system package" },
+				{ value: "parallel", label: "Parallel", description: "Requires PARALLEL_API_KEY" },
+				{ value: "jina", label: "Jina", description: "Uses r.jina.ai reader (JINA_API_KEY optional)" },
+			],
 		},
 	},
 	"provider.appendOnlyContext": {
@@ -3307,6 +3339,7 @@ export interface RetrySettings {
 	maxRetries: number;
 	baseDelayMs: number;
 	maxDelayMs: number;
+	modelFallback: boolean;
 }
 
 export interface MemoriesSettings {
diff --git a/packages/coding-agent/src/config/settings.ts b/packages/coding-agent/src/config/settings.ts
index f5dc33299..6f51f703d 100644
--- a/packages/coding-agent/src/config/settings.ts
+++ b/packages/coding-agent/src/config/settings.ts
@@ -240,11 +240,13 @@ export class Settings {
 		return promise.then(
 			instance => {
 				globalInstance = instance;
+				clearBoundSettingsMethods();
 				globalInstancePromise = Promise.resolve(instance);
 				return instance;
 			},
 			error => {
 				globalInstance = null;
+				clearBoundSettingsMethods();
 				throw error;
 			},
 		);
@@ -712,6 +714,17 @@ export class Settings {
 			}
 		}
 
+		// providers.parallelFetch (boolean) replaced by the providers.fetch reader
+		// priority enum. The new default ("auto") supersedes both old values —
+		// Parallel is now a deep fallback in the auto chain rather than the first
+		// choice — so drop the legacy key (flat and nested) and let the enum
+		// default apply.
+		const providersObj = raw.providers as Record<string, unknown> | undefined;
+		if (providersObj && "parallelFetch" in providersObj) {
+			delete providersObj.parallelFetch;
+		}
+		delete raw["providers.parallelFetch"];
+
 		// Map legacy `memories.enabled` boolean to the explicit `memory.backend`
 		// enum if the latter hasn't been set yet. Idempotent: subsequent
 		// migrations are no-ops once memory.backend is materialised.
@@ -967,6 +980,13 @@ export function onHindsightScopeChanged(cb: () => void): () => void {
 
 let globalInstance: Settings | null = null;
 let globalInstancePromise: Promise<Settings> | null = null;
+let boundSettingsInstance: Settings | null = null;
+let boundSettingsMethods = new Map<PropertyKey, unknown>();
+
+function clearBoundSettingsMethods(): void {
+	boundSettingsInstance = null;
+	boundSettingsMethods = new Map<PropertyKey, unknown>();
+}
 
 export function isSettingsInitialized(): boolean {
 	return globalInstance !== null;
@@ -979,6 +999,7 @@ export function isSettingsInitialized(): boolean {
 export function resetSettingsForTest(): void {
 	globalInstance = null;
 	globalInstancePromise = null;
+	clearBoundSettingsMethods();
 }
 
 /**
@@ -990,9 +1011,17 @@ export const settings = new Proxy({} as Settings, {
 		if (!globalInstance) {
 			throw new Error("Settings not initialized. Call Settings.init() first.");
 		}
-		const value = (globalInstance as unknown as Record<string | symbol, unknown>)[prop];
+		if (boundSettingsInstance !== globalInstance) {
+			clearBoundSettingsMethods();
+			boundSettingsInstance = globalInstance;
+		}
+		const value = (globalInstance as unknown as Record<PropertyKey, unknown>)[prop];
 		if (typeof value === "function") {
-			return value.bind(globalInstance);
+			const cached = boundSettingsMethods.get(prop);
+			if (cached) return cached;
+			const bound = value.bind(globalInstance);
+			boundSettingsMethods.set(prop, bound);
+			return bound;
 		}
 		return value;
 	},
diff --git a/packages/coding-agent/src/dap/client.ts b/packages/coding-agent/src/dap/client.ts
index 79b333a82..a93e1df9c 100644
--- a/packages/coding-agent/src/dap/client.ts
+++ b/packages/coding-agent/src/dap/client.ts
@@ -1,4 +1,5 @@
-import { logger, ptree } from "@oh-my-pi/pi-utils";
+import * as fs from "node:fs/promises";
+import { isEnoent, logger, ptree } from "@oh-my-pi/pi-utils";
 import { NON_INTERACTIVE_ENV } from "../exec/non-interactive-env";
 import { ToolAbortError } from "../tools/tool-errors";
 import type {
@@ -165,19 +166,7 @@ export class DapClient {
 			detached: true,
 		});
 
-		// Wait for the socket file to appear (dlv needs to start listening)
-		await waitForCondition(
-			() => {
-				try {
-					Bun.file(socketPath).size;
-					return true;
-				} catch {
-					return false;
-				}
-			},
-			10_000,
-			proc,
-		);
+		await waitForCondition(() => isUnixSocketReady(socketPath), 10_000, proc);
 
 		const { readable, writeSink, socket } = await connectSocket({ unix: socketPath });
 		const client = new DapClient(adapter, cwd, proc, { readable, writeSink, socket });
@@ -553,15 +542,24 @@ export class DapClient {
 	}
 }
 
+async function isUnixSocketReady(socketPath: string): Promise<boolean> {
+	try {
+		return (await fs.stat(socketPath)).isSocket();
+	} catch (error) {
+		if (isEnoent(error)) return false;
+		throw error;
+	}
+}
+
 /** Poll a condition until it returns true, or timeout/process exit. */
 async function waitForCondition(
-	check: () => boolean,
+	check: () => boolean | Promise<boolean>,
 	timeoutMs: number,
 	proc: { exitCode: number | null },
 ): Promise<void> {
 	const deadline = Date.now() + timeoutMs;
 	while (Date.now() < deadline) {
-		if (check()) return;
+		if (await check()) return;
 		if (proc.exitCode !== null) {
 			throw new Error("Adapter process exited before socket was ready");
 		}
diff --git a/packages/coding-agent/src/dap/config.ts b/packages/coding-agent/src/dap/config.ts
index 271a38fc6..31f870b91 100644
--- a/packages/coding-agent/src/dap/config.ts
+++ b/packages/coding-agent/src/dap/config.ts
@@ -27,6 +27,7 @@ function normalizeAdapterConfig(config: unknown): DapAdapterConfig | null {
 		rootMarkers: normalizeStringArray(config.rootMarkers),
 		launchDefaults: normalizeObject(config.launchDefaults),
 		attachDefaults: normalizeObject(config.attachDefaults),
+		acceptsDirectoryProgram: config.acceptsDirectoryProgram === true,
 		...(connectMode ? { connectMode } : {}),
 	};
 }
@@ -64,6 +65,7 @@ export function resolveAdapter(adapterName: string, cwd: string): DapResolvedAda
 		launchDefaults: config.launchDefaults ?? {},
 		attachDefaults: config.attachDefaults ?? {},
 		connectMode: config.connectMode ?? "stdio",
+		acceptsDirectoryProgram: config.acceptsDirectoryProgram === true,
 	};
 }
 
@@ -124,12 +126,19 @@ function sortAdaptersForLaunch(program: string, cwd: string, adapters: DapResolv
 	return rootAware.map(entry => entry.adapter);
 }
 
-export function selectLaunchAdapter(program: string, cwd: string, adapterName?: string): DapResolvedAdapter | null {
+export function selectLaunchAdapter(
+	program: string,
+	cwd: string,
+	adapterName?: string,
+	programKind: LaunchProgramKind = "file",
+): DapResolvedAdapter | null {
 	if (adapterName) {
 		return resolveAdapter(adapterName, cwd);
 	}
 	const matches = getMatchingAdapters(program, cwd);
-	const sorted = sortAdaptersForLaunch(program, cwd, matches);
+	const candidates =
+		programKind === "directory" ? matches.filter(adapter => adapter.acceptsDirectoryProgram) : matches;
+	const sorted = sortAdaptersForLaunch(program, cwd, candidates.length > 0 ? candidates : matches);
 	return sorted[0] ?? null;
 }
 
@@ -148,3 +157,33 @@ export function selectAttachAdapter(cwd: string, adapterName?: string, port?: nu
 	}
 	return available[0] ?? null;
 }
+
+/** How the launch `program` resolves on disk. `"missing"` is reserved for
+ *  programs the adapter creates on demand (rare); we treat them like files. */
+export type LaunchProgramKind = "file" | "directory" | "missing";
+
+/** Compute adapter-specific launch arguments that depend on the resolved
+ *  program. Returned values are spread over `adapter.launchDefaults` so they
+ *  take precedence over the static defaults but can still be overridden by
+ *  the fields `DapSessionManager.launch` sets explicitly (program, cwd, args).
+ *
+ *  Currently scoped to dlv, where `mode` selects how the program path is
+ *  interpreted: directories and `.go` source files debug as a Go package
+ *  (`mode=debug`), anything else is treated as a compiled binary (`mode=exec`).
+ */
+export function resolveLaunchOverrides(
+	adapter: DapResolvedAdapter,
+	program: string,
+	programKind: LaunchProgramKind,
+): Record<string, unknown> {
+	if (adapter.name === "dlv") {
+		const extension = path.extname(program).toLowerCase();
+		if (programKind === "directory" || extension === ".go") {
+			return { mode: "debug" };
+		}
+		if (programKind === "file") {
+			return { mode: "exec" };
+		}
+	}
+	return {};
+}
diff --git a/packages/coding-agent/src/dap/defaults.json b/packages/coding-agent/src/dap/defaults.json
index dd3725236..5ea74ef25 100644
--- a/packages/coding-agent/src/dap/defaults.json
+++ b/packages/coding-agent/src/dap/defaults.json
@@ -65,6 +65,7 @@
 		"languages": ["go"],
 		"fileTypes": [".go"],
 		"rootMarkers": ["go.mod", "go.sum"],
+		"acceptsDirectoryProgram": true,
 		"launchDefaults": {
 			"request": "launch",
 			"mode": "debug",
diff --git a/packages/coding-agent/src/dap/session.ts b/packages/coding-agent/src/dap/session.ts
index 0dc66c63b..57be0afc1 100644
--- a/packages/coding-agent/src/dap/session.ts
+++ b/packages/coding-agent/src/dap/session.ts
@@ -259,6 +259,7 @@ export class DapSessionManager {
 			session.needsConfigurationDone = session.capabilities.supportsConfigurationDoneRequest === true;
 			const launchArguments: DapLaunchArguments = {
 				...options.adapter.launchDefaults,
+				...(options.extraLaunchArguments ?? {}),
 				program: options.program,
 				cwd: options.cwd,
 				args: options.args,
diff --git a/packages/coding-agent/src/dap/types.ts b/packages/coding-agent/src/dap/types.ts
index eb4659de9..542e9088c 100644
--- a/packages/coding-agent/src/dap/types.ts
+++ b/packages/coding-agent/src/dap/types.ts
@@ -488,6 +488,10 @@ export interface DapAdapterConfig {
 	 *  On Linux, connects via a unix domain socket.
 	 *  On macOS, the adapter dials into a local TCP listener (--client-addr). */
 	connectMode?: "stdio" | "socket";
+	/** When true, the adapter accepts a directory as the launch `program`
+	 *  (e.g. dlv treats it as a Go package path). When false/undefined, the
+	 *  debug tool rejects directory programs upfront. */
+	acceptsDirectoryProgram?: boolean;
 }
 
 export interface DapResolvedAdapter {
@@ -501,6 +505,7 @@ export interface DapResolvedAdapter {
 	launchDefaults: Record<string, unknown>;
 	attachDefaults: Record<string, unknown>;
 	connectMode: "stdio" | "socket";
+	acceptsDirectoryProgram: boolean;
 }
 
 export interface DapBreakpointRecord {
@@ -589,6 +594,11 @@ export interface DapLaunchSessionOptions {
 	program: string;
 	args?: string[];
 	cwd: string;
+	/** Per-launch overrides merged over `adapter.launchDefaults`. Used to
+	 *  inject adapter-specific values that depend on the resolved program
+	 *  (e.g. dlv's `mode` switches between `debug` and `exec` based on
+	 *  whether `program` is a Go package path or a compiled binary). */
+	extraLaunchArguments?: Record<string, unknown>;
 }
 
 export interface DapAttachSessionOptions {
diff --git a/packages/coding-agent/src/debug/index.ts b/packages/coding-agent/src/debug/index.ts
index eece0d938..17840cc4b 100644
--- a/packages/coding-agent/src/debug/index.ts
+++ b/packages/coding-agent/src/debug/index.ts
@@ -19,6 +19,7 @@ import {
 } from "@oh-my-pi/pi-tui";
 import { getSessionsDir } from "@oh-my-pi/pi-utils";
 import { DynamicBorder } from "../modes/components/dynamic-border";
+import { TranscriptBlock } from "../modes/components/transcript-container";
 import { getSelectListTheme, getSymbolTheme, theme } from "../modes/theme/theme";
 import type { InteractiveModeContext } from "../modes/types";
 import { formatBytes } from "../tools/render-utils";
@@ -150,13 +151,13 @@ export class DebugSelectorComponent extends Container {
 		}
 
 		// Show message and wait for keypress
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Text(theme.fg("accent", `${theme.status.info} CPU profiling started`), 1, 0));
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(
+		const block = new TranscriptBlock();
+		block.addChild(new Text(theme.fg("accent", `${theme.status.info} CPU profiling started`), 1, 0));
+		block.addChild(new Spacer(1));
+		block.addChild(
 			new Text(theme.fg("muted", "Reproduce the performance issue, then press Enter to stop profiling."), 1, 0),
 		);
-		this.ctx.ui.requestRender();
+		this.ctx.present(block);
 
 		// Wait for Enter keypress
 		const { promise, resolve } = Promise.withResolvers<void>();
@@ -201,19 +202,16 @@ export class DebugSelectorComponent extends Container {
 			loader.stop();
 			this.ctx.statusContainer.clear();
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
-				new Text(theme.fg("success", `${theme.status.success} Performance report saved`), 1, 0),
-			);
-			this.ctx.chatContainer.addChild(new Text(theme.fg("dim", formatFileHyperlink(result.path)), 1, 0));
-			this.ctx.chatContainer.addChild(new Text(theme.fg("dim", `Files: ${result.files.length}`), 1, 0));
+			const block = new TranscriptBlock();
+			block.addChild(new Text(theme.fg("success", `${theme.status.success} Performance report saved`), 1, 0));
+			block.addChild(new Text(theme.fg("dim", formatFileHyperlink(result.path)), 1, 0));
+			block.addChild(new Text(theme.fg("dim", `Files: ${result.files.length}`), 1, 0));
+			this.ctx.present(block);
 		} catch (err) {
 			loader.stop();
 			this.ctx.statusContainer.clear();
 			this.ctx.showError(`Failed to create report: ${err instanceof Error ? err.message : String(err)}`);
 		}
-
-		this.ctx.ui.requestRender();
 	}
 
 	async #handleWorkReport(): Promise<void> {
@@ -231,15 +229,13 @@ export class DebugSelectorComponent extends Container {
 
 			openPath(tmpPath);
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
+			this.ctx.present([
+				new Spacer(1),
 				new Text(theme.fg("dim", `Opened flamegraph (${workProfile.sampleCount} samples)`), 1, 0),
-			);
+			]);
 		} catch (err) {
 			this.ctx.showError(`Failed to open profile: ${err instanceof Error ? err.message : String(err)}`);
 		}
-
-		this.ctx.ui.requestRender();
 	}
 
 	async #handleDumpReport(): Promise<void> {
@@ -262,19 +258,16 @@ export class DebugSelectorComponent extends Container {
 			loader.stop();
 			this.ctx.statusContainer.clear();
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
-				new Text(theme.fg("success", `${theme.status.success} Report bundle saved`), 1, 0),
-			);
-			this.ctx.chatContainer.addChild(new Text(theme.fg("dim", formatFileHyperlink(result.path)), 1, 0));
-			this.ctx.chatContainer.addChild(new Text(theme.fg("dim", `Files: ${result.files.length}`), 1, 0));
+			const block = new TranscriptBlock();
+			block.addChild(new Text(theme.fg("success", `${theme.status.success} Report bundle saved`), 1, 0));
+			block.addChild(new Text(theme.fg("dim", formatFileHyperlink(result.path)), 1, 0));
+			block.addChild(new Text(theme.fg("dim", `Files: ${result.files.length}`), 1, 0));
+			this.ctx.present(block);
 		} catch (err) {
 			loader.stop();
 			this.ctx.statusContainer.clear();
 			this.ctx.showError(`Failed to create report: ${err instanceof Error ? err.message : String(err)}`);
 		}
-
-		this.ctx.ui.requestRender();
 	}
 
 	async #handleMemoryReport(): Promise<void> {
@@ -301,19 +294,16 @@ export class DebugSelectorComponent extends Container {
 			loader.stop();
 			this.ctx.statusContainer.clear();
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
-				new Text(theme.fg("success", `${theme.status.success} Memory report saved`), 1, 0),
-			);
-			this.ctx.chatContainer.addChild(new Text(theme.fg("dim", formatFileHyperlink(result.path)), 1, 0));
-			this.ctx.chatContainer.addChild(new Text(theme.fg("dim", `Files: ${result.files.length}`), 1, 0));
+			const block = new TranscriptBlock();
+			block.addChild(new Text(theme.fg("success", `${theme.status.success} Memory report saved`), 1, 0));
+			block.addChild(new Text(theme.fg("dim", formatFileHyperlink(result.path)), 1, 0));
+			block.addChild(new Text(theme.fg("dim", `Files: ${result.files.length}`), 1, 0));
+			this.ctx.present(block);
 		} catch (err) {
 			loader.stop();
 			this.ctx.statusContainer.clear();
 			this.ctx.showError(`Failed to create report: ${err instanceof Error ? err.message : String(err)}`);
 		}
-
-		this.ctx.ui.requestRender();
 	}
 
 	async #handleViewLogs(): Promise<void> {
@@ -365,15 +355,14 @@ export class DebugSelectorComponent extends Container {
 			const info = await collectSystemInfo();
 			const formatted = formatSystemInfo(info);
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(new DynamicBorder());
-			this.ctx.chatContainer.addChild(new Text(formatted, 1, 0));
-			this.ctx.chatContainer.addChild(new DynamicBorder());
+			const block = new TranscriptBlock();
+			block.addChild(new DynamicBorder());
+			block.addChild(new Text(formatted, 1, 0));
+			block.addChild(new DynamicBorder());
+			this.ctx.present(block);
 		} catch (err) {
 			this.ctx.showError(`Failed to collect system info: ${err instanceof Error ? err.message : String(err)}`);
 		}
-
-		this.ctx.ui.requestRender();
 	}
 
 	async #handleViewTerminalState(): Promise<void> {
@@ -384,11 +373,11 @@ export class DebugSelectorComponent extends Container {
 		});
 		const formatted = formatTerminalState(info);
 
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new DynamicBorder());
-		this.ctx.chatContainer.addChild(new Text(formatted, 1, 0));
-		this.ctx.chatContainer.addChild(new DynamicBorder());
-		this.ctx.ui.requestRender();
+		const block = new TranscriptBlock();
+		block.addChild(new DynamicBorder());
+		block.addChild(new Text(formatted, 1, 0));
+		block.addChild(new DynamicBorder());
+		this.ctx.present(block);
 	}
 
 	async #handleViewProtocols(): Promise<void> {
@@ -407,15 +396,14 @@ export class DebugSelectorComponent extends Container {
 			TERMINAL.sendNotification(notification);
 		}
 
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(
+		this.ctx.present([
+			new Spacer(1),
 			new ProtocolProbeComponent({
 				image: buildSampleImage(),
 				imageBudget: this.ctx.ui.imageBudget,
 				notificationSuppressed: suppressed,
 			}),
-		);
-		this.ctx.ui.requestRender();
+		]);
 	}
 
 	async #handleTranscriptExport(): Promise<void> {
@@ -487,21 +475,19 @@ export class DebugSelectorComponent extends Container {
 			loader.stop();
 			this.ctx.statusContainer.clear();
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
+			this.ctx.present([
+				new Spacer(1),
 				new Text(
 					theme.fg("success", `${theme.status.success} Cleared ${result.removed} artifact directories`),
 					1,
 					0,
 				),
-			);
+			]);
 		} catch (err) {
 			loader.stop();
 			this.ctx.statusContainer.clear();
 			this.ctx.showError(`Failed to clear cache: ${err instanceof Error ? err.message : String(err)}`);
 		}
-
-		this.ctx.ui.requestRender();
 	}
 
 	#getResolvedSettings(): Record<string, unknown> {
diff --git a/packages/coding-agent/src/debug/raw-sse.ts b/packages/coding-agent/src/debug/raw-sse.ts
index a369fb246..3be286152 100644
--- a/packages/coding-agent/src/debug/raw-sse.ts
+++ b/packages/coding-agent/src/debug/raw-sse.ts
@@ -1,4 +1,12 @@
-import { type Component, matchesKey, padding, replaceTabs, truncateToWidth, visibleWidth } from "@oh-my-pi/pi-tui";
+import {
+	type Component,
+	matchesKey,
+	padding,
+	replaceTabs,
+	ScrollView,
+	truncateToWidth,
+	visibleWidth,
+} from "@oh-my-pi/pi-tui";
 import { sanitizeText } from "@oh-my-pi/pi-utils";
 import { theme } from "../modes/theme/theme";
 import { copyToClipboard } from "../utils/clipboard";
@@ -146,14 +154,20 @@ export class RawSseViewerComponent implements Component {
 		const innerWidth = Math.max(1, this.#lastRenderWidth - 2);
 		const bodyHeight = this.#bodyHeight();
 		const rawLines = this.#renderRawLines(innerWidth);
-		const body = rawLines.slice(this.#scrollOffset, this.#scrollOffset + bodyHeight);
-		while (body.length < bodyHeight) body.push("");
+		const sv = new ScrollView(rawLines.slice(this.#scrollOffset, this.#scrollOffset + bodyHeight), {
+			height: bodyHeight,
+			scrollbar: "auto",
+			totalRows: rawLines.length,
+			theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+		});
+		sv.setScrollOffset(this.#scrollOffset);
+		const bodyRows = sv.render(innerWidth);
 
 		return [
 			this.#frameTop(innerWidth),
 			this.#frameLine(this.#summaryText(), innerWidth),
 			this.#frameSeparator(innerWidth),
-			...body.map(line => this.#frameLine(line, innerWidth)),
+			...bodyRows.map(line => this.#frameLine(line, innerWidth)),
 			this.#frameLine(this.#statusText(), innerWidth),
 			this.#frameBottom(innerWidth),
 		];
diff --git a/packages/coding-agent/src/edit/file-snapshot-store.ts b/packages/coding-agent/src/edit/file-snapshot-store.ts
index ca51dbd11..467aab33e 100644
--- a/packages/coding-agent/src/edit/file-snapshot-store.ts
+++ b/packages/coding-agent/src/edit/file-snapshot-store.ts
@@ -14,7 +14,7 @@ import { normalizeToLF } from "./normalize";
 /**
  * Upper bound on the file size we snapshot. A section tag is a content hash of
  * the *whole* file, so minting one means holding the full normalized text in
- * the store. Files above this cap emit no `¶path#tag` header — line-anchored
+ * the store. Files above this cap emit no `[path#tag]` header — line-anchored
  * editing of multi-megabyte files is out of scope under the full-content model.
  */
 export const SNAPSHOT_MAX_BYTES = 4 * 1024 * 1024;
diff --git a/packages/coding-agent/src/edit/index.ts b/packages/coding-agent/src/edit/index.ts
index 6686c4d5c..8bbb30e31 100644
--- a/packages/coding-agent/src/edit/index.ts
+++ b/packages/coding-agent/src/edit/index.ts
@@ -275,7 +275,7 @@ function extractApprovalPath(args: unknown): string {
 	const record = args && typeof args === "object" ? (args as Record<string, unknown>) : {};
 	const input = typeof record.input === "string" ? record.input : undefined;
 	if (input) {
-		const hashlineMatch = /^(?:¶|§|@)([^\s#]+)/m.exec(input);
+		const hashlineMatch = /^\[([^#\r\n]+)(?:#[0-9a-fA-F]{4})?\]/m.exec(input);
 		if (hashlineMatch?.[1]) return hashlineMatch[1];
 
 		const applyPatchMatch = /^\*\*\* (?:Add|Update|Delete) File:\s*(.+)$/m.exec(input);
diff --git a/packages/coding-agent/src/edit/renderer.ts b/packages/coding-agent/src/edit/renderer.ts
index f2de22869..6bedb5947 100644
--- a/packages/coding-agent/src/edit/renderer.ts
+++ b/packages/coding-agent/src/edit/renderer.ts
@@ -2,9 +2,9 @@
  * Edit tool renderer and LSP batching helpers.
  */
 
-import { HL_FILE_PREFIX } from "@oh-my-pi/hashline";
+import { HL_FILE_PREFIX, HL_FILE_SUFFIX } from "@oh-my-pi/hashline";
 import type { Component } from "@oh-my-pi/pi-tui";
-import { Text, visibleWidth, wrapTextWithAnsi } from "@oh-my-pi/pi-tui";
+import { visibleWidth, wrapTextWithAnsi } from "@oh-my-pi/pi-tui";
 import { sanitizeText } from "@oh-my-pi/pi-utils";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import type { FileDiagnosticsResult } from "../lsp";
@@ -16,7 +16,6 @@ import {
 	formatDiffStats,
 	formatExpandHint,
 	formatStatusIcon,
-	formatTitle,
 	getDiffStats,
 	getLspBatchRequest,
 	type LspBatchRequest,
@@ -25,7 +24,7 @@ import {
 	shortenPath,
 	truncateDiffByHunk,
 } from "../tools/render-utils";
-import { fileHyperlink, Hasher, type RenderCache, renderStatusLine, truncateToWidth } from "../tui";
+import { fileHyperlink, framedBlock, Hasher, type RenderCache, renderStatusLine, truncateToWidth } from "../tui";
 import type { EditMode } from "../utils/edit-mode";
 import type { DiffError, DiffResult } from "./diff";
 import { type ApplyPatchEntry, expandApplyPatchToEntries, expandApplyPatchToPreviewEntries } from "./modes/apply-patch";
@@ -179,11 +178,6 @@ function countEditFiles(edits: EditRenderEntry[]): number {
 	return new Set(edits.map(edit => filePathFromEditEntry(edit.path)).filter(Boolean)).size;
 }
 
-function countLines(text: string): number {
-	if (!text) return 0;
-	return text.split("\n").length;
-}
-
 function getOperationTitle(op: Operation | undefined): string {
 	return op === "create" ? "Create" : op === "delete" ? "Delete" : "Edit";
 }
@@ -191,10 +185,14 @@ function getOperationTitle(op: Operation | undefined): string {
 function formatEditPathDisplay(
 	rawPath: string,
 	uiTheme: Theme,
-	options?: { rename?: string; firstChangedLine?: number },
+	options?: { rename?: string; firstChangedLine?: number; linkPath?: string; renameLinkPath?: string },
 ): string {
+	// `rawPath`/`rename` are shown (cwd-relative) but the OSC 8 link targets the
+	// absolute path when known — a relative `rawPath` would yield a `file:///rel`
+	// URI that resolves against filesystem root instead of cwd.
+	const linkTarget = options?.linkPath || rawPath;
 	let pathDisplay = rawPath
-		? fileHyperlink(rawPath, uiTheme.fg("accent", shortenPath(rawPath)))
+		? fileHyperlink(linkTarget, uiTheme.fg("accent", shortenPath(rawPath)))
 		: uiTheme.fg("toolOutput", "…");
 
 	if (options?.firstChangedLine) {
@@ -202,7 +200,8 @@ function formatEditPathDisplay(
 	}
 
 	if (options?.rename) {
-		pathDisplay += ` ${uiTheme.fg("dim", "→")} ${fileHyperlink(options.rename, uiTheme.fg("accent", shortenPath(options.rename)))}`;
+		const renameTarget = options.renameLinkPath || options.rename;
+		pathDisplay += ` ${uiTheme.fg("dim", "→")} ${fileHyperlink(renameTarget, uiTheme.fg("accent", shortenPath(options.rename)))}`;
 	}
 
 	return pathDisplay;
@@ -211,7 +210,7 @@ function formatEditPathDisplay(
 function formatEditDescription(
 	rawPath: string,
 	uiTheme: Theme,
-	options?: { rename?: string; firstChangedLine?: number },
+	options?: { rename?: string; firstChangedLine?: number; linkPath?: string; renameLinkPath?: string },
 ): { language: string; description: string } {
 	const language = getLanguageFromPath(rawPath) ?? "text";
 	const icon = uiTheme.fg("muted", uiTheme.getLangIcon(language));
@@ -233,19 +232,22 @@ function renderPlainTextPreview(text: string, uiTheme: Theme, filePath?: string)
 	return preview.trimEnd();
 }
 
-function formatStreamingDiff(diff: string, rawPath: string, uiTheme: Theme, label = "streaming"): string {
+function formatStreamingDiff(
+	diff: string,
+	rawPath: string,
+	uiTheme: Theme,
+	expanded: boolean,
+	label = "streaming",
+): string {
 	if (!diff) return "";
-	// "Cursor" tail window: pin the last EDIT_STREAMING_PREVIEW_LINES rows to the
-	// bottom of the diff so freshly streamed changes stay on screen, and accept
-	// the trailing rows "from the back" once the diff outgrows the window. The
-	// whole-file diff is recomputed on every streamed chunk and its Myers
-	// alignment is not monotonic in payload length, so a hunk-aware window that
-	// kept whole change segments gained and lost rows tick to tick — the box
-	// stuttered, and the earlier high-water fix traded that for a half-empty
-	// rectangle. A strict fixed-height window keeps the box steady and always
-	// full of real diff context instead of blank padding.
+	// Collapsed uses a "Cursor" tail window: pin the last
+	// EDIT_STREAMING_PREVIEW_LINES rows to the bottom so freshly streamed changes
+	// stay on screen. The whole-file diff is recomputed on every streamed chunk
+	// and its Myers alignment is not monotonic in payload length, so a hunk-aware
+	// window stutters as rows move between hunks. Expanded deliberately lifts that
+	// cap for the approval-time full view.
 	const allLines = diff.replace(/\n+$/u, "").split("\n");
-	const hiddenLines = Math.max(0, allLines.length - EDIT_STREAMING_PREVIEW_LINES);
+	const hiddenLines = expanded ? 0 : Math.max(0, allLines.length - EDIT_STREAMING_PREVIEW_LINES);
 	const visible = hiddenLines > 0 ? allLines.slice(hiddenLines) : allLines;
 	let text = "\n\n";
 	if (hiddenLines > 0) {
@@ -256,19 +258,11 @@ function formatStreamingDiff(diff: string, rawPath: string, uiTheme: Theme, labe
 		text += `${uiTheme.fg("dim", `… (${remainder.join(", ")} above)`)}\n`;
 	}
 	text += renderDiffColored(visible.join("\n"), { filePath: rawPath });
-	text += uiTheme.fg("dim", `\n(${label})`);
+	if (!expanded || label !== "preview") text += uiTheme.fg("dim", `\n(${label})`);
 	return text;
 }
 
-function formatMetadataLine(lineCount: number | null, language: string | undefined, uiTheme: Theme): string {
-	const icon = uiTheme.getLangIcon(language);
-	if (lineCount !== null) {
-		return uiTheme.fg("dim", `${icon} ${lineCount} lines`);
-	}
-	return uiTheme.fg("dim", `${icon}`);
-}
-
-function formatMultiFileStreamingDiff(previews: PerFileDiffPreview[], uiTheme: Theme): string {
+function formatMultiFileStreamingDiff(previews: PerFileDiffPreview[], uiTheme: Theme, expanded: boolean): string {
 	const parts: string[] = [];
 	for (const preview of previews) {
 		if (!preview.diff && !preview.error) continue;
@@ -278,7 +272,7 @@ function formatMultiFileStreamingDiff(previews: PerFileDiffPreview[], uiTheme: T
 			continue;
 		}
 		if (preview.diff) {
-			parts.push(`${header}${formatStreamingDiff(preview.diff, preview.path, uiTheme, "preview")}`);
+			parts.push(`${header}${formatStreamingDiff(preview.diff, preview.path, uiTheme, expanded, "preview")}`);
 		}
 	}
 	return parts.join("");
@@ -289,16 +283,17 @@ function getCallPreview(
 	rawPath: string,
 	uiTheme: Theme,
 	renderContext: EditRenderContext | undefined,
+	expanded: boolean,
 ): string {
 	const multi = renderContext?.perFileDiffPreview;
 	if (multi && multi.length > 1 && multi.some(p => p.diff || p.error)) {
-		return formatMultiFileStreamingDiff(multi, uiTheme);
+		return formatMultiFileStreamingDiff(multi, uiTheme, expanded);
 	}
 	if (args.previewDiff) {
-		return formatStreamingDiff(args.previewDiff, rawPath, uiTheme, "preview");
+		return formatStreamingDiff(args.previewDiff, rawPath, uiTheme, expanded, "preview");
 	}
 	if (args.diff && args.op) {
-		return formatStreamingDiff(args.diff, rawPath, uiTheme);
+		return formatStreamingDiff(args.diff, rawPath, uiTheme, expanded);
 	}
 	if (args.diff) {
 		return renderPlainTextPreview(args.diff, uiTheme, rawPath);
@@ -328,12 +323,12 @@ function normalizeHashlineInputPreviewPath(rawPath: string): string {
 }
 
 function parseHashlineInputPreviewHeader(line: string): string | null {
-	if (!line.startsWith(HL_FILE_PREFIX)) return null;
-	// Mirror hashline/input.ts: strip every leading file marker so canonical
-	// `¶ PATH` headers and stray `¶¶ PATH` / `¶¶¶PATH` runs render clean paths.
-	let prefixEnd = 0;
-	while (prefixEnd < line.length && line[prefixEnd] === HL_FILE_PREFIX) prefixEnd++;
-	const body = line.slice(prefixEnd).trim();
+	const trimmed = line.trimEnd();
+	if (!trimmed.startsWith(HL_FILE_PREFIX)) return null;
+	// Keep streaming previews tolerant while the closing bracket is still
+	// being generated; the parser enforces the final `[path#TAG]` shape.
+	const bodyEnd = trimmed.endsWith(HL_FILE_SUFFIX) ? trimmed.length - HL_FILE_SUFFIX.length : trimmed.length;
+	const body = trimmed.slice(HL_FILE_PREFIX.length, bodyEnd).trim();
 	const previewPath = normalizeHashlineInputPreviewPath(body);
 	return previewPath.length > 0 ? previewPath : null;
 }
@@ -383,6 +378,13 @@ function getApplyPatchRenderSummary(
 	}
 }
 
+function formatDiffStatsSuffix(diff: string, uiTheme: Theme): string {
+	const { added, removed, hunks } = getDiffStats(diff);
+	const stats = formatDiffStats(added, removed, hunks, uiTheme);
+	if (!stats) return "";
+	return ` ${uiTheme.fg("dim", uiTheme.format.bracketLeft)}${stats}${uiTheme.fg("dim", uiTheme.format.bracketRight)}`;
+}
+
 function renderDiffSection(
 	diff: string,
 	rawPath: string,
@@ -390,15 +392,6 @@ function renderDiffSection(
 	uiTheme: Theme,
 	renderDiffFn: (t: string, o?: { filePath?: string }) => string,
 ): string {
-	let text = "";
-	const diffStats = getDiffStats(diff);
-	text += `\n${uiTheme.fg("dim", uiTheme.format.bracketLeft)}${formatDiffStats(
-		diffStats.added,
-		diffStats.removed,
-		diffStats.hunks,
-		uiTheme,
-	)}${uiTheme.fg("dim", uiTheme.format.bracketRight)}`;
-
 	const {
 		text: truncatedDiff,
 		hiddenHunks,
@@ -407,7 +400,7 @@ function renderDiffSection(
 		? { text: diff, hiddenHunks: 0, hiddenLines: 0 }
 		: truncateDiffByHunk(diff, PREVIEW_LIMITS.DIFF_COLLAPSED_HUNKS, PREVIEW_LIMITS.DIFF_COLLAPSED_LINES);
 
-	text += `\n\n${renderDiffFn(truncatedDiff, { filePath: rawPath })}`;
+	let text = `\n${renderDiffFn(truncatedDiff, { filePath: rawPath })}`;
 	if (!expanded && (hiddenHunks > 0 || hiddenLines > 0)) {
 		const remainder: string[] = [];
 		if (hiddenHunks > 0) remainder.push(`${hiddenHunks} more hunks`);
@@ -470,23 +463,31 @@ export const editToolRenderer = {
 		const rename = editArgs.rename || firstEdit?.rename || firstEdit?.move || firstApplyPatchEntry?.rename;
 		const op = editArgs.op || firstEdit?.op || firstApplyPatchEntry?.op;
 		const { description } = formatEditDescription(rawPath, uiTheme, { rename });
-		const spinner =
-			options?.spinnerFrame !== undefined ? formatStatusIcon("running", uiTheme, options.spinnerFrame) : "";
-		let text = `${formatTitle(getOperationTitle(op), uiTheme)} ${spinner ? `${spinner} ` : ""}${description}`;
-		// Show file count hint for multi-file edits
 		let fileCount = hashlineInputSummary?.entries.length ?? applyPatchSummary?.entries.length ?? 0;
 		if (Array.isArray(editArgs.edits)) {
 			fileCount = countEditFiles(editArgs.edits);
 		}
-		if (fileCount > 1) {
-			text += uiTheme.fg("dim", ` (+${fileCount - 1} more)`);
-		}
-		text += getCallPreview(editArgs, rawPath, uiTheme, renderContext);
-		if (applyPatchSummary?.error) {
-			text += `\n\n${uiTheme.fg("error", truncateToWidth(replaceTabs(applyPatchSummary.error, rawPath), CALL_TEXT_PREVIEW_WIDTH))}`;
-		}
-
-		return new Text(text, 0, 0);
+		return framedBlock(uiTheme, width => {
+			let header = renderStatusLine(
+				{ icon: "pending", spinnerFrame: options?.spinnerFrame, title: getOperationTitle(op), description },
+				uiTheme,
+			);
+			if (fileCount > 1) header += uiTheme.fg("dim", ` (+${fileCount - 1} more)`);
+			let body = getCallPreview(editArgs, rawPath, uiTheme, renderContext, options.expanded);
+			if (applyPatchSummary?.error) {
+				body += `\n${uiTheme.fg("error", truncateToWidth(replaceTabs(applyPatchSummary.error, rawPath), Math.max(1, width - 2)))}`;
+			}
+			const bodyLines = body ? body.split("\n") : [];
+			while (bodyLines.length > 0 && bodyLines[0].trim() === "") bodyLines.shift();
+			return {
+				header,
+				sections: bodyLines.length > 0 ? [{ lines: bodyLines }] : [],
+				state: applyPatchSummary?.error ? "error" : "pending",
+				borderColor: applyPatchSummary?.error ? "error" : "borderMuted",
+				width,
+				contentPaddingLeft: 0,
+			};
+		});
 	},
 
 	renderResult(
@@ -528,11 +529,6 @@ function renderSingleFileResult(
 		"";
 	const op = args?.op || firstEdit?.op || details?.op;
 	const rename = args?.rename || firstEdit?.rename || firstEdit?.move || details?.move;
-	const { language } = formatEditDescription(rawPath, uiTheme, { rename });
-
-	const editTextSource = args?.newText ?? args?.oldText ?? args?.diff ?? args?.patch;
-	const metadataLineCount = editTextSource ? countLines(editTextSource) : null;
-	const metadataLine = op !== "delete" ? `\n${formatMetadataLine(metadataLineCount, language, uiTheme)}` : "";
 
 	const displayErrorText = isError && details && "displayErrorText" in details ? details.displayErrorText : undefined;
 	const errorText = isError
@@ -541,61 +537,57 @@ function renderSingleFileResult(
 			(result.content?.find(c => c.type === "text")?.text ?? "")
 		: "";
 
-	let cached: RenderCache | undefined;
+	return framedBlock(uiTheme, width => {
+		const { expanded, renderContext } = options;
+		const editDiffPreview = renderContext?.editDiffPreview;
+		const renderDiffFn = renderContext?.renderDiff ?? ((t: string) => t);
 
-	return {
-		render(width) {
-			const { expanded, renderContext } = options;
-			const editDiffPreview = renderContext?.editDiffPreview;
-			const renderDiffFn = renderContext?.renderDiff ?? ((t: string) => t);
-			const key = new Hasher().bool(expanded).u32(width).digest();
-			if (cached?.key === key) return cached.lines;
+		const firstChangedLine =
+			(editDiffPreview && "firstChangedLine" in editDiffPreview ? editDiffPreview.firstChangedLine : undefined) ||
+			(details && !isError ? details.firstChangedLine : undefined);
+		const linkPath = details && "path" in details ? details.path : undefined;
+		const { description } = formatEditDescription(rawPath, uiTheme, { rename, firstChangedLine, linkPath });
 
-			const firstChangedLine =
-				(editDiffPreview && "firstChangedLine" in editDiffPreview ? editDiffPreview.firstChangedLine : undefined) ||
-				(details && !isError ? details.firstChangedLine : undefined);
-			const { description } = formatEditDescription(rawPath, uiTheme, { rename, firstChangedLine });
+		// Change stats ride inline on the header bar next to the path.
+		const previewDiff = editDiffPreview && !("error" in editDiffPreview) ? editDiffPreview.diff : undefined;
+		const headerDiff = isError ? undefined : details?.diff || previewDiff;
+		const statsSuffix = headerDiff ? formatDiffStatsSuffix(headerDiff, uiTheme) : "";
+		const header =
+			renderStatusLine({ icon: isError ? "error" : "success", title: getOperationTitle(op), description }, uiTheme) +
+			statsSuffix;
 
-			const header = renderStatusLine(
-				{
-					icon: isError ? "error" : "success",
-					title: getOperationTitle(op),
-					description,
-				},
-				uiTheme,
+		let body = "";
+		if (isError) {
+			if (errorText) body = uiTheme.fg("error", replaceTabs(errorText, rawPath));
+		} else if (details?.diff) {
+			body = renderDiffSection(details.diff, rawPath, expanded, uiTheme, renderDiffFn);
+		} else if (editDiffPreview) {
+			if ("error" in editDiffPreview) body = uiTheme.fg("error", replaceTabs(editDiffPreview.error, rawPath));
+			else if (editDiffPreview.diff)
+				body = renderDiffSection(editDiffPreview.diff, rawPath, expanded, uiTheme, renderDiffFn);
+		}
+		if (details?.diagnostics) {
+			body += formatDiagnostics(details.diagnostics, expanded, uiTheme, (fp: string) =>
+				uiTheme.getLangIcon(getLanguageFromPath(fp)),
 			);
-			let text = header;
-			text += metadataLine;
+		}
 
-			if (isError) {
-				if (errorText) {
-					text += `\n\n${uiTheme.fg("error", replaceTabs(errorText, rawPath))}`;
-				}
-			} else if (details?.diff) {
-				text += renderDiffSection(details.diff, rawPath, expanded, uiTheme, renderDiffFn);
-			} else if (editDiffPreview) {
-				if ("error" in editDiffPreview) {
-					text += `\n\n${uiTheme.fg("error", replaceTabs(editDiffPreview.error, rawPath))}`;
-				} else if (editDiffPreview.diff) {
-					text += renderDiffSection(editDiffPreview.diff, rawPath, expanded, uiTheme, renderDiffFn);
-				}
-			}
+		// Diff lines self-wrap with a continuation gutter; pre-wrap to the frame's
+		// inner width so renderOutputBlock's generic wrap is a no-op. Edit frames
+		// use a flush left border because code-frame gutters already provide padding.
+		const innerWidth = Math.max(1, width - 2);
+		const bodyLines = body.length > 0 ? body.split("\n").flatMap(line => wrapEditRendererLine(line, innerWidth)) : [];
+		while (bodyLines.length > 0 && bodyLines[0].trim() === "") bodyLines.shift();
 
-			if (details?.diagnostics) {
-				text += formatDiagnostics(details.diagnostics, expanded, uiTheme, (fp: string) =>
-					uiTheme.getLangIcon(getLanguageFromPath(fp)),
-				);
-			}
-
-			const lines =
-				width > 0 ? text.split("\n").flatMap(line => wrapEditRendererLine(line, width)) : text.split("\n");
-			cached = { key, lines };
-			return lines;
-		},
-		invalidate() {
-			cached = undefined;
-		},
-	};
+		return {
+			header,
+			sections: bodyLines.length > 0 ? [{ lines: bodyLines }] : [],
+			state: isError ? "error" : options.isPartial ? "pending" : "success",
+			borderColor: isError ? "error" : "borderMuted",
+			width,
+			contentPaddingLeft: 0,
+		};
+	});
 }
 
 function renderMultiFileResult(
@@ -650,7 +642,7 @@ function renderMultiFileResult(
 		},
 		invalidate() {
 			cached = undefined;
-			for (const c of fileComponents) c.invalidate();
+			for (const c of fileComponents) c.invalidate?.();
 		},
 	};
 }
diff --git a/packages/coding-agent/src/edit/streaming.ts b/packages/coding-agent/src/edit/streaming.ts
index b411ed70b..718d164db 100644
--- a/packages/coding-agent/src/edit/streaming.ts
+++ b/packages/coding-agent/src/edit/streaming.ts
@@ -424,7 +424,7 @@ const hashlineStrategy: EditStreamingStrategy<HashlineArgs> = {
 		return previews.length > 0 ? previews : null;
 	},
 	renderStreamingFallback() {
-		// Never leak raw hashline syntax (`64:`, `|payload`, `¶path#hash`)
+		// Never leak raw hashline syntax (`64:`, `|payload`, `[path#hash]`)
 		// to the user — the streaming preview already projects every
 		// parseable op onto the real file via applyPartialTo, and an
 		// unparseable trailing chunk renders as "no preview yet" rather
diff --git a/packages/coding-agent/src/eval/__tests__/agent-bridge.test.ts b/packages/coding-agent/src/eval/__tests__/agent-bridge.test.ts
index b33e43727..838ed6c1a 100644
--- a/packages/coding-agent/src/eval/__tests__/agent-bridge.test.ts
+++ b/packages/coding-agent/src/eval/__tests__/agent-bridge.test.ts
@@ -10,7 +10,7 @@ import { AgentOutputManager } from "../../task/output-manager";
 import type { AgentDefinition, AgentProgress, SingleResult } from "../../task/types";
 import type { ToolSession } from "../../tools";
 import { EVAL_AGENT_MAX_DEPTH, runEvalAgent } from "../agent-bridge";
-import { EVAL_HEARTBEAT_OP, setBridgeHeartbeatIntervalMs } from "../heartbeat";
+import { EVAL_TIMEOUT_PAUSE_OP, EVAL_TIMEOUT_RESUME_OP } from "../bridge-timeout";
 import { IdleTimeout } from "../idle-timeout";
 import { disposeAllVmContexts } from "../js/context-manager";
 import { executeJs } from "../js/executor";
@@ -231,12 +231,62 @@ describe("runEvalAgent", () => {
 		});
 		await expect(runEvalAgent({ prompt: "fail" }, { session: makeSession() })).rejects.toThrow("boom");
 	});
+
+	// Regression: a runtime-limit abort returns exitCode=1, stderr="", error=undefined,
+	// aborted=true, abortReason="Subagent runtime limit exceeded (...)". The previous
+	// failure-message coalesce stopped at the empty `stderr` (since `??` only skips
+	// nullish values) and shipped an empty error through the bridge — Python then
+	// surfaced the generic `bridge call '__agent__' failed`. See #2006.
+	it("surfaces abortReason for aborts that leave stderr empty", async () => {
+		mockAgents();
+		const runSpy = vi.spyOn(taskExecutor, "runSubprocess");
+		runSpy.mockImplementationOnce(async options =>
+			singleResult(options, {
+				exitCode: 1,
+				output: "",
+				stderr: "",
+				error: undefined,
+				aborted: true,
+				abortReason: "Subagent runtime limit exceeded (task.maxRuntimeMs=900000)",
+			}),
+		);
+		runSpy.mockImplementationOnce(async options =>
+			singleResult(options, {
+				exitCode: 1,
+				output: "",
+				stderr: "   ",
+				error: "   ",
+				aborted: true,
+				abortReason: "Cancelled by caller",
+			}),
+		);
+		runSpy.mockImplementationOnce(async options =>
+			singleResult(options, {
+				exitCode: 1,
+				output: "",
+				stderr: "",
+				error: undefined,
+			}),
+		);
+
+		await expect(runEvalAgent({ prompt: "slow" }, { session: makeSession() })).rejects.toThrow(
+			"Subagent runtime limit exceeded (task.maxRuntimeMs=900000)",
+		);
+		// Whitespace-only stderr/error must not mask abortReason either.
+		await expect(runEvalAgent({ prompt: "cancelled" }, { session: makeSession() })).rejects.toThrow(
+			"Cancelled by caller",
+		);
+		// Last resort: still produce a non-empty message even when nothing useful is set,
+		// so Python never falls back to `bridge call '__agent__' failed`.
+		await expect(runEvalAgent({ prompt: "blank" }, { session: makeSession() })).rejects.toThrow(
+			"agent() subagent 'task' failed.",
+		);
+	});
 });
 
 describe("agent() through eval runtimes", () => {
 	afterEach(() => {
 		vi.restoreAllMocks();
-		setBridgeHeartbeatIntervalMs();
 	});
 
 	afterAll(async () => {
@@ -327,18 +377,6 @@ describe("agent() through eval runtimes", () => {
 			singleResult(options, { output: "hello from python" }),
 		);
 
-		const probe = await executePython('print("probe")', {
-			cwd: tempDir.path(),
-			sessionId: `${sessionId}:probe`,
-			sessionFile,
-			kernelMode: "per-call",
-		});
-		if (probe.exitCode === undefined && probe.cancelled) {
-			expect(probe.output).toBe("");
-			return;
-		}
-		expect(probe.exitCode).toBe(0);
-
 		const result = await executePython('print(agent("hi"))', {
 			cwd: tempDir.path(),
 			sessionId,
@@ -346,6 +384,10 @@ describe("agent() through eval runtimes", () => {
 			kernelMode: "per-call",
 			toolSession: session,
 		});
+		if (result.exitCode === undefined && result.cancelled) {
+			expect(result.output).toBe("");
+			return; // kernel unavailable in this environment
+		}
 
 		expect(result.exitCode).toBe(0);
 		expect(result.output.trim()).toBe("hello from python");
@@ -374,22 +416,14 @@ describe("agent() through eval runtimes", () => {
 			}
 		});
 
-		const probe = await executePython('print("probe")', {
-			cwd: tempDir.path(),
-			sessionId: `${sessionId}:probe`,
-			sessionFile,
-			kernelMode: "per-call",
-		});
-		if (probe.exitCode === undefined && probe.cancelled) {
-			expect(probe.output).toBe("");
-			return;
-		}
-		expect(probe.exitCode).toBe(0);
-
 		const result = await executePython(
 			'import json\nprint(json.dumps(parallel([lambda n=n: agent(n) for n in ["a", "b", "c", "d"]])))',
 			{ cwd: tempDir.path(), sessionId, sessionFile, kernelMode: "per-call", toolSession: session },
 		);
+		if (result.exitCode === undefined && result.cancelled) {
+			expect(result.output).toBe("");
+			return; // kernel unavailable in this environment
+		}
 
 		expect(result.exitCode).toBe(0);
 		expect(JSON.parse(result.output.trim())).toEqual(["a", "b", "c", "d"]);
@@ -413,7 +447,14 @@ describe("agent() through eval runtimes", () => {
 		// The host must respond the instant the cell aborts so the kernel can
 		// unwind via KeyboardInterrupt instead of being hard-killed (which used to
 		// surface "[kernel] Python kernel shutdown" and lose all session state).
+		let inFlight = 0;
+		let markSaturated: (() => void) | undefined;
+		const saturated = new Promise<void>(resolve => {
+			markSaturated = resolve;
+		});
 		vi.spyOn(taskExecutor, "runSubprocess").mockImplementation(async options => {
+			// task.maxConcurrency=6 → six bridge calls block at once; signal then.
+			if (++inFlight >= 6) markSaturated?.();
 			await Bun.sleep(9000); // deliberately ignores options.signal
 			return singleResult(options, { output: options.assignment ?? "" });
 		});
@@ -433,8 +474,9 @@ describe("agent() through eval runtimes", () => {
 		expect(seed.exitCode).toBe(0);
 
 		const ac = new AbortController();
-		// Abort ~1s in, after the worker threads are blocked in their bridge calls.
-		setTimeout(() => ac.abort(new Error("external interrupt")), 1000);
+		// Abort the instant all six worker threads are confirmed blocked in their
+		// bridge calls (condition-driven) instead of waiting a fixed wall second.
+		void saturated.then(() => ac.abort(new Error("external interrupt")));
 
 		const start = Date.now();
 		const result = await executePython(
@@ -560,52 +602,52 @@ describe("agent() through eval runtimes", () => {
 		expect(displayAgentEvents.length).toBe(2);
 	});
 
-	it("keeps the idle watchdog armed while a quiet agent() runs past the budget", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-agent-heartbeat-");
-		const { session } = makeEvalSession(tempDir, "js-agent-heartbeat");
+	it("pauses the idle watchdog while a quiet agent() runs past the budget", async () => {
+		using tempDir = TempDir.createSync("@omp-eval-agent-timeout-pause-");
+		const { session } = makeEvalSession(tempDir, "js-agent-timeout-pause");
 		mockAgents();
-		// Heartbeat cadence well under the idle budget so a working-but-silent
-		// subagent re-arms the watchdog several times before it could expire.
-		setBridgeHeartbeatIntervalMs(15);
 
-		// runSubprocess runs far past the budget and emits NO progress of its own
-		// — the only thing standing between the subagent and a spurious idle abort
-		// is the heartbeat keepalive the bridge pumps while it awaits.
+		// runSubprocess runs far past the eval timeout budget and emits NO progress
+		// of its own. The bridge pause must make that delegated time invisible to
+		// the watchdog.
 		vi.spyOn(taskExecutor, "runSubprocess").mockImplementation(async options => {
-			await Bun.sleep(200);
+			await Bun.sleep(40);
 			return singleResult(options, { output: "done" });
 		});
 
-		// Mirror the eval tool's wiring: an IdleTimeout drives cancellation and
-		// ONLY a bridge heartbeat re-arms it.
-		using idle = new IdleTimeout(60);
+		const ops: string[] = [];
+		using idle = new IdleTimeout(20);
 		const result = await runEvalAgent(
 			{ prompt: "investigate" },
 			{
 				session,
 				signal: idle.signal,
 				emitStatus: event => {
-					if (event.op === EVAL_HEARTBEAT_OP) idle.bump();
+					ops.push(event.op);
+					if (event.op === EVAL_TIMEOUT_PAUSE_OP) idle.pause();
+					if (event.op === EVAL_TIMEOUT_RESUME_OP) idle.resume();
 				},
 			},
 		);
 
-		expect(idle.signal.aborted).toBe(false);
 		expect(result.text).toBe("done");
+		expect(ops).toEqual([EVAL_TIMEOUT_PAUSE_OP, EVAL_TIMEOUT_RESUME_OP]);
+		expect(idle.signal.aborted).toBe(false);
+
+		await Bun.sleep(60);
+		expect(idle.signal.aborted).toBe(true);
 	});
 
-	it("does not let agent() progress snapshots re-arm the watchdog without a heartbeat", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-agent-progress-no-rearm-");
-		const { session } = makeEvalSession(tempDir, "js-agent-progress-no-rearm");
+	it("keeps timeout paused despite agent() progress snapshots", async () => {
+		using tempDir = TempDir.createSync("@omp-eval-agent-progress-timeout-pause-");
+		const { session } = makeEvalSession(tempDir, "js-agent-progress-timeout-pause");
 		mockAgents();
-		// Heartbeat slower than the budget: only the immediate beat at call start
-		// fires, so after the budget elapses nothing re-arms the watchdog.
-		setBridgeHeartbeatIntervalMs(10_000);
 
 		// Stream frequent progress snapshots (op:"agent") for well past the budget.
-		// Progress is rendered but MUST NOT count as activity — only heartbeats do.
+		// They render as status, but timeout accounting is controlled only by the
+		// bridge pause/resume events.
 		vi.spyOn(taskExecutor, "runSubprocess").mockImplementation(async options => {
-			for (let i = 0; i < 40; i++) {
+			for (let i = 0; i < 20; i++) {
 				options.onProgress?.({
 					index: options.index,
 					id: options.id,
@@ -622,28 +664,30 @@ describe("agent() through eval runtimes", () => {
 					cost: 0,
 					durationMs: i * 10,
 				});
-				await Bun.sleep(10);
+				await Bun.sleep(5);
 			}
 			return singleResult(options, { output: "done" });
 		});
 
 		const ops: string[] = [];
-		using idle = new IdleTimeout(80);
-		await runEvalAgent(
+		using idle = new IdleTimeout(40);
+		const result = await runEvalAgent(
 			{ prompt: "investigate" },
 			{
 				session,
 				signal: idle.signal,
 				emitStatus: event => {
 					ops.push(event.op);
-					if (event.op === EVAL_HEARTBEAT_OP) idle.bump();
+					if (event.op === EVAL_TIMEOUT_PAUSE_OP) idle.pause();
+					if (event.op === EVAL_TIMEOUT_RESUME_OP) idle.resume();
 				},
 			},
 		);
 
-		// Progress streamed, but the watchdog still fired: agent snapshots never
-		// re-armed it, and the lone start heartbeat lapsed before the call ended.
+		expect(result.text).toBe("done");
+		expect(ops[0]).toBe(EVAL_TIMEOUT_PAUSE_OP);
 		expect(ops).toContain("agent");
-		expect(idle.signal.aborted).toBe(true);
+		expect(ops.at(-1)).toBe(EVAL_TIMEOUT_RESUME_OP);
+		expect(idle.signal.aborted).toBe(false);
 	});
 });
diff --git a/packages/coding-agent/src/eval/__tests__/bridge-timeout.test.ts b/packages/coding-agent/src/eval/__tests__/bridge-timeout.test.ts
new file mode 100644
index 000000000..8cde55110
--- /dev/null
+++ b/packages/coding-agent/src/eval/__tests__/bridge-timeout.test.ts
@@ -0,0 +1,64 @@
+import { describe, expect, it } from "bun:test";
+import {
+	EVAL_TIMEOUT_PAUSE_OP,
+	EVAL_TIMEOUT_RESUME_OP,
+	isEvalTimeoutControlEvent,
+	withBridgeTimeoutPause,
+} from "../bridge-timeout";
+import type { JsStatusEvent } from "../js/shared/types";
+
+describe("withBridgeTimeoutPause", () => {
+	it("emits one pause before the operation and one resume after it settles", async () => {
+		const events: JsStatusEvent[] = [];
+
+		const value = await withBridgeTimeoutPause(
+			event => events.push(event),
+			async () => {
+				await Bun.sleep(80);
+				return "done";
+			},
+		);
+
+		expect(value).toBe("done");
+		expect(events.map(event => event.op)).toEqual([EVAL_TIMEOUT_PAUSE_OP, EVAL_TIMEOUT_RESUME_OP]);
+
+		const settledCount = events.length;
+		await Bun.sleep(40);
+		expect(events.length).toBe(settledCount);
+	});
+
+	it("resumes timeout accounting even when the operation throws", async () => {
+		const events: JsStatusEvent[] = [];
+
+		await expect(
+			withBridgeTimeoutPause(
+				event => events.push(event),
+				async () => {
+					await Bun.sleep(20);
+					throw new Error("boom");
+				},
+			),
+		).rejects.toThrow("boom");
+
+		expect(events.map(event => event.op)).toEqual([EVAL_TIMEOUT_PAUSE_OP, EVAL_TIMEOUT_RESUME_OP]);
+	});
+
+	it("runs the operation without emitting when no status sink is wired", async () => {
+		let ran = 0;
+
+		const value = await withBridgeTimeoutPause(undefined, async () => {
+			ran++;
+			await Bun.sleep(20);
+			return 42;
+		});
+
+		expect(value).toBe(42);
+		expect(ran).toBe(1);
+	});
+
+	it("identifies timeout-control events as non-renderable status", () => {
+		expect(isEvalTimeoutControlEvent({ op: EVAL_TIMEOUT_PAUSE_OP })).toBe(true);
+		expect(isEvalTimeoutControlEvent({ op: EVAL_TIMEOUT_RESUME_OP })).toBe(true);
+		expect(isEvalTimeoutControlEvent({ op: "agent", id: "subagent-1" })).toBe(false);
+	});
+});
diff --git a/packages/coding-agent/src/eval/__tests__/heartbeat.test.ts b/packages/coding-agent/src/eval/__tests__/heartbeat.test.ts
deleted file mode 100644
index 4daac8aee..000000000
--- a/packages/coding-agent/src/eval/__tests__/heartbeat.test.ts
+++ /dev/null
@@ -1,84 +0,0 @@
-import { afterEach, describe, expect, it } from "bun:test";
-import { EVAL_HEARTBEAT_OP, setBridgeHeartbeatIntervalMs, withBridgeHeartbeat } from "../heartbeat";
-import type { JsStatusEvent } from "../js/shared/types";
-
-describe("withBridgeHeartbeat", () => {
-	afterEach(() => {
-		setBridgeHeartbeatIntervalMs();
-	});
-
-	it("pumps heartbeat events on cadence while the operation is pending, then stops", async () => {
-		setBridgeHeartbeatIntervalMs(20);
-		const events: JsStatusEvent[] = [];
-
-		const value = await withBridgeHeartbeat(
-			event => events.push(event),
-			async () => {
-				await Bun.sleep(130);
-				return "done";
-			},
-		);
-
-		expect(value).toBe("done");
-		// ~6 ticks fit in 130ms at a 20ms cadence; assert it ticked repeatedly
-		// without pinning the exact count (scheduler jitter).
-		expect(events.length).toBeGreaterThanOrEqual(3);
-		expect(events.every(event => event.op === EVAL_HEARTBEAT_OP)).toBe(true);
-
-		// The interval is cleared once the operation settles: no further ticks.
-		const settledCount = events.length;
-		await Bun.sleep(80);
-		expect(events.length).toBe(settledCount);
-	});
-
-	it("emits a heartbeat immediately so a bridge call extends the budget at once", async () => {
-		// Interval far longer than the operation: the only beat that can fire is
-		// the immediate one at call start. It must still reach the sink.
-		setBridgeHeartbeatIntervalMs(10_000);
-		const events: JsStatusEvent[] = [];
-
-		await withBridgeHeartbeat(
-			event => events.push(event),
-			async () => {
-				await Bun.sleep(30);
-				return "done";
-			},
-		);
-
-		expect(events.length).toBe(1);
-		expect(events[0]?.op).toBe(EVAL_HEARTBEAT_OP);
-	});
-
-	it("runs the operation without emitting when no status sink is wired", async () => {
-		setBridgeHeartbeatIntervalMs(5);
-		let ran = 0;
-
-		const value = await withBridgeHeartbeat(undefined, async () => {
-			ran++;
-			await Bun.sleep(40);
-			return 42;
-		});
-
-		expect(value).toBe(42);
-		expect(ran).toBe(1);
-	});
-
-	it("clears the heartbeat even when the operation throws", async () => {
-		setBridgeHeartbeatIntervalMs(15);
-		const events: JsStatusEvent[] = [];
-
-		await expect(
-			withBridgeHeartbeat(
-				event => events.push(event),
-				async () => {
-					await Bun.sleep(60);
-					throw new Error("boom");
-				},
-			),
-		).rejects.toThrow("boom");
-
-		const afterThrow = events.length;
-		await Bun.sleep(60);
-		expect(events.length).toBe(afterThrow);
-	});
-});
diff --git a/packages/coding-agent/src/eval/__tests__/idle-timeout.test.ts b/packages/coding-agent/src/eval/__tests__/idle-timeout.test.ts
index 2aef57494..e4b7185da 100644
--- a/packages/coding-agent/src/eval/__tests__/idle-timeout.test.ts
+++ b/packages/coding-agent/src/eval/__tests__/idle-timeout.test.ts
@@ -32,21 +32,34 @@ describe("IdleTimeout", () => {
 		expect((idle.signal.reason as DOMException).name).toBe("TimeoutError");
 	});
 
-	it("re-arms on every bump and only fires after activity stops", async () => {
-		using idle = new IdleTimeout(150);
-		// Bump well past a single window; each bump must push the deadline forward
-		// so the watchdog never trips while activity continues.
-		for (let i = 0; i < 6; i++) {
-			await Bun.sleep(40);
-			idle.bump();
-		}
+	it("ignores elapsed time while paused and resumes with a fresh window", async () => {
+		using idle = new IdleTimeout(80);
+		idle.pause();
+		await Bun.sleep(160);
 		expect(idle.signal.aborted).toBe(false);
 
-		// Activity stopped — the watchdog should now fire within roughly one window.
-		const fired = await abortedWithin(idle.signal, 800);
+		idle.resume();
+		const firedEarly = await abortedWithin(idle.signal, 30);
+		expect(firedEarly).toBe(false);
+		const fired = await abortedWithin(idle.signal, 500);
 		expect(fired).toBe(true);
 	});
 
+	it("reference-counts overlapping pauses", async () => {
+		using idle = new IdleTimeout(60);
+		idle.pause();
+		idle.pause();
+		await Bun.sleep(120);
+		expect(idle.signal.aborted).toBe(false);
+
+		idle.resume();
+		await Bun.sleep(90);
+		expect(idle.signal.aborted).toBe(false);
+
+		idle.resume();
+		const fired = await abortedWithin(idle.signal, 500);
+		expect(fired).toBe(true);
+	});
 	it("never fires after dispose()", async () => {
 		const idle = new IdleTimeout(30);
 		idle.dispose();
@@ -55,12 +68,13 @@ describe("IdleTimeout", () => {
 		expect(idle.signal.aborted).toBe(false);
 	});
 
-	it("ignores bump() after the watchdog has already fired", async () => {
+	it("ignores pause/resume after the watchdog has already fired", async () => {
 		using idle = new IdleTimeout(30);
 		await abortedWithin(idle.signal, 500);
 		expect(idle.signal.aborted).toBe(true);
 		// Late activity must not un-abort or rearm a settled watchdog.
-		idle.bump();
+		idle.pause();
+		idle.resume();
 		expect(idle.signal.aborted).toBe(true);
 	});
 });
diff --git a/packages/coding-agent/src/eval/__tests__/kernel-spawn.test.ts b/packages/coding-agent/src/eval/__tests__/kernel-spawn.test.ts
new file mode 100644
index 000000000..0d7cdd4cc
--- /dev/null
+++ b/packages/coding-agent/src/eval/__tests__/kernel-spawn.test.ts
@@ -0,0 +1,103 @@
+import { afterEach, describe, expect, it } from "bun:test";
+import {
+	__resetWindowsConsoleProbeCache,
+	consoleAttachedViaTTY,
+	hostHasInheritableConsole,
+	shouldHideKernelWindow,
+} from "../py/spawn-options";
+
+/**
+ * `shouldHideKernelWindow` decides whether the long-lived Python kernel
+ * subprocess is spawned with `windowsHide: true`. On Windows, Bun maps that
+ * option to `CREATE_NO_WINDOW`, which detaches the child from any inherited
+ * console — breaking both (a) `LoadLibraryExW` for NumPy/pandas native
+ * extensions and (b) SIGINT delivery via `GenerateConsoleCtrlEvent`. See
+ * issue #1960. The tests below pin the three layered concerns the PR review
+ * surfaced:
+ *
+ * 1. `shouldHideKernelWindow` — pure predicate over a single boolean.
+ * 2. `consoleAttachedViaTTY` — the TTY-OR fallback used when the Win32 FFI
+ *    probe is unavailable; covers the partial-redirection cases.
+ * 3. `hostHasInheritableConsole` — the integration boundary. Off-Windows it
+ *    short-circuits to the TTY fallback; on Windows it is expected to
+ *    consult `kernel32!GetConsoleWindow()` first, which is the authoritative
+ *    signal even for the all-stdio-redirected case.
+ */
+describe("shouldHideKernelWindow", () => {
+	it("inherits the host console on Windows when one is attached", () => {
+		// Reporter's repro: omp launched in Windows Terminal, host has a
+		// console, kernel must inherit so `import pandas` doesn't deadlock in
+		// `_multiarray_umath` and SIGINT can recover the cell.
+		expect(shouldHideKernelWindow({ platform: "win32", hostHasInheritableConsole: true })).toBe(false);
+	});
+
+	it("hides on Windows only when the host has no console at all (true service / daemon)", () => {
+		// CREATE_NO_WINDOW here suppresses the console window Windows would
+		// otherwise auto-allocate for the console-app Python kernel.
+		expect(shouldHideKernelWindow({ platform: "win32", hostHasInheritableConsole: false })).toBe(true);
+	});
+
+	it("never sets windowsHide off-Windows (the option is a Win32-only flag)", () => {
+		// On POSIX `windowsHide` is a no-op; the predicate must return false
+		// everywhere off-Windows so the spawn site matches pre-fix behavior.
+		expect(shouldHideKernelWindow({ platform: "linux", hostHasInheritableConsole: true })).toBe(false);
+		expect(shouldHideKernelWindow({ platform: "linux", hostHasInheritableConsole: false })).toBe(false);
+		expect(shouldHideKernelWindow({ platform: "darwin", hostHasInheritableConsole: true })).toBe(false);
+		expect(shouldHideKernelWindow({ platform: "darwin", hostHasInheritableConsole: false })).toBe(false);
+	});
+});
+
+describe("consoleAttachedViaTTY (FFI fallback heuristic)", () => {
+	// The OR of three TTY signals correctly classifies the realistic shell
+	// redirection scenarios that motivated widening the check beyond stdout
+	// in the first review pass (PR #1961). The all-three-redirected case
+	// (false here) is the gap that the Win32 FFI probe in
+	// `hostHasInheritableConsole` is meant to close — this fallback is best-
+	// effort.
+
+	it("treats a fully interactive launch as console-attached", () => {
+		expect(consoleAttachedViaTTY({ stdinIsTTY: true, stdoutIsTTY: true, stderrIsTTY: true })).toBe(true);
+	});
+
+	it("treats `omp -p '...' > out.txt` (stdout-only redirect) as console-attached", () => {
+		// The reviewer's first-pass repro: stdout off the terminal, stdin
+		// and stderr still attached. OR keeps the console.
+		expect(consoleAttachedViaTTY({ stdinIsTTY: true, stdoutIsTTY: false, stderrIsTTY: true })).toBe(true);
+	});
+
+	it("treats stdin-only redirects (`< in.txt`) as console-attached", () => {
+		expect(consoleAttachedViaTTY({ stdinIsTTY: false, stdoutIsTTY: true, stderrIsTTY: true })).toBe(true);
+	});
+
+	it("treats stderr-only redirects (`2> err.log`) as console-attached", () => {
+		expect(consoleAttachedViaTTY({ stdinIsTTY: true, stdoutIsTTY: true, stderrIsTTY: false })).toBe(true);
+	});
+
+	it("returns false only when none of stdin/stdout/stderr is a TTY", () => {
+		// This is the gap: a real Windows Terminal session with all three
+		// streams redirected (`omp ... < in > out 2> err`) lands here.
+		// `hostHasInheritableConsole` uses the Win32 FFI probe to recover
+		// the right answer in that scenario; this helper is the fallback.
+		expect(consoleAttachedViaTTY({ stdinIsTTY: false, stdoutIsTTY: false, stderrIsTTY: false })).toBe(false);
+	});
+});
+
+describe("hostHasInheritableConsole", () => {
+	afterEach(() => {
+		__resetWindowsConsoleProbeCache();
+	});
+
+	if (process.platform !== "win32") {
+		it("matches the TTY-OR fallback off-Windows", () => {
+			// Off-Windows, `windowsHide` is a no-op anyway, but we still
+			// expose `hostHasInheritableConsole` symmetrically. Confirm it
+			// degrades to the same OR the call site would compute by hand.
+			const tty = consoleAttachedViaTTY({
+				stdinIsTTY: !!process.stdin.isTTY,
+				stdoutIsTTY: !!process.stdout.isTTY,
+				stderrIsTTY: !!process.stderr.isTTY,
+			});
+			expect(hostHasInheritableConsole()).toBe(tty);
+		});
+	}
+});
diff --git a/packages/coding-agent/src/eval/__tests__/llm-bridge.test.ts b/packages/coding-agent/src/eval/__tests__/llm-bridge.test.ts
index 2c0612333..67260c8f2 100644
--- a/packages/coding-agent/src/eval/__tests__/llm-bridge.test.ts
+++ b/packages/coding-agent/src/eval/__tests__/llm-bridge.test.ts
@@ -4,16 +4,17 @@ import type { Api, AssistantMessage, Model } from "@oh-my-pi/pi-ai";
 import * as ai from "@oh-my-pi/pi-ai";
 import { Effort } from "@oh-my-pi/pi-ai";
 import { TempDir } from "@oh-my-pi/pi-utils";
+import { $ } from "bun";
 import type { ModelRegistry } from "../../config/model-registry";
 import { Settings } from "../../config/settings";
 import type { ToolSession } from "../../tools";
 import { ToolError } from "../../tools/tool-errors";
-import { EVAL_HEARTBEAT_OP, setBridgeHeartbeatIntervalMs } from "../heartbeat";
+import { EVAL_TIMEOUT_PAUSE_OP, EVAL_TIMEOUT_RESUME_OP } from "../bridge-timeout";
 import { IdleTimeout } from "../idle-timeout";
 import { disposeAllVmContexts } from "../js/context-manager";
 import { executeJs } from "../js/executor";
 import { runEvalLlm } from "../llm-bridge";
-import { disposeAllKernelSessions, executePython } from "../py/executor";
+import { disposeAllKernelSessions, type PythonResult } from "../py/executor";
 
 function makeModel(provider: string, id: string, extra: Partial<Model<Api>> = {}): Model<Api> {
 	return {
@@ -57,6 +58,7 @@ function makeSession(opts: SessionOptions = {}): ToolSession {
 	const modelRegistry = {
 		getAvailable: () => opts.available ?? [SMOL, DEFAULT, SLOW],
 		getApiKey: async () => (opts.apiKey === undefined ? "test-key" : opts.apiKey),
+		resolver: () => async () => (opts.apiKey === undefined ? "test-key" : opts.apiKey),
 	} as unknown as ModelRegistry;
 	return {
 		settings,
@@ -96,10 +98,80 @@ function assistant(opts: {
 	};
 }
 
+async function runPythonLlmInSubprocess(options: { structured: boolean; tempDir: TempDir }): Promise<PythonResult> {
+	const repoRoot = path.resolve(import.meta.dir, "../../../..");
+	const scriptPath = path.join(options.tempDir.path(), "run-python-llm.ts");
+	const resultPath = path.join(options.tempDir.path(), "python-llm-result.json");
+	const aiPath = path.resolve(import.meta.dir, "../../../../ai/src/index.ts");
+	const executorPath = path.resolve(import.meta.dir, "../py/executor.ts");
+	const settingsPath = path.resolve(import.meta.dir, "../../config/settings.ts");
+	const code = options.structured
+		? 'import json\nprint(json.dumps(llm("hi", schema={"type": "object"})))'
+		: 'print(llm("hi", model="smol"))';
+	const responseContent = options.structured
+		? '[{ type: "toolCall", id: "tc-1", name: "respond", arguments: { ok: true } }]'
+		: '[{ type: "text", text: "hello from python" }]';
+	await Bun.write(
+		scriptPath,
+		`
+import { vi } from "bun:test";
+import * as ai from ${JSON.stringify(aiPath)};
+import { executePython } from ${JSON.stringify(executorPath)};
+import { Settings } from ${JSON.stringify(settingsPath)};
+
+const SMOL = {
+	id: "smol",
+	name: "smol",
+	api: "openai-responses",
+	provider: "p",
+	baseUrl: "https://example.test/v1",
+	reasoning: false,
+	input: ["text"],
+	cost: { input: 1, output: 1, cacheRead: 0, cacheWrite: 1 },
+	contextWindow: 128000,
+	maxTokens: 4096,
+};
+const settings = Settings.isolated({ "async.enabled": false, "task.isolation.mode": "none" });
+settings.setModelRole("smol", "p/smol");
+settings.setModelRole("slow", "p/slow");
+const session = {
+	settings,
+	modelRegistry: {
+		getAvailable: () => [SMOL],
+		getApiKey: async () => "test-key",
+		resolver: () => async () => "test-key",
+	},
+	getActiveModelString: () => "p/smol",
+};
+vi.spyOn(ai, "completeSimple").mockResolvedValue({
+	role: "assistant",
+	api: "openai-responses",
+	provider: "p",
+	model: "smol",
+	stopReason: "stop",
+	content: ${responseContent},
+});
+const result = await executePython(${JSON.stringify(code)}, {
+	cwd: ${JSON.stringify(options.tempDir.path())},
+	sessionId: ${JSON.stringify(`py-llm:${options.structured ? "struct" : "plain"}`)},
+	sessionFile: ${JSON.stringify(path.join(options.tempDir.path(), "session.jsonl"))},
+	toolSession: session,
+	kernelMode: "per-call",
+});
+await Bun.write(${JSON.stringify(resultPath)}, JSON.stringify(result));
+process.exit(0);
+`,
+	);
+	const child = await $`bun ${scriptPath}`.cwd(repoRoot).quiet().nothrow();
+	const stdout = child.stdout.toString();
+	const stderr = child.stderr.toString();
+	if (child.exitCode !== 0) throw new Error(stderr || stdout || `Python llm subprocess exited with ${child.exitCode}`);
+	return (await Bun.file(resultPath).json()) as PythonResult;
+}
+
 describe("runEvalLlm", () => {
 	afterEach(() => {
 		vi.restoreAllMocks();
-		setBridgeHeartbeatIntervalMs();
 	});
 
 	it("resolves each tier to its expected model", async () => {
@@ -217,31 +289,32 @@ describe("runEvalLlm", () => {
 		);
 	});
 
-	it("keeps the idle watchdog armed while a slow llm() request is in flight", async () => {
-		// A oneshot completion emits no status until it returns; a slow request
-		// must not look like a stalled cell. The bridge pumps a heartbeat while it
-		// awaits, re-arming the watchdog through emitStatus.
-		setBridgeHeartbeatIntervalMs(15);
+	it("pauses the idle watchdog while a slow llm() request is in flight", async () => {
+		// A oneshot completion emits no status until it returns; delegated model
+		// time must be invisible to the eval timeout budget.
 		vi.spyOn(ai, "completeSimple").mockImplementation(async () => {
 			await Bun.sleep(200);
 			return assistant({ text: "the answer" });
 		});
 
+		const ops: string[] = [];
 		using idle = new IdleTimeout(60);
 		const result = await runEvalLlm(
 			{ prompt: "q", model: "smol" },
 			{
 				session: makeSession(),
 				signal: idle.signal,
-				// Mirror the eval tool: only a bridge heartbeat re-arms the watchdog.
 				emitStatus: event => {
-					if (event.op === EVAL_HEARTBEAT_OP) idle.bump();
+					ops.push(event.op);
+					if (event.op === EVAL_TIMEOUT_PAUSE_OP) idle.pause();
+					if (event.op === EVAL_TIMEOUT_RESUME_OP) idle.resume();
 				},
 			},
 		);
 
-		expect(idle.signal.aborted).toBe(false);
 		expect(result.text).toBe("the answer");
+		expect(ops).toEqual([EVAL_TIMEOUT_PAUSE_OP, EVAL_TIMEOUT_RESUME_OP, "llm"]);
+		expect(idle.signal.aborted).toBe(false);
 	});
 });
 
@@ -290,38 +363,24 @@ describe("llm() through eval runtimes", () => {
 	});
 
 	it("exposes llm() in the Python runtime", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-llm-py-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-llm:${crypto.randomUUID()}`;
-		vi.spyOn(ai, "completeSimple").mockResolvedValue(assistant({ text: "hello from python" }));
-
-		const result = await executePython('print(llm("hi", model="smol"))', {
-			cwd: tempDir.path(),
-			sessionId,
-			sessionFile,
-			toolSession: makeSession(),
-		});
-
-		expect(result.exitCode).toBe(0);
-		expect(result.output.trim()).toBe("hello from python");
+		const tempDir = TempDir.createSync("@omp-eval-llm-py-");
+		try {
+			const result = await runPythonLlmInSubprocess({ structured: false, tempDir });
+			expect(result.exitCode).toBe(0);
+			expect(result.output.trim()).toBe("hello from python");
+		} finally {
+			tempDir.removeSync();
+		}
 	});
 
 	it("parses structured llm() output in the Python runtime", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-llm-py-struct-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-llm-struct:${crypto.randomUUID()}`;
-		vi.spyOn(ai, "completeSimple").mockResolvedValue(
-			assistant({ toolCall: { name: "respond", arguments: { ok: true } } }),
-		);
-
-		const result = await executePython('import json\nprint(json.dumps(llm("hi", schema={"type": "object"})))', {
-			cwd: tempDir.path(),
-			sessionId,
-			sessionFile,
-			toolSession: makeSession(),
-		});
-
-		expect(result.exitCode).toBe(0);
-		expect(JSON.parse(result.output.trim())).toEqual({ ok: true });
+		const tempDir = TempDir.createSync("@omp-eval-llm-py-struct-");
+		try {
+			const result = await runPythonLlmInSubprocess({ structured: true, tempDir });
+			expect(result.exitCode).toBe(0);
+			expect(JSON.parse(result.output.trim())).toEqual({ ok: true });
+		} finally {
+			tempDir.removeSync();
+		}
 	});
 });
diff --git a/packages/coding-agent/src/eval/__tests__/shared-executors.test.ts b/packages/coding-agent/src/eval/__tests__/shared-executors.test.ts
deleted file mode 100644
index 772a37c31..000000000
--- a/packages/coding-agent/src/eval/__tests__/shared-executors.test.ts
+++ /dev/null
@@ -1,609 +0,0 @@
-import { afterAll, afterEach, describe, expect, it, vi } from "bun:test";
-import * as fs from "node:fs/promises";
-import * as path from "node:path";
-import type { AssistantMessage } from "@oh-my-pi/pi-ai";
-import { TempDir } from "@oh-my-pi/pi-utils";
-import type { ModelRegistry } from "../../config/model-registry";
-import { Settings } from "../../config/settings";
-import type { LoadExtensionsResult } from "../../extensibility/extensions/types";
-import type { CreateAgentSessionOptions, CreateAgentSessionResult } from "../../sdk";
-import * as sdkModule from "../../sdk";
-import type { AgentSession, AgentSessionEvent, PromptOptions } from "../../session/agent-session";
-import { TaskTool } from "../../task";
-import * as discoveryModule from "../../task/discovery";
-import type { AgentDefinition, TaskParams } from "../../task/types";
-import type { ToolSession } from "../../tools";
-import { EventBus } from "../../utils/event-bus";
-import { disposeAllVmContexts } from "../js/context-manager";
-import { executeJs } from "../js/executor";
-import { disposeAllKernelSessions, executePython } from "../py/executor";
-
-function createToolSession(cwd: string, sessionFile: string | null, evalSessionId?: string): ToolSession {
-	const modelRegistry = {
-		authStorage: undefined,
-		refresh: async () => {},
-		getAvailable: () => [],
-		getApiKey: async () => null,
-	} as unknown as ModelRegistry;
-	return {
-		cwd,
-		hasUI: false,
-		settings: Settings.isolated({
-			"async.enabled": false,
-			"task.isolation.mode": "none",
-		}),
-		getSessionFile: () => sessionFile,
-		getSessionSpawns: () => "*",
-		getEvalSessionId: evalSessionId ? () => evalSessionId : undefined,
-		modelRegistry,
-	} as unknown as ToolSession;
-}
-
-function createBridgeToolSession(resultText: string, calls: unknown[]): ToolSession {
-	const readTool = {
-		name: "read",
-		label: "read",
-		description: "read",
-		parameters: { type: "object" },
-		async execute(_id: string, args: unknown) {
-			calls.push(args);
-			return { content: [{ type: "text" as const, text: resultText }] };
-		},
-	};
-	const tools = new Map<string, unknown>([["read", readTool]]);
-	return { getToolByName: (name: string) => tools.get(name) } as unknown as ToolSession;
-}
-
-function assistantStopMessage(text: string): AssistantMessage {
-	return {
-		role: "assistant",
-		content: text ? [{ type: "text", text }] : [],
-		api: "openai-responses",
-		provider: "openai",
-		model: "mock",
-		usage: {
-			input: 0,
-			output: 0,
-			cacheRead: 0,
-			cacheWrite: 0,
-			totalTokens: 0,
-			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
-		},
-		stopReason: "stop",
-		timestamp: Date.now(),
-	};
-}
-
-function createYieldingSubagentSession(onPrompt: () => Promise<void>): AgentSession {
-	const listeners: Array<(event: AgentSessionEvent) => void> = [];
-	const state = { messages: [] as AssistantMessage[] };
-	const emit = (event: AgentSessionEvent) => {
-		for (const listener of listeners) listener(event);
-	};
-	return {
-		state,
-		agent: { state: { systemPrompt: ["test"] } },
-		model: undefined,
-		extensionRunner: undefined,
-		sessionManager: {
-			appendSessionInit: () => {},
-		},
-		getActiveToolNames: () => ["eval", "yield"],
-		setActiveToolsByName: async () => {},
-		subscribe: (listener: (event: AgentSessionEvent) => void) => {
-			listeners.push(listener);
-			return () => {
-				const index = listeners.indexOf(listener);
-				if (index >= 0) listeners.splice(index, 1);
-			};
-		},
-		prompt: async (_text: string, _options?: PromptOptions) => {
-			await onPrompt();
-			state.messages.push(assistantStopMessage("done"));
-			emit({
-				type: "tool_execution_end",
-				toolCallId: "yield-call",
-				toolName: "yield",
-				result: {
-					content: [{ type: "text", text: "Result submitted." }],
-					details: { status: "success", data: { ok: true } },
-				},
-				isError: false,
-			});
-		},
-		waitForIdle: async () => {},
-		getLastAssistantMessage: () => state.messages[state.messages.length - 1],
-		abort: async () => {},
-		dispose: async () => {},
-	} as unknown as AgentSession;
-}
-
-const taskAgent: AgentDefinition = {
-	name: "task",
-	description: "Task agent",
-	systemPrompt: "Read eval state and yield.",
-	source: "bundled",
-	tools: ["eval", "yield"],
-};
-
-const taskParams: TaskParams = {
-	agent: "task",
-	tasks: [{ id: "ReadEval", description: "Read eval state", assignment: "Read parent eval state." }],
-};
-
-describe("shared eval executors", () => {
-	afterEach(() => {
-		vi.restoreAllMocks();
-	});
-
-	afterAll(async () => {
-		await disposeAllVmContexts();
-		await disposeAllKernelSessions();
-	});
-
-	it("shares JavaScript state across executeJs calls with one session id", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-js-shared-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `js-shared:${crypto.randomUUID()}`;
-		const session = createToolSession(tempDir.path(), sessionFile);
-
-		await executeJs("globalThis.x = 41;", { sessionId, session, sessionFile });
-		const result = await executeJs("return globalThis.x + 1;", { sessionId, session, sessionFile });
-
-		expect(result.exitCode).toBe(0);
-		expect(result.output.trim()).toBe("42");
-	});
-
-	it("treats idleTimeoutMs as an inactivity budget, not a fixed timer", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-js-idle-budget-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `js-idle-budget:${crypto.randomUUID()}`;
-		const session = createToolSession(tempDir.path(), sessionFile);
-
-		// With no wall-clock deadlineMs/timeoutMs and no aborting signal, a cell that
-		// runs well past idleTimeoutMs must still complete: the backend must never
-		// derive a competing fixed timer from the inactivity budget.
-		const result = await executeJs("await Bun.sleep(120); return 'done';", {
-			sessionId,
-			session,
-			sessionFile,
-			idleTimeoutMs: 30,
-		});
-
-		expect(result.cancelled).toBe(false);
-		expect(result.exitCode).toBe(0);
-		expect(result.output.trim()).toBe("done");
-	});
-
-	it("shares Python state across executePython calls with one session id", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-py-shared-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-shared:${crypto.randomUUID()}`;
-
-		await executePython("x = 41", { cwd: tempDir.path(), sessionId, sessionFile });
-		const result = await executePython("print(x + 1)", { cwd: tempDir.path(), sessionId, sessionFile });
-
-		expect(result.exitCode).toBe(0);
-		expect(result.output.trim()).toBe("42");
-	});
-
-	it("deduplicates concurrent first JavaScript session acquisition", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-js-cold-start-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `js-cold-start:${crypto.randomUUID()}`;
-		const session = createToolSession(tempDir.path(), sessionFile);
-
-		const [first, second] = await Promise.all([
-			executeJs(
-				"globalThis.sharedMarker ??= crypto.randomUUID(); await Bun.sleep(50); return globalThis.sharedMarker;",
-				{
-					sessionId,
-					session,
-					sessionFile,
-				},
-			),
-			executeJs("globalThis.sharedMarker ??= crypto.randomUUID(); return globalThis.sharedMarker;", {
-				sessionId,
-				session,
-				sessionFile,
-			}),
-		]);
-		const third = await executeJs("return globalThis.sharedMarker;", { sessionId, session, sessionFile });
-
-		expect(first.exitCode).toBe(0);
-		expect(second.exitCode).toBe(0);
-		expect(third.exitCode).toBe(0);
-		expect(first.output.trim()).toBe(second.output.trim());
-		expect(third.output.trim()).toBe(first.output.trim());
-	});
-
-	it("deduplicates concurrent first Python session acquisition", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-py-cold-start-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-cold-start:${crypto.randomUUID()}`;
-
-		const [first, second] = await Promise.all([
-			executePython(
-				`import asyncio, uuid
-shared_marker = globals().get("shared_marker") or str(uuid.uuid4())
-globals()["shared_marker"] = shared_marker
-await asyncio.sleep(0.05)
-print(shared_marker)`,
-				{ cwd: tempDir.path(), sessionId, sessionFile },
-			),
-			executePython(
-				`import uuid
-shared_marker = globals().get("shared_marker") or str(uuid.uuid4())
-globals()["shared_marker"] = shared_marker
-print(shared_marker)`,
-				{ cwd: tempDir.path(), sessionId, sessionFile },
-			),
-		]);
-		const third = await executePython("print(shared_marker)", { cwd: tempDir.path(), sessionId, sessionFile });
-
-		expect(first.exitCode).toBe(0);
-		expect(second.exitCode).toBe(0);
-		expect(third.exitCode).toBe(0);
-		expect(first.output.trim()).toBe(second.output.trim());
-		expect(third.output.trim()).toBe(first.output.trim());
-	});
-
-	it("splits retained Python kernels by cwd for one shared session id", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-py-cwd-");
-		const dirA = path.join(tempDir.path(), "a");
-		const dirB = path.join(tempDir.path(), "b");
-		await fs.mkdir(dirA);
-		await fs.mkdir(dirB);
-		const realDirA = await fs.realpath(dirA);
-		const realDirB = await fs.realpath(dirB);
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-cwd:${crypto.randomUUID()}`;
-
-		const first = await executePython(
-			`import os
-token = "from-a"
-print(os.getcwd())`,
-			{
-				cwd: dirA,
-				sessionId,
-				sessionFile,
-			},
-		);
-		const second = await executePython(
-			`import os
-print(os.getcwd())
-print("token" in globals())`,
-			{
-				cwd: dirB,
-				sessionId,
-				sessionFile,
-			},
-		);
-		const third = await executePython("print(token)", { cwd: dirA, sessionId, sessionFile });
-
-		expect(first.exitCode).toBe(0);
-		expect(first.output.trim()).toBe(realDirA);
-		expect(second.exitCode).toBe(0);
-		expect(second.output.trim().split("\n")).toEqual([realDirB, "False"]);
-		expect(third.exitCode).toBe(0);
-		expect(third.output.trim()).toBe("from-a");
-	});
-
-	it("interrupts timed out synchronous Python cells before they mutate shared state", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-py-sync-timeout-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-sync-timeout:${crypto.randomUUID()}`;
-
-		const timedOut = await executePython("import time\ntime.sleep(0.2)\nleaked_after_timeout = True", {
-			cwd: tempDir.path(),
-			sessionId,
-			sessionFile,
-			timeoutMs: 20,
-		});
-		await Bun.sleep(250);
-		const probe = await executePython('print("leaked_after_timeout" in globals())', {
-			cwd: tempDir.path(),
-			sessionId,
-			sessionFile,
-		});
-
-		expect(timedOut.cancelled).toBe(true);
-		expect(probe.exitCode).toBe(0);
-		expect(probe.output.trim()).toBe("False");
-	});
-
-	it("settles Python cells that raise SystemExit", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-py-system-exit-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-system-exit:${crypto.randomUUID()}`;
-
-		const result = await executePython('raise SystemExit("bye")', {
-			cwd: tempDir.path(),
-			sessionId,
-			sessionFile,
-			timeoutMs: 500,
-		});
-
-		expect(result.exitCode).toBe(1);
-		expect(result.output).toContain("SystemExit");
-		expect(result.output).toContain("bye");
-	});
-
-	it("lets a subagent inherit parent JavaScript and Python eval state", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-subagent-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const evalSessionId = `session:${sessionFile}:cwd:${tempDir.path()}`;
-		const parentSession = createToolSession(tempDir.path(), sessionFile, evalSessionId);
-		let seenJs = "";
-		let seenPy = "";
-		let capturedOptions: CreateAgentSessionOptions | undefined;
-
-		await executeJs('globalThis.parentSecret = "hello-js";', {
-			sessionId: `js:${evalSessionId}`,
-			session: parentSession,
-			sessionFile,
-		});
-		await executePython('parent_secret = "hello-py"', {
-			cwd: tempDir.path(),
-			sessionId: `python:${evalSessionId}`,
-			sessionFile,
-		});
-
-		vi.spyOn(discoveryModule, "discoverAgents").mockResolvedValue({ agents: [taskAgent], projectAgentsDir: null });
-		vi.spyOn(sdkModule, "createAgentSession").mockImplementation(async (options = {}) => {
-			capturedOptions = options;
-			const inherited = options.parentEvalSessionId;
-			if (!inherited) throw new Error("Missing parent eval session id");
-			return {
-				session: createYieldingSubagentSession(async () => {
-					const jsResult = await executeJs("return globalThis.parentSecret;", {
-						sessionId: `js:${inherited}`,
-						session: parentSession,
-						sessionFile,
-					});
-					const pyResult = await executePython("print(parent_secret)", {
-						cwd: tempDir.path(),
-						sessionId: `python:${inherited}`,
-						sessionFile,
-					});
-					seenJs = jsResult.output.trim();
-					seenPy = pyResult.output.trim();
-				}),
-				extensionsResult: {} as unknown as LoadExtensionsResult,
-				setToolUIContext: () => {},
-				eventBus: new EventBus(),
-			} satisfies CreateAgentSessionResult;
-		});
-
-		const tool = await TaskTool.create(parentSession);
-		await tool.execute("tool-call", taskParams);
-
-		expect(capturedOptions?.parentEvalSessionId).toBe(evalSessionId);
-		expect(seenJs).toBe("hello-js");
-		expect(seenPy).toBe("hello-py");
-	});
-
-	it("routes interleaved JavaScript display output to the matching run", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-js-interleave-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `js-interleave:${crypto.randomUUID()}`;
-		const session = createToolSession(tempDir.path(), sessionFile);
-
-		const first = executeJs('await Bun.sleep(80); display({ label: "A" });', {
-			sessionId,
-			session,
-			sessionFile,
-		});
-		await Bun.sleep(10);
-		const second = executeJs('display({ label: "B" });', {
-			sessionId,
-			session,
-			sessionFile,
-		});
-
-		const [firstResult, secondResult] = await Promise.all([first, second]);
-		expect(firstResult.exitCode).toBe(0);
-		expect(secondResult.exitCode).toBe(0);
-		expect(firstResult.displayOutputs).toEqual([{ type: "json", data: { label: "A" } }]);
-		expect(secondResult.displayOutputs).toEqual([{ type: "json", data: { label: "B" } }]);
-	});
-
-	it("routes interleaved Python display output to the matching run", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-py-interleave-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-interleave:${crypto.randomUUID()}`;
-
-		const first = executePython(
-			`import asyncio
-await asyncio.sleep(0.08)
-display({"label": "A"})`,
-			{
-				cwd: tempDir.path(),
-				sessionId,
-				sessionFile,
-			},
-		);
-		await Bun.sleep(10);
-		const second = executePython('display({"label": "B"})', {
-			cwd: tempDir.path(),
-			sessionId,
-			sessionFile,
-		});
-
-		const [firstResult, secondResult] = await Promise.all([first, second]);
-		expect(firstResult.exitCode).toBe(0);
-		expect(secondResult.exitCode).toBe(0);
-		expect(firstResult.displayOutputs).toEqual([{ type: "json", data: { label: "A" } }]);
-		expect(secondResult.displayOutputs).toEqual([{ type: "json", data: { label: "B" } }]);
-	});
-	it("preserves module-level singleton state across re-imports of an unchanged file", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-js-mtime-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `js-mtime:${crypto.randomUUID()}`;
-		const session = createToolSession(tempDir.path(), sessionFile);
-		const modulePath = path.join(tempDir.path(), "singleton.ts");
-		const moduleSpec = JSON.stringify(modulePath);
-		await Bun.write(
-			modulePath,
-			"let value = 0;\nexport function set(v) { value = v; }\nexport function get() { return value; }\n",
-		);
-
-		const initResult = await executeJs(`const mod = await import(${moduleSpec}); mod.set(42); return mod.get();`, {
-			sessionId,
-			session,
-			sessionFile,
-		});
-		expect(initResult.exitCode).toBe(0);
-		expect(initResult.output.trim()).toBe("42");
-
-		// Unchanged file: re-import must reuse the existing module namespace so the
-		// counter is still 42. This is the regression — the previous unconditional
-		// `delete require.cache[target]` reset singletons on every dynamic import.
-		const reuseResult = await executeJs(`const mod = await import(${moduleSpec}); return mod.get();`, {
-			sessionId,
-			session,
-			sessionFile,
-		});
-		expect(reuseResult.exitCode).toBe(0);
-		expect(reuseResult.output.trim()).toBe("42");
-
-		// Bump mtime by 5s to simulate an edit; the next import must evict the cache
-		// and re-evaluate the file, dropping the counter back to its initializer.
-		const future = new Date(Date.now() + 5_000);
-		await fs.utimes(modulePath, future, future);
-
-		const reloadResult = await executeJs(`const mod = await import(${moduleSpec}); return mod.get();`, {
-			sessionId,
-			session,
-			sessionFile,
-		});
-		expect(reloadResult.exitCode).toBe(0);
-		expect(reloadResult.output.trim()).toBe("0");
-	});
-
-	it("reloads a local re-export when a transitive dependency changes", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-js-transitive-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `js-transitive:${crypto.randomUUID()}`;
-		const session = createToolSession(tempDir.path(), sessionFile);
-		const leafPath = path.join(tempDir.path(), "leaf.ts");
-		const entryPath = path.join(tempDir.path(), "entry.ts");
-		const entrySpec = JSON.stringify(entryPath);
-		await Bun.write(leafPath, "export const value = 1;\n");
-		await Bun.write(entryPath, 'export { value } from "./leaf.ts";\n');
-
-		const initial = await executeJs(`const mod = await import(${entrySpec}); return mod.value;`, {
-			sessionId,
-			session,
-			sessionFile,
-		});
-		expect(initial.exitCode).toBe(0);
-		expect(initial.output.trim()).toBe("1");
-
-		await Bun.write(leafPath, "export const value = 2;\n");
-		const future = new Date(Date.now() + 5_000);
-		await fs.utimes(leafPath, future, future);
-
-		const reloaded = await executeJs(`const mod = await import(${entrySpec}); return mod.value;`, {
-			sessionId,
-			session,
-			sessionFile,
-		});
-		expect(reloaded.exitCode).toBe(0);
-		expect(reloaded.output.trim()).toBe("2");
-	});
-
-	it("links a cyclic local module graph without crashing", async () => {
-		// Regression: the loader used to link()+evaluate() each local module individually
-		// inside the recursive linker callback. On any import cycle that re-entered Bun's
-		// node:vm linker mid-instantiation and segfaulted the process (SIGTRAP,
-		// getImportedModule on a null record) — e.g. `await import("…/edit/streaming.ts")`,
-		// whose relative-import subtree is cyclic. The graph must now link in a single pass.
-		using tempDir = TempDir.createSync("@omp-eval-js-cycle-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `js-cycle:${crypto.randomUUID()}`;
-		const session = createToolSession(tempDir.path(), sessionFile);
-		const alphaPath = path.join(tempDir.path(), "alpha.ts");
-		const betaPath = path.join(tempDir.path(), "beta.ts");
-		const alphaSpec = JSON.stringify(alphaPath);
-		const betaSpec = JSON.stringify(betaPath);
-		await Bun.write(
-			alphaPath,
-			'import { betaName } from "./beta.ts";\nexport const alphaName = "alpha";\nexport function combined() { return alphaName + ":" + betaName; }\n',
-		);
-		await Bun.write(
-			betaPath,
-			'import { alphaName } from "./alpha.ts";\nexport const betaName = "beta";\nexport function viaAlpha() { return alphaName; }\n',
-		);
-
-		const result = await executeJs(
-			`const a = await import(${alphaSpec});\nconst b = await import(${betaSpec});\nreturn [a.combined(), b.viaAlpha()].join("|");`,
-			{ sessionId, session, sessionFile },
-		);
-
-		expect(result.exitCode).toBe(0);
-		expect(result.output.trim()).toBe("alpha:beta|alpha");
-	});
-
-	it("loads TypeScript type-only imports in cells and local modules", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-js-type-imports-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `js-type-imports:${crypto.randomUUID()}`;
-		const session = createToolSession(tempDir.path(), sessionFile);
-		const typesPath = path.join(tempDir.path(), "types.ts");
-		const valuesPath = path.join(tempDir.path(), "values.ts");
-		const entryPath = path.join(tempDir.path(), "entry.ts");
-		const typesSpec = JSON.stringify(typesPath);
-		const entrySpec = JSON.stringify(entryPath);
-		await Bun.write(typesPath, "export interface TypeOnly { value: number }\n");
-		await Bun.write(valuesPath, "export interface InlineOnly { value: number }\nexport const imported = 41;\n");
-		await Bun.write(
-			entryPath,
-			[
-				'import type { TypeOnly } from "./types.ts";',
-				'import { type InlineOnly, imported } from "./values.ts";',
-				"export const typeOnly = 1;",
-				"export const inlineType = imported;",
-				"",
-			].join("\n"),
-		);
-
-		const result = await executeJs(
-			`import type { TypeOnly } from ${typesSpec};\nconst mod = await import(${entrySpec});\nreturn mod.typeOnly + mod.inlineType;`,
-			{
-				sessionId,
-				session,
-				sessionFile,
-			},
-		);
-
-		expect(result.exitCode).toBe(0);
-		expect(result.output.trim()).toBe("42");
-	});
-
-	it("refreshes the Python tool proxy when bridge env appears after kernel warm-up", async () => {
-		using tempDir = TempDir.createSync("@omp-eval-py-tool-proxy-");
-		const sessionFile = path.join(tempDir.path(), "session.jsonl");
-		const sessionId = `py-tool-proxy:${crypto.randomUUID()}`;
-		const bridgeCalls: unknown[] = [];
-		const bridgeSession = createBridgeToolSession("bridge-ok", bridgeCalls);
-
-		const withoutBridge = await executePython(
-			'try:\n    print(tool.read({"path": "foo.txt"}))\nexcept Exception as exc:\n    print(type(exc).__name__)\n    print(str(exc))',
-			{ cwd: tempDir.path(), sessionId, sessionFile },
-		);
-		const withBridge = await executePython('print(tool.read({"path": "foo.txt"}))', {
-			cwd: tempDir.path(),
-			sessionId,
-			sessionFile,
-			toolSession: bridgeSession,
-		});
-
-		expect(withoutBridge.exitCode).toBe(0);
-		expect(withoutBridge.output).toContain("RuntimeError");
-		expect(withoutBridge.output).toContain("tool bridge is unavailable");
-		expect(withBridge.exitCode).toBe(0);
-		expect(withBridge.output.trim()).toBe("bridge-ok");
-		expect(bridgeCalls).toEqual([{ path: "foo.txt", _i: "py prelude" }]);
-	});
-});
diff --git a/packages/coding-agent/src/eval/agent-bridge.ts b/packages/coding-agent/src/eval/agent-bridge.ts
index 54dcecad1..d66601c22 100644
--- a/packages/coding-agent/src/eval/agent-bridge.ts
+++ b/packages/coding-agent/src/eval/agent-bridge.ts
@@ -13,10 +13,10 @@ import subagentUserPromptTemplate from "../prompts/system/subagent-user-prompt.m
 import * as taskDiscovery from "../task/discovery";
 import * as taskExecutor from "../task/executor";
 import { AgentOutputManager } from "../task/output-manager";
-import type { AgentDefinition, AgentProgress } from "../task/types";
+import type { AgentDefinition, AgentProgress, SingleResult } from "../task/types";
 import type { ToolSession } from "../tools";
 import { ToolError } from "../tools/tool-errors";
-import { withBridgeHeartbeat } from "./heartbeat";
+import { withBridgeTimeoutPause } from "./bridge-timeout";
 import type { JsStatusEvent } from "./js/shared/types";
 // Import review tools for side effects (registers subagent tool handlers).
 import "../tools/review";
@@ -173,6 +173,26 @@ function emitProgressStatus(emitStatus: ((event: JsStatusEvent) => void) | undef
 	});
 }
 
+/**
+ * Coalesce a subagent failure into a non-empty, human-meaningful error message.
+ *
+ * When the executor aborts a subagent (runtime limit, parent cancellation, …)
+ * the actionable explanation lives on `abortReason`, while `error`/`stderr`
+ * are routinely empty strings. Plain `??` coalescing stops at the empty string
+ * and ships an empty error through the bridge — Python then surfaces only the
+ * generic `bridge call '__agent__' failed`. See #2006.
+ */
+function buildSubagentFailureMessage(agentName: string, result: SingleResult): string {
+	const abortReason = trimToUndefined(result.abortReason);
+	if (result.aborted && abortReason) return abortReason;
+	return (
+		trimToUndefined(result.error) ??
+		trimToUndefined(result.stderr) ??
+		abortReason ??
+		`agent() subagent '${agentName}' failed.`
+	);
+}
+
 /**
  * Run a single subagent on behalf of an eval cell's `agent()` call.
  */
@@ -225,17 +245,15 @@ export async function runEvalAgent(args: unknown, options: EvalAgentBridgeOption
 		getSessionId: options.session.getSessionId ?? (() => null),
 	};
 	const parentArtifactManager = options.session.getArtifactManager?.() ?? undefined;
-	const parentEvalSessionId = options.session.getEvalSessionId?.() ?? undefined;
 	const mcpManager = options.session.mcpManager ?? MCPManager.instance();
 	const { sessionFile, artifactsDir, contextFile } = await getArtifacts(options.session);
 	const outputManager = getOutputManager(options.session);
 	const id = await outputManager.allocate(outputIdBase(parsed.label, agentName));
 	const assignment = parsed.prompt.trim();
 	const context = trimToUndefined(parsed.context);
-	// Pump a heartbeat while the subagent runs so the eval idle watchdog stays
-	// armed across quiet stretches (time-to-first-token, long nested tools)
-	// where `onProgress` would otherwise emit no status to re-arm it.
-	const result = await withBridgeHeartbeat(options.emitStatus, () =>
+	// Suspend eval timeout accounting while the subagent owns control. The
+	// timeout clock restarts once the bridge returns to the cell runtime.
+	const result = await withBridgeTimeoutPause(options.emitStatus, () =>
 		taskExecutor.runSubprocess({
 			cwd: options.session.cwd,
 			agent: effectiveAgent,
@@ -261,6 +279,12 @@ export async function runEvalAgent(args: unknown, options: EvalAgentBridgeOption
 			authStorage: options.session.authStorage,
 			modelRegistry: options.session.modelRegistry,
 			settings: options.session.settings,
+			// Eval `agent()` subagents are never wall-clock capped: the parent
+			// cell's idle watchdog is suspended for the whole bridge call
+			// (withBridgeTimeoutPause), so a long-running phase/recovery workflow
+			// must not be killed by `task.maxRuntimeMs`. Force the limit off
+			// regardless of the inherited session setting.
+			maxRuntimeMs: 0,
 			mcpManager,
 			contextFiles,
 			skills: availableSkills,
@@ -272,14 +296,16 @@ export async function runEvalAgent(args: unknown, options: EvalAgentBridgeOption
 			parentHindsightSessionState: options.session.getHindsightSessionState?.(),
 			parentMnemopiSessionState: options.session.getMnemopiSessionState?.(),
 			parentTelemetry: options.session.getTelemetry?.(),
-			parentEvalSessionId,
+			// Deliberately omit parentEvalSessionId: the parent's Python kernel is
+			// blocked on this bridge call, so sharing the eval session would deadlock
+			// (subagent queues behind the parent's in-flight execution, parent waits
+			// for subagent → circular). Each bridge-spawned subagent gets its own
+			// eval session with an independent kernel.
 		}),
 	);
 
-	if (result.exitCode !== 0 || result.error) {
-		const failureMessage =
-			result.error ?? result.stderr ?? result.abortReason ?? `agent() subagent '${agentName}' failed.`;
-		throw new ToolError(failureMessage);
+	if (result.exitCode !== 0 || result.error || result.aborted) {
+		throw new ToolError(buildSubagentFailureMessage(agentName, result));
 	}
 
 	options.session.recordEvalSubagentUsage?.(result.usage?.output ?? 0);
diff --git a/packages/coding-agent/src/eval/backend.ts b/packages/coding-agent/src/eval/backend.ts
index 0ae69a0c4..ed8a85f12 100644
--- a/packages/coding-agent/src/eval/backend.ts
+++ b/packages/coding-agent/src/eval/backend.ts
@@ -10,12 +10,12 @@ export interface ExecutorBackendExecOptions {
 	signal?: AbortSignal;
 	session: ToolSession;
 	/**
-	 * Inactivity budget in milliseconds (the cell's `timeout`). Cancellation is
-	 * driven entirely by `signal`, which the eval tool arms as an idle watchdog
-	 * that fires a `TimeoutError` reason after this much time with no progress
-	 * (status) events. Backends use this value only for timeout-annotation text
-	 * and as cold-start headroom; they MUST NOT derive a competing wall-clock
-	 * timer from it.
+	 * Runtime-work budget in milliseconds (the cell's `timeout`). Cancellation is
+	 * driven entirely by `signal`, which the eval tool arms as a watchdog that
+	 * pauses on bridge timeout-control status events and fires a `TimeoutError`
+	 * reason only while the Python/JS runtime owns control. Backends use this
+	 * value only for timeout-annotation text and as cold-start headroom; they MUST
+	 * NOT derive a competing wall-clock timer from it.
 	 */
 	idleTimeoutMs: number;
 	reset: boolean;
diff --git a/packages/coding-agent/src/eval/bridge-timeout.ts b/packages/coding-agent/src/eval/bridge-timeout.ts
new file mode 100644
index 000000000..bef0798cc
--- /dev/null
+++ b/packages/coding-agent/src/eval/bridge-timeout.ts
@@ -0,0 +1,44 @@
+/**
+ * Timeout suspension for in-flight host-side eval bridge calls.
+ *
+ * The eval watchdog caps a cell's `timeout` as a budget on the cell runtime's
+ * own work. Host-side `agent()` / `parallel()` / `llm()` bridge calls hand
+ * control to the outer TypeScript process, where the Python kernel or JS VM is
+ * only waiting for a result. While that delegated work is in flight, the cell
+ * timeout must be ignored completely; once the bridge returns and the runtime is
+ * back in control, the watchdog starts a fresh timeout window.
+ *
+ * Bridge helpers express that handoff with synthetic pause/resume status events
+ * on the existing `emitStatus → onStatus` path. Consumers MUST treat these as
+ * timeout-control events only: update the watchdog and drop them from rendered
+ * or persisted cell output.
+ */
+import type { JsStatusEvent } from "./js/shared/types";
+
+/** Synthetic status op emitted when a bridge call leaves the cell runtime. */
+export const EVAL_TIMEOUT_PAUSE_OP = "timeout-pause";
+
+/** Synthetic status op emitted when a bridge call returns control to the runtime. */
+export const EVAL_TIMEOUT_RESUME_OP = "timeout-resume";
+
+/** Whether a status event is pure eval-timeout control and should not render. */
+export function isEvalTimeoutControlEvent(event: JsStatusEvent): boolean {
+	return event.op === EVAL_TIMEOUT_PAUSE_OP || event.op === EVAL_TIMEOUT_RESUME_OP;
+}
+
+/**
+ * Run {@link operation} while suspending the eval watchdog through
+ * {@link emitStatus}. A no-op wrapper when no status sink is wired.
+ */
+export async function withBridgeTimeoutPause<T>(
+	emitStatus: ((event: JsStatusEvent) => void) | undefined,
+	operation: () => Promise<T>,
+): Promise<T> {
+	if (!emitStatus) return operation();
+	emitStatus({ op: EVAL_TIMEOUT_PAUSE_OP });
+	try {
+		return await operation();
+	} finally {
+		emitStatus({ op: EVAL_TIMEOUT_RESUME_OP });
+	}
+}
diff --git a/packages/coding-agent/src/eval/heartbeat.ts b/packages/coding-agent/src/eval/heartbeat.ts
deleted file mode 100644
index 295bcdf20..000000000
--- a/packages/coding-agent/src/eval/heartbeat.ts
+++ /dev/null
@@ -1,74 +0,0 @@
-/**
- * Keepalive for in-flight host-side eval bridge calls.
- *
- * The eval watchdog ({@link ../tools/eval IdleTimeout}) caps a cell's `timeout`
- * as a wall-clock budget on the cell's *own* work, but pauses that budget while
- * a host-side `agent()`/`parallel()` (via `runSubprocess`) or `llm()` (a single
- * completion) call is in flight. Those calls are the only thing that re-arms the
- * watchdog — and they can run for long stretches with **no** status of their own
- * (a subagent's time-to-first-token on a reasoning model, a long quiet nested
- * tool, or the entire body of a oneshot `llm()` call). Without a keepalive the
- * watchdog would mistake that delegated work for the cell stalling and abort it
- * mid-flight, killing the subagent.
- *
- * {@link withBridgeHeartbeat} bridges that gap by emitting a synthetic
- * {@link EVAL_HEARTBEAT_OP} status event immediately when the call begins and
- * then on a fixed cadence until it settles. The event rides the same
- * `emitStatus → onStatus` channel both runtimes already forward, so it re-arms
- * the watchdog without any new plumbing. The heartbeat is the *sole* signal that
- * extends the budget: consumers MUST treat it as a pure keepalive — bump the
- * watchdog and drop it (never persist or render it) — see the executor display
- * sinks and the eval tool's `onStatus` handler. Every other status event
- * (compute helpers, `log()`/`phase()`, tool results) counts against the budget.
- */
-import type { JsStatusEvent } from "./js/shared/types";
-
-/**
- * Synthetic status op emitted purely to keep the eval idle watchdog alive while
- * a host-side bridge call is in flight. Carries no payload.
- */
-export const EVAL_HEARTBEAT_OP = "heartbeat";
-
-/**
- * Heartbeat cadence. Comfortably below the default 30s idle budget (and the
- * larger budgets long fanouts run under), so a working bridge call always bumps
- * the watchdog before it expires, while a genuine stall is still bounded once
- * the call settles and the heartbeat stops.
- */
-const HEARTBEAT_INTERVAL_MS = 5_000;
-
-let heartbeatIntervalMs = HEARTBEAT_INTERVAL_MS;
-
-/**
- * Test seam: override the heartbeat cadence so integration tests can exercise
- * the keepalive within a sub-second idle budget. Pass no value to restore the
- * production default.
- */
-export function setBridgeHeartbeatIntervalMs(ms?: number): void {
-	heartbeatIntervalMs = ms === undefined ? HEARTBEAT_INTERVAL_MS : Math.max(1, Math.floor(ms));
-}
-
-/**
- * Run {@link operation}, pumping {@link EVAL_HEARTBEAT_OP} status events through
- * {@link emitStatus} — one immediately, then on a fixed cadence — until it
- * settles. The immediate beat pauses the watchdog the instant the call begins,
- * so a bridge call that starts close to the budget edge (after the cell already
- * spent most of it computing) is not aborted before the first interval tick. A
- * no-op wrapper when no `emitStatus` sink is wired (the heartbeat would reach
- * nobody).
- */
-export async function withBridgeHeartbeat<T>(
-	emitStatus: ((event: JsStatusEvent) => void) | undefined,
-	operation: () => Promise<T>,
-): Promise<T> {
-	if (!emitStatus) return operation();
-	emitStatus({ op: EVAL_HEARTBEAT_OP });
-	const timer = setInterval(() => emitStatus({ op: EVAL_HEARTBEAT_OP }), heartbeatIntervalMs);
-	// Never keep the event loop alive for the heartbeat alone.
-	timer.unref?.();
-	try {
-		return await operation();
-	} finally {
-		clearInterval(timer);
-	}
-}
diff --git a/packages/coding-agent/src/eval/idle-timeout.ts b/packages/coding-agent/src/eval/idle-timeout.ts
index 2fbba1de1..44c438a65 100644
--- a/packages/coding-agent/src/eval/idle-timeout.ts
+++ b/packages/coding-agent/src/eval/idle-timeout.ts
@@ -1,17 +1,15 @@
 /**
- * Inactivity watchdog for eval cells.
+ * Watchdog for eval cell work.
  *
- * A cell's `timeout` is treated as an *idle* budget rather than a hard
- * wall-clock deadline: the watchdog aborts {@link signal} (with a
- * `TimeoutError` reason, matching `AbortSignal.timeout`) only once `idleMs`
- * elapses with no {@link bump}. Every progress signal re-arms it, so a
- * long-running fanout that keeps reporting progress (e.g. `agent()` status
- * updates, `log()`/`phase()`) never trips the timeout, while a genuinely
- * stalled cell still gets interrupted.
+ * A cell's `timeout` bounds time while the Python kernel or JS VM is in control.
+ * Host-side bridge calls can {@link pause} the watchdog so delegated
+ * `agent()`/`parallel()`/`llm()` work is ignored completely, then {@link resume}
+ * starts a fresh timeout window once the runtime gets control back.
  *
- * The timer self-reschedules instead of being torn down and recreated on every
- * bump, so a high-frequency stream of bumps (sub-second agent progress) costs
- * one timestamp write per event rather than churning a timer each time.
+ * The active timer self-reschedules instead of being torn down on every
+ * activity event, so frequent activity costs one timestamp write per event.
+ * Pause is reference-counted because `parallel()` can have multiple bridge calls
+ * in flight at once.
  */
 export class IdleTimeout {
 	readonly #controller = new AbortController();
@@ -20,6 +18,7 @@ export class IdleTimeout {
 	#deadlineMs: number;
 	#timer: NodeJS.Timeout | undefined;
 	#settled = false;
+	#pauseDepth = 0;
 
 	constructor(idleMs: number) {
 		this.#idleMs = Math.max(1, Math.floor(idleMs));
@@ -27,21 +26,40 @@ export class IdleTimeout {
 		this.#arm(this.#idleMs);
 	}
 
-	/** Aborts with a `TimeoutError` reason once the inactivity budget is exhausted. */
+	/** Aborts with a `TimeoutError` reason once the active timeout window is exhausted. */
 	get signal(): AbortSignal {
 		return this.#controller.signal;
 	}
 
-	/** Configured inactivity budget in milliseconds. */
+	/** Configured active timeout window in milliseconds. */
 	get idleMs(): number {
 		return this.#idleMs;
 	}
 
-	/** Record activity, pushing the inactivity deadline forward by `idleMs`. */
+	/** Record runtime activity, pushing the active deadline forward by `idleMs`. */
 	bump(): void {
-		if (this.#settled) return;
+		if (this.#settled || this.#pauseDepth > 0) return;
 		this.#deadlineMs = Date.now() + this.#idleMs;
 	}
+	/** Suspend timeout accounting while control is delegated to host-side work. */
+	pause(): void {
+		if (this.#settled) return;
+		this.#pauseDepth++;
+		if (this.#pauseDepth !== 1) return;
+		if (this.#timer) {
+			clearTimeout(this.#timer);
+			this.#timer = undefined;
+		}
+	}
+
+	/** Resume timeout accounting with a fresh timeout window. */
+	resume(): void {
+		if (this.#settled || this.#pauseDepth === 0) return;
+		this.#pauseDepth--;
+		if (this.#pauseDepth > 0) return;
+		this.#deadlineMs = Date.now() + this.#idleMs;
+		this.#arm(this.#idleMs);
+	}
 
 	/** Stop the watchdog. Safe to call multiple times. */
 	dispose(): void {
@@ -65,7 +83,7 @@ export class IdleTimeout {
 	}
 
 	#onExpire(): void {
-		if (this.#settled) return;
+		if (this.#settled || this.#pauseDepth > 0) return;
 		const remainingMs = this.#deadlineMs - Date.now();
 		if (remainingMs > 0) {
 			// A bump moved the deadline forward after this timer was armed; wait
diff --git a/packages/coding-agent/src/eval/js/executor.ts b/packages/coding-agent/src/eval/js/executor.ts
index 2577b263d..430b38348 100644
--- a/packages/coding-agent/src/eval/js/executor.ts
+++ b/packages/coding-agent/src/eval/js/executor.ts
@@ -1,7 +1,7 @@
 import { DEFAULT_MAX_BYTES, OutputSink } from "../../session/streaming-output";
 import type { ToolSession } from "../../tools";
 import { resolveOutputMaxColumns, resolveOutputSinkHeadBytes } from "../../tools/output-meta";
-import { EVAL_HEARTBEAT_OP } from "../heartbeat";
+import { isEvalTimeoutControlEvent } from "../bridge-timeout";
 import { executeInVmContext, type JsDisplayOutput } from "./context-manager";
 import type { JsStatusEvent } from "./shared/types";
 
@@ -10,9 +10,9 @@ export interface JsExecutorOptions {
 	timeoutMs?: number;
 	deadlineMs?: number;
 	/**
-	 * Inactivity budget (ms). Used for worker cold-start headroom and
-	 * timeout-annotation text when the caller drives cancellation via an
-	 * idle-aware `signal` instead of `deadlineMs`/`timeoutMs`. Never arms a timer.
+	 * Runtime-work budget (ms). Used for worker cold-start headroom and
+	 * timeout-annotation text when the caller drives cancellation via the eval
+	 * watchdog `signal` instead of `deadlineMs`/`timeoutMs`. Never arms a timer.
 	 */
 	idleTimeoutMs?: number;
 	onChunk?: (chunk: string) => Promise<void> | void;
@@ -85,9 +85,9 @@ export async function executeJs(code: string, options: JsExecutorOptions): Promi
 		options.signal && timeoutSignal
 			? AbortSignal.any([options.signal, timeoutSignal])
 			: (options.signal ?? timeoutSignal);
-	// The eval tool drives cancellation via an idle-aware `signal` and passes only
-	// an inactivity budget; use it solely as worker cold-start headroom and never
-	// derive a competing fixed timer from it.
+	// The eval tool drives cancellation via its own watchdog `signal` and passes
+	// only the runtime-work budget; use it solely as worker cold-start headroom
+	// and never derive a competing fixed timer from it.
 	const acquireBudgetMs = legacyTimeoutMs ?? options.idleTimeoutMs;
 
 	try {
@@ -105,10 +105,10 @@ export async function executeJs(code: string, options: JsExecutorOptions): Promi
 				onText: chunk => outputSink.push(chunk),
 				onDisplay: output => {
 					if (output.type === "status") {
-						// Heartbeats are pure idle-watchdog keepalives: forward them so
-						// the eval tool re-arms its timer, but never store or render them.
+						// Timeout-control events drive the eval watchdog only; never
+						// store or render them as cell output.
 						options.onStatus?.(output.event);
-						if (output.event.op === EVAL_HEARTBEAT_OP) return;
+						if (isEvalTimeoutControlEvent(output.event)) return;
 					}
 					displayOutputs.push(output);
 				},
diff --git a/packages/coding-agent/src/eval/llm-bridge.ts b/packages/coding-agent/src/eval/llm-bridge.ts
index 39fd1168f..1f37dc2c9 100644
--- a/packages/coding-agent/src/eval/llm-bridge.ts
+++ b/packages/coding-agent/src/eval/llm-bridge.ts
@@ -15,10 +15,11 @@ import { instrumentedCompleteSimple, resolveTelemetry } from "@oh-my-pi/pi-agent
 import { type Api, Effort, getSupportedEfforts, type Model, type Tool } from "@oh-my-pi/pi-ai";
 import * as z from "zod/v4";
 import { extractTextContent, extractToolCall, parseJsonPayload } from "../commit/utils";
+
 import { expandRoleAlias, formatModelString, resolveModelFromString } from "../config/model-resolver";
 import type { ToolSession } from "../tools";
 import { ToolError } from "../tools/tool-errors";
-import { withBridgeHeartbeat } from "./heartbeat";
+import { withBridgeTimeoutPause } from "./bridge-timeout";
 import type { JsStatusEvent } from "./js/shared/types";
 
 /** Synthetic bridge name reserved for the `llm()` helper across both runtimes. */
@@ -112,8 +113,9 @@ export async function runEvalLlm(args: unknown, options: EvalLlmBridgeOptions):
 		);
 	}
 
-	const apiKey = await options.session.modelRegistry?.getApiKey(model);
-	if (!apiKey) {
+	const registry = options.session.modelRegistry;
+	const apiKey = await registry?.getApiKey(model);
+	if (!registry || !apiKey) {
 		throw new ToolError(
 			`llm() has no API key for ${formatModelString(model)}. Configure credentials for this provider or choose another tier.`,
 		);
@@ -132,10 +134,9 @@ export async function runEvalLlm(args: unknown, options: EvalLlmBridgeOptions):
 
 	const telemetry = resolveTelemetry(options.session.getTelemetry?.(), options.session.getSessionId?.() ?? undefined);
 
-	// A oneshot completion emits no status until it returns, so pump a heartbeat
-	// while it runs to keep the eval idle watchdog armed across a slow (e.g.
-	// reasoning-tier) request that would otherwise look like a stalled cell.
-	const response = await withBridgeHeartbeat(options.emitStatus, () =>
+	// Suspend eval timeout accounting while the model request owns control. The
+	// timeout clock restarts once the bridge returns to the cell runtime.
+	const response = await withBridgeTimeoutPause(options.emitStatus, () =>
 		instrumentedCompleteSimple(
 			model,
 			{
@@ -144,7 +145,10 @@ export async function runEvalLlm(args: unknown, options: EvalLlmBridgeOptions):
 				tools,
 			},
 			{
-				apiKey,
+				apiKey: registry.resolver(model.provider, {
+					sessionId: options.session.getSessionId?.() ?? undefined,
+					baseUrl: model.baseUrl,
+				}),
 				signal: options.signal,
 				reasoning: reasoningForTier(tier, model),
 				toolChoice: schema ? { type: "tool", name: STRUCTURED_TOOL_NAME } : undefined,
diff --git a/packages/coding-agent/src/eval/py/executor.ts b/packages/coding-agent/src/eval/py/executor.ts
index 788fa27ea..6f678527c 100644
--- a/packages/coding-agent/src/eval/py/executor.ts
+++ b/packages/coding-agent/src/eval/py/executor.ts
@@ -5,7 +5,7 @@ import { Settings } from "../../config/settings";
 import { OutputSink } from "../../session/streaming-output";
 import type { ToolSession } from "../../tools";
 import { resolveOutputMaxColumns, resolveOutputSinkHeadBytes } from "../../tools/output-meta";
-import { EVAL_HEARTBEAT_OP } from "../heartbeat";
+import { isEvalTimeoutControlEvent } from "../bridge-timeout";
 import type { JsStatusEvent } from "../js/shared/types";
 import {
 	checkPythonKernelAvailability,
@@ -27,8 +27,8 @@ export interface PythonExecutorOptions {
 	/** Absolute wall-clock deadline in milliseconds since epoch */
 	deadlineMs?: number;
 	/**
-	 * Inactivity budget (ms). Used only for timeout-annotation text when the
-	 * caller drives cancellation via an idle-aware `signal` instead of a
+	 * Runtime-work budget (ms). Used only for timeout-annotation text when the
+	 * caller drives cancellation via the eval watchdog `signal` instead of a
 	 * wall-clock `deadlineMs`/`timeoutMs`. Does not arm a timer.
 	 */
 	idleTimeoutMs?: number;
@@ -492,10 +492,10 @@ async function executeWithKernel(
 	// long-running bridge helpers (e.g. `agent()`) surface progress mid-cell.
 	const collectDisplay = (output: KernelDisplayOutput) => {
 		if (output.type === "status") {
-			// Heartbeats are pure idle-watchdog keepalives: forward them so the
-			// eval tool re-arms its timer, but never store or render them.
+			// Timeout-control events drive the eval watchdog only; never store or
+			// render them as cell output.
 			options?.onStatus?.(output.event);
-			if (output.event.op === EVAL_HEARTBEAT_OP) return;
+			if (isEvalTimeoutControlEvent(output.event)) return;
 		}
 		displayOutputs.push(output);
 	};
diff --git a/packages/coding-agent/src/eval/py/kernel.ts b/packages/coding-agent/src/eval/py/kernel.ts
index 0d056cf1d..6d741c309 100644
--- a/packages/coding-agent/src/eval/py/kernel.ts
+++ b/packages/coding-agent/src/eval/py/kernel.ts
@@ -18,6 +18,7 @@ import { type KernelDisplayOutput, renderKernelDisplay } from "./display";
 import { PYTHON_PRELUDE } from "./prelude";
 import RUNNER_SCRIPT from "./runner.py" with { type: "text" };
 import { enumeratePythonRuntimes, filterEnv, type PythonRuntime, resolvePythonRuntime } from "./runtime";
+import { hostHasInheritableConsole, shouldHideKernelWindow } from "./spawn-options";
 
 export type { KernelDisplayOutput, PythonStatusEvent } from "./display";
 export { renderKernelDisplay } from "./display";
@@ -253,7 +254,16 @@ export class PythonKernel {
 			stdin: "pipe",
 			stdout: "pipe",
 			stderr: "pipe",
-			windowsHide: true,
+			// Detached from any inherited console only when the host itself
+			// has no console — kernel32!GetConsoleWindow() is authoritative
+			// (works even when every stdio stream is redirected), with a
+			// TTY-OR fallback when the FFI probe is unavailable. See #1960
+			// for the numpy/pandas LoadLibraryExW hang + SIGINT-recovery
+			// failure that motivates the predicate.
+			windowsHide: shouldHideKernelWindow({
+				platform: process.platform,
+				hostHasInheritableConsole: hostHasInheritableConsole(),
+			}),
 		});
 		kernel.#proc = proc;
 		kernel.#stdin = proc.stdin;
diff --git a/packages/coding-agent/src/eval/py/spawn-options.ts b/packages/coding-agent/src/eval/py/spawn-options.ts
new file mode 100644
index 000000000..c422577f9
--- /dev/null
+++ b/packages/coding-agent/src/eval/py/spawn-options.ts
@@ -0,0 +1,126 @@
+/**
+ * Subprocess spawn-option helpers for the Python kernel.
+ *
+ * Pure helpers (`shouldHideKernelWindow`, `consoleAttachedViaTTY`) live here
+ * so they can be unit-tested without dragging in the kernel's runtime
+ * dependencies. The effectful `hostHasInheritableConsole` wraps a Win32 FFI
+ * probe with a TTY fallback and is the function `kernel.ts` actually calls.
+ */
+import { dlopen, FFIType } from "bun:ffi";
+
+/**
+ * Decide whether the long-lived Python kernel subprocess should be spawned
+ * with `windowsHide: true`.
+ *
+ * On Windows, Bun maps `windowsHide: true` to the `CREATE_NO_WINDOW` flag,
+ * which detaches the child from any inherited console. The Python kernel
+ * runs user code that imports NumPy/pandas; those native extensions
+ * (`numpy/_core/_multiarray_umath.pyd` + bundled OpenBLAS/SLEEF thread-pool
+ * init) can deadlock inside `LoadLibraryExW` when no console is attached,
+ * and a console-less child cannot receive SIGINT via
+ * `GenerateConsoleCtrlEvent` (the recovery path the host relies on). See
+ * issue #1960.
+ *
+ * So on Windows we hide only when the host itself has no console to share.
+ * In any launch where a console is attached — even one with every stdio
+ * stream redirected — the kernel inherits the parent's console, matching
+ * `python.exe` invoked from `cmd.exe`, which keeps native imports and
+ * SIGINT recovery working.
+ *
+ * Short-lived helper subprocesses elsewhere in the codebase (LSP probes,
+ * git, plugin installs) keep `windowsHide: true` because they don't load
+ * complex native modules and the brief console flash would be user-visible
+ * noise.
+ */
+export function shouldHideKernelWindow(opts: {
+	platform: NodeJS.Platform;
+	hostHasInheritableConsole: boolean;
+}): boolean {
+	if (opts.platform !== "win32") return false;
+	return !opts.hostHasInheritableConsole;
+}
+
+/**
+ * TTY-based fallback used when the Win32 console probe is unavailable.
+ *
+ * Returns `true` if any of stdin/stdout/stderr is currently a TTY. This
+ * correctly detects the common interactive launches and the partial-
+ * redirection cases (`omp -p > out.txt`, `< in.txt`, `2> err.log`) where at
+ * least one stream stays bound to the terminal. The all-stdio-redirected
+ * case (`< in > out 2> err` from a console) is the reason we prefer the
+ * Win32 probe over this fallback whenever possible.
+ */
+export function consoleAttachedViaTTY(opts: {
+	stdinIsTTY: boolean;
+	stdoutIsTTY: boolean;
+	stderrIsTTY: boolean;
+}): boolean {
+	return opts.stdinIsTTY || opts.stdoutIsTTY || opts.stderrIsTTY;
+}
+
+/**
+ * Probe `kernel32.dll!GetConsoleWindow()` to detect whether the current
+ * Windows process owns a console window.
+ *
+ * Returns `true` for a non-NULL HWND, `false` when NULL (no console — true
+ * service / `DETACHED_PROCESS` / GUI parent), and `null` when the probe
+ * itself fails (off-Windows, FFI disabled, or unexpected kernel32 layout).
+ * A `null` return means "don't trust me, use the TTY fallback".
+ *
+ * Cached on first call because in practice the console attachment of a
+ * long-lived OMP host never changes for the lifetime of the process, and
+ * we don't want to re-dlopen kernel32 on every kernel spawn.
+ */
+type ConsoleProbeResult = boolean | null;
+let cachedWindowsConsoleProbe: { value: ConsoleProbeResult } | undefined;
+
+function probeWindowsConsoleWindow(): ConsoleProbeResult {
+	if (cachedWindowsConsoleProbe) return cachedWindowsConsoleProbe.value;
+	let value: ConsoleProbeResult = null;
+	try {
+		const lib = dlopen("kernel32.dll", {
+			GetConsoleWindow: { args: [], returns: FFIType.ptr },
+		});
+		try {
+			const hwnd = lib.symbols.GetConsoleWindow();
+			// FFIType.ptr returns `Pointer | null`; a 0 pointer should also be
+			// treated as NULL defensively in case Bun ever returns 0n / 0.
+			value = hwnd !== null && hwnd !== 0;
+		} finally {
+			lib.close();
+		}
+	} catch {
+		value = null;
+	}
+	cachedWindowsConsoleProbe = { value };
+	return value;
+}
+
+/** Reset the cached Win32 probe result. Test-only; not part of the public surface. */
+export function __resetWindowsConsoleProbeCache(): void {
+	cachedWindowsConsoleProbe = undefined;
+}
+
+/**
+ * Whether the host process owns a console its children can inherit.
+ *
+ * - On Windows, the authoritative signal is `GetConsoleWindow()`. It returns
+ *   a non-NULL HWND whenever the process has a console attached, regardless
+ *   of how the standard streams are redirected — so an `omp -p ... < in.txt
+ *   > out.txt 2> err.log` launched from a real Windows Terminal session is
+ *   correctly classified as console-attached and the kernel keeps its
+ *   inheritable console.
+ * - On any other platform, or if the FFI probe fails, fall back to the
+ *   TTY-OR heuristic. That still catches the common interactive cases.
+ */
+export function hostHasInheritableConsole(): boolean {
+	if (process.platform === "win32") {
+		const native = probeWindowsConsoleWindow();
+		if (native !== null) return native;
+	}
+	return consoleAttachedViaTTY({
+		stdinIsTTY: !!process.stdin.isTTY,
+		stdoutIsTTY: !!process.stdout.isTTY,
+		stderrIsTTY: !!process.stderr.isTTY,
+	});
+}
diff --git a/packages/coding-agent/src/export/ttsr.ts b/packages/coding-agent/src/export/ttsr.ts
index fa973b420..8dba7a399 100644
--- a/packages/coding-agent/src/export/ttsr.ts
+++ b/packages/coding-agent/src/export/ttsr.ts
@@ -294,6 +294,9 @@ export class TtsrManager {
 
 	/** Add a TTSR rule to be monitored. */
 	addRule(rule: Rule): boolean {
+		if (!this.#settings.enabled) {
+			return false;
+		}
 		if (this.#rules.has(rule.name)) {
 			return false;
 		}
@@ -357,6 +360,9 @@ export class TtsrManager {
 	}
 
 	#matchBuffer(buffer: string, context: TtsrMatchContext): Rule[] {
+		if (!this.#settings.enabled) {
+			return [];
+		}
 		const matches: Rule[] = [];
 		for (const [name, entry] of this.#rules) {
 			if (!this.#canTrigger(name)) {
@@ -433,6 +439,9 @@ export class TtsrManager {
 
 	/** Check if any TTSR rules are registered. */
 	hasRules(): boolean {
+		if (!this.#settings.enabled) {
+			return false;
+		}
 		return this.#rules.size > 0;
 	}
 
diff --git a/packages/coding-agent/src/extensibility/custom-commands/bundled/ci-green/index.ts b/packages/coding-agent/src/extensibility/custom-commands/bundled/ci-green/index.ts
index a4032b792..78adb624f 100644
--- a/packages/coding-agent/src/extensibility/custom-commands/bundled/ci-green/index.ts
+++ b/packages/coding-agent/src/extensibility/custom-commands/bundled/ci-green/index.ts
@@ -12,6 +12,35 @@ async function getHeadTag(api: CustomCommandAPI): Promise<string | undefined> {
 	}
 }
 
+async function getCurrentBranch(api: CustomCommandAPI): Promise<string> {
+	try {
+		return (await git.branch.current(api.cwd)) ?? "HEAD";
+	} catch {
+		return "HEAD";
+	}
+}
+
+async function getPushRemote(api: CustomCommandAPI, branch: string): Promise<string | undefined> {
+	try {
+		return (
+			(await git.config.getBranch(api.cwd, branch, "pushRemote")) ??
+			(await git.config.getBranch(api.cwd, branch, "remote"))
+		);
+	} catch {
+		return undefined;
+	}
+}
+
+async function getHeadTagContext(api: CustomCommandAPI): Promise<{ branch: string; headTag?: string; remote: string }> {
+	const branch = await getCurrentBranch(api);
+	const [headTag, pushRemote] = await Promise.all([getHeadTag(api), getPushRemote(api, branch)]);
+	return {
+		headTag,
+		branch,
+		remote: pushRemote ?? "origin",
+	};
+}
+
 export class GreenCommand implements CustomCommand {
 	name = "green";
 	description = "Generate a prompt to iterate on CI failures until the branch is green";
@@ -19,7 +48,7 @@ export class GreenCommand implements CustomCommand {
 	constructor(private api: CustomCommandAPI) {}
 
 	async execute(_args: string[], _ctx: HookCommandContext): Promise<string> {
-		const headTag = await getHeadTag(this.api);
-		return prompt.render(ciGreenRequestTemplate, { headTag });
+		const { headTag, branch, remote } = await getHeadTagContext(this.api);
+		return prompt.render(ciGreenRequestTemplate, { headTag, branch, remote });
 	}
 }
diff --git a/packages/coding-agent/src/extensibility/extensions/runner.ts b/packages/coding-agent/src/extensibility/extensions/runner.ts
index 3b23d3c73..76abcf464 100644
--- a/packages/coding-agent/src/extensibility/extensions/runner.ts
+++ b/packages/coding-agent/src/extensibility/extensions/runner.ts
@@ -354,6 +354,9 @@ export class ExtensionRunner {
 		"ctrl+o": true,
 		"ctrl+t": true,
 		"ctrl+g": true,
+		"alt+m": true,
+		// Default chord for `app.message.followUp` (Windows Terminal can't deliver Ctrl+Enter; #1903).
+		"ctrl+q": true,
 		"shift+tab": true,
 		"shift+ctrl+p": true,
 		"alt+enter": true,
diff --git a/packages/coding-agent/src/extensibility/plugins/doctor.ts b/packages/coding-agent/src/extensibility/plugins/doctor.ts
index ff33cc562..5bd600c67 100644
--- a/packages/coding-agent/src/extensibility/plugins/doctor.ts
+++ b/packages/coding-agent/src/extensibility/plugins/doctor.ts
@@ -25,7 +25,6 @@ export async function runDoctorChecks(): Promise<DoctorCheck[]> {
 	const apiKeys = [
 		{ name: "ANTHROPIC_API_KEY", description: "Anthropic API" },
 		{ name: "OPENAI_API_KEY", description: "OpenAI API" },
-		{ name: "PERPLEXITY_API_KEY", description: "Perplexity search" },
 		{ name: "EXA_API_KEY", description: "Exa search" },
 	];
 
diff --git a/packages/coding-agent/src/extensibility/plugins/marketplace-auto-update.ts b/packages/coding-agent/src/extensibility/plugins/marketplace-auto-update.ts
new file mode 100644
index 000000000..f81da80a2
--- /dev/null
+++ b/packages/coding-agent/src/extensibility/plugins/marketplace-auto-update.ts
@@ -0,0 +1,49 @@
+import { getProjectDir, logger } from "@oh-my-pi/pi-utils";
+
+type MarketplaceAutoUpdateMode = "off" | "notify" | "auto";
+
+interface MarketplaceAutoUpdateOptions {
+	autoUpdate: MarketplaceAutoUpdateMode;
+	resolveActiveProjectRegistryPath: (cwd: string) => Promise<string | null>;
+	clearPluginRootsCache: () => void;
+}
+
+export function scheduleMarketplaceAutoUpdate(options: MarketplaceAutoUpdateOptions): void {
+	if (options.autoUpdate === "off") {
+		return;
+	}
+
+	void runMarketplaceAutoUpdate(options);
+}
+
+async function runMarketplaceAutoUpdate(options: MarketplaceAutoUpdateOptions): Promise<void> {
+	try {
+		// Startup perf: marketplace manager pulls scraper/fetch/cache code; keep it out of the initial TUI graph.
+		const {
+			MarketplaceManager,
+			getInstalledPluginsRegistryPath,
+			getMarketplacesCacheDir,
+			getMarketplacesRegistryPath,
+			getPluginsCacheDir,
+		} = await import("./marketplace");
+		const mgr = new MarketplaceManager({
+			marketplacesRegistryPath: getMarketplacesRegistryPath(),
+			installedRegistryPath: getInstalledPluginsRegistryPath(),
+			projectInstalledRegistryPath: (await options.resolveActiveProjectRegistryPath(getProjectDir())) ?? undefined,
+			marketplacesCacheDir: getMarketplacesCacheDir(),
+			pluginsCacheDir: getPluginsCacheDir(),
+			clearPluginRootsCache: options.clearPluginRootsCache,
+		});
+		await mgr.refreshStaleMarketplaces();
+		const updates = await mgr.checkForUpdates();
+		if (updates.length === 0) return;
+		if (options.autoUpdate === "auto") {
+			await mgr.upgradeAllPlugins();
+			logger.debug(`Auto-upgraded ${updates.length} marketplace plugin(s)`);
+		} else {
+			logger.debug(`${updates.length} marketplace plugin update(s) available — /marketplace upgrade`);
+		}
+	} catch {
+		// Silently ignore — network failure, corrupt data, offline.
+	}
+}
diff --git a/packages/coding-agent/src/goals/tools/goal-tool.ts b/packages/coding-agent/src/goals/tools/goal-tool.ts
index 1c94cb54f..94b9545bf 100644
--- a/packages/coding-agent/src/goals/tools/goal-tool.ts
+++ b/packages/coding-agent/src/goals/tools/goal-tool.ts
@@ -8,9 +8,9 @@ import type { Theme, ThemeColor } from "../../modes/theme/theme";
 import goalDescription from "../../prompts/tools/goal.md" with { type: "text" };
 import { formatDuration } from "../../slash-commands/helpers/format";
 import type { ToolSession } from "../../tools";
-import { formatErrorMessage, TRUNCATE_LENGTHS } from "../../tools/render-utils";
+import { formatErrorDetail, TRUNCATE_LENGTHS } from "../../tools/render-utils";
 import { ToolError } from "../../tools/tool-errors";
-import { renderStatusLine, truncateToWidth } from "../../tui";
+import { framedBlock, renderStatusLine, truncateToWidth } from "../../tui";
 import { completionBudgetReport, remainingTokens } from "../runtime";
 import type { Goal, GoalStatus, GoalToolDetails } from "../state";
 
@@ -173,8 +173,7 @@ export const goalToolRenderer = {
 		if (args.op === "create" && args.token_budget !== undefined) {
 			meta.push(`budget ${formatNumber(args.token_budget)}`);
 		}
-		const text = renderStatusLine({ icon: "pending", title: "Goal", description, meta }, uiTheme);
-		return new Text(text, 0, 0);
+		return new Text(renderStatusLine({ icon: "pending", title: "Goal", description, meta }, uiTheme), 0, 0);
 	},
 
 	renderResult(
@@ -190,51 +189,62 @@ export const goalToolRenderer = {
 
 		if (result.isError) {
 			const header = renderStatusLine({ icon: "error", title: "Goal", description }, uiTheme);
-			const body = formatErrorMessage(fallbackText || "Goal tool failed", uiTheme);
-			return new Text([header, body].join("\n"), 0, 0);
+			return framedBlock(uiTheme, width => ({
+				header,
+				sections: [{ lines: formatErrorDetail(fallbackText || "Goal tool failed", uiTheme).split("\n") }],
+				state: "error",
+				borderColor: "error",
+				width,
+			}));
 		}
 
 		const goal = details?.goal ?? null;
 		if (!goal) {
-			const header = renderStatusLine({ icon: "warning", title: "Goal", description }, uiTheme);
-			const body = uiTheme.fg("muted", "No active goal.");
-			return new Text([header, body].join("\n"), 0, 0);
+			return new Text(
+				renderStatusLine({ icon: "warning", title: "Goal", description, meta: ["no active goal"] }, uiTheme),
+				0,
+				0,
+			);
 		}
 
-		const lines: string[] = [];
-		lines.push(
-			renderStatusLine(
-				{
-					icon: "success",
-					title: "Goal",
-					description,
-					badge: { label: goal.status, color: goalBadgeColor(goal.status) },
-				},
-				uiTheme,
-			),
+		const header = renderStatusLine(
+			{
+				icon: "success",
+				title: "Goal",
+				description,
+				badge: { label: goal.status, color: goalBadgeColor(goal.status) },
+			},
+			uiTheme,
 		);
 
+		const lines: string[] = [];
 		const objectiveText = truncateToWidth(goal.objective.trim(), TRUNCATE_LENGTHS.LONG);
-		lines.push(`  ${uiTheme.italic(uiTheme.fg("muted", `"${objectiveText}"`))}`);
+		lines.push(uiTheme.italic(uiTheme.fg("muted", `"${objectiveText}"`)));
 
 		const used = formatNumber(goal.tokensUsed);
 		const tokensLine =
 			goal.tokenBudget !== undefined
 				? `${used} / ${formatNumber(goal.tokenBudget)} tokens (${formatNumber(Math.max(0, goal.tokenBudget - goal.tokensUsed))} left)`
 				: `${used} tokens`;
-		lines.push(`  ${uiTheme.fg("dim", tokensLine)}`);
-
+		const metaParts = [tokensLine];
 		if (goal.timeUsedSeconds > 0) {
-			lines.push(`  ${uiTheme.fg("dim", `${formatDuration(goal.timeUsedSeconds * 1000)} elapsed`)}`);
+			metaParts.push(`${formatDuration(goal.timeUsedSeconds * 1000)} elapsed`);
 		}
+		lines.push(uiTheme.fg("dim", metaParts.join(" · ")));
 
 		const report = details?.completionBudgetReport;
+		const sections: Array<{ label?: string; lines: string[] }> = [{ lines }];
 		if (report) {
-			lines.push("");
-			lines.push(uiTheme.italic(uiTheme.fg("muted", report)));
+			sections.push({ label: "Report", lines: report.split("\n").map(line => uiTheme.fg("muted", line)) });
 		}
 
-		return new Text(lines.join("\n"), 0, 0);
+		return framedBlock(uiTheme, width => ({
+			header,
+			sections,
+			state: "success",
+			borderColor: "borderMuted",
+			width,
+		}));
 	},
 
 	mergeCallAndResult: true,
diff --git a/packages/coding-agent/src/lsp/client.ts b/packages/coding-agent/src/lsp/client.ts
index c93df3abe..8127cdfc4 100644
--- a/packages/coding-agent/src/lsp/client.ts
+++ b/packages/coding-agent/src/lsp/client.ts
@@ -1,3 +1,4 @@
+import * as path from "node:path";
 import { isEnoent, logger, ptree, untilAborted } from "@oh-my-pi/pi-utils";
 import { ToolAbortError, throwIfAborted } from "../tools/tool-errors";
 import { applyWorkspaceEdit } from "./edits";
@@ -147,6 +148,7 @@ const CLIENT_CAPABILITIES = {
 			failureHandling: "textOnlyTransactional",
 		},
 		configuration: true,
+		workspaceFolders: true,
 		symbol: {
 			dynamicRegistration: false,
 			symbolKind: {
@@ -172,49 +174,69 @@ const CLIENT_CAPABILITIES = {
 // LSP Message Protocol
 // =============================================================================
 
-/**
- * Parse a single LSP message from a buffer.
- * Returns the parsed message and remaining buffer, or null if incomplete.
- */
-function parseMessage(
-	buffer: Buffer,
-): { message: LspJsonRpcResponse | LspJsonRpcNotification; remaining: Buffer } | null {
-	// Only decode enough to find the header
-	const headerEndIndex = findHeaderEnd(buffer);
-	if (headerEndIndex === -1) return null;
-
-	const headerText = new TextDecoder().decode(buffer.slice(0, headerEndIndex));
-	const contentLengthMatch = headerText.match(/Content-Length: (\d+)/i);
-	if (!contentLengthMatch) return null;
-
-	const contentLength = Number.parseInt(contentLengthMatch[1], 10);
-	const messageStart = headerEndIndex + 4; // Skip \r\n\r\n
-	const messageEnd = messageStart + contentLength;
-
-	if (buffer.length < messageEnd) return null;
-
-	const messageBytes = buffer.subarray(messageStart, messageEnd);
-	const messageText = new TextDecoder().decode(messageBytes);
-	const remaining = buffer.subarray(messageEnd);
-
-	return {
-		message: JSON.parse(messageText),
-		remaining,
-	};
-}
+// Reused for all full (non-streaming) decodes; each decode() resets state, so a
+// single instance is safe and avoids per-message TextDecoder allocation.
+const MESSAGE_DECODER = new TextDecoder("utf-8");
 
 /**
- * Find the end of the header section (before \r\n\r\n)
+ * Locate the `\r\n\r\n` header terminator across the pending chunk list.
+ * Returns the absolute byte index of the first `\r`, or -1 when not present.
+ * Equivalent to scanning the contiguous concatenation of the chunks.
  */
-function findHeaderEnd(buffer: Uint8Array): number {
-	for (let i = 0; i < buffer.length - 3; i++) {
-		if (buffer[i] === 13 && buffer[i + 1] === 10 && buffer[i + 2] === 13 && buffer[i + 3] === 10) {
-			return i;
+function findHeaderEndInChunks(chunks: Buffer[]): number {
+	let global = 0;
+	let b0 = -1;
+	let b1 = -1;
+	let b2 = -1;
+	for (const chunk of chunks) {
+		for (let i = 0; i < chunk.length; i++) {
+			const b3 = chunk[i];
+			if (b0 === 13 && b1 === 10 && b2 === 13 && b3 === 10) {
+				return global - 3;
+			}
+			b0 = b1;
+			b1 = b2;
+			b2 = b3;
+			global++;
 		}
 	}
 	return -1;
 }
 
+/** Copy the byte range [from, to) out of the pending chunk list into one Buffer. */
+function copyChunkRange(chunks: Buffer[], from: number, to: number): Buffer {
+	const out = Buffer.allocUnsafe(to - from);
+	let global = 0;
+	let written = 0;
+	for (const chunk of chunks) {
+		const chunkEnd = global + chunk.length;
+		if (chunkEnd > from && global < to) {
+			const start = Math.max(from, global) - global;
+			const end = Math.min(to, chunkEnd) - global;
+			chunk.copy(out, written, start, end);
+			written += end - start;
+		}
+		global = chunkEnd;
+		if (global >= to) break;
+	}
+	return out;
+}
+
+/** Drop the first `count` bytes from the pending chunk list in place. */
+function dropChunkFront(chunks: Buffer[], count: number): void {
+	let removed = 0;
+	while (chunks.length > 0) {
+		const head = chunks[0];
+		if (removed + head.length <= count) {
+			removed += head.length;
+			chunks.shift();
+		} else {
+			chunks[0] = head.subarray(count - removed);
+			break;
+		}
+	}
+}
+
 async function writeMessage(
 	sink: Bun.FileSink,
 	message: LspJsonRpcRequest | LspJsonRpcNotification | LspJsonRpcResponse,
@@ -247,22 +269,43 @@ async function startMessageReader(client: LspClient): Promise<void> {
 
 	const reader = (client.proc.stdout as ReadableStream<Uint8Array>).getReader();
 
+	// Incoming bytes are buffered as a list of chunks and only joined when a full
+	// message is framed. Concatenating the accumulator on every read was O(n^2)
+	// for messages that span many reads (e.g. a large initial diagnostics burst).
+	const pendingChunks: Buffer[] = [];
+	let pendingLen = 0;
+	if (client.messageBuffer.length > 0) {
+		const seed = Buffer.from(client.messageBuffer);
+		pendingChunks.push(seed);
+		pendingLen = seed.length;
+	}
+
 	try {
 		while (true) {
 			const { done, value } = await reader.read();
 			if (done) break;
 
-			// Atomically update buffer before processing
-			const currentBuffer: Buffer = Buffer.concat([client.messageBuffer, value]);
-			client.messageBuffer = currentBuffer;
+			pendingChunks.push(Buffer.from(value));
+			pendingLen += value.length;
 
-			// Process all complete messages in buffer
-			// Use local variable to avoid race with concurrent buffer updates
-			let workingBuffer = currentBuffer;
-			let parsed = parseMessage(workingBuffer);
-			while (parsed) {
-				const { message, remaining } = parsed;
-				workingBuffer = remaining;
+			// Drain every complete message currently buffered.
+			while (true) {
+				const headerEnd = findHeaderEndInChunks(pendingChunks);
+				if (headerEnd === -1) break;
+
+				const headerText = MESSAGE_DECODER.decode(copyChunkRange(pendingChunks, 0, headerEnd));
+				const contentLengthMatch = headerText.match(/Content-Length: (\d+)/i);
+				if (!contentLengthMatch) break;
+
+				const contentLength = Number.parseInt(contentLengthMatch[1], 10);
+				const messageStart = headerEnd + 4; // Skip \r\n\r\n
+				const messageEnd = messageStart + contentLength;
+				if (pendingLen < messageEnd) break;
+
+				const messageText = MESSAGE_DECODER.decode(copyChunkRange(pendingChunks, messageStart, messageEnd));
+				const message: LspJsonRpcResponse | LspJsonRpcNotification = JSON.parse(messageText);
+				dropChunkFront(pendingChunks, messageEnd);
+				pendingLen -= messageEnd;
 
 				// Route message
 				if ("id" in message && message.id !== undefined) {
@@ -299,12 +342,7 @@ async function startMessageReader(client: LspClient): Promise<void> {
 						}
 					}
 				}
-
-				parsed = parseMessage(workingBuffer);
 			}
-
-			// Atomically commit processed buffer
-			client.messageBuffer = workingBuffer;
 		}
 	} catch (err) {
 		// Connection closed or error - reject all pending requests
@@ -313,11 +351,34 @@ async function startMessageReader(client: LspClient): Promise<void> {
 		}
 		client.pendingRequests.clear();
 	} finally {
+		// Persist any unparsed remainder so a restarted reader resumes mid-message.
+		client.messageBuffer =
+			pendingChunks.length === 0
+				? new Uint8Array(0)
+				: pendingChunks.length === 1
+					? pendingChunks[0]
+					: Buffer.concat(pendingChunks, pendingLen);
 		reader.releaseLock();
 		client.isReading = false;
 	}
 }
 
+/**
+ * Build the workspace folder list advertised to the server. Identical shape
+ * for `initialize` params and `workspace/workspaceFolders` server requests.
+ */
+function currentWorkspaceFolders(client: LspClient): Array<{ uri: string; name: string }> {
+	return [{ uri: fileToUri(client.cwd), name: path.basename(client.cwd) || "workspace" }];
+}
+
+/**
+ * Handle workspace/workspaceFolders requests from the server.
+ */
+async function handleWorkspaceFoldersRequest(client: LspClient, message: LspJsonRpcRequest): Promise<void> {
+	if (typeof message.id !== "number") return;
+	await sendResponse(client, message.id, currentWorkspaceFolders(client), "workspace/workspaceFolders");
+}
+
 /**
  * Handle workspace/configuration requests from the server.
  */
@@ -364,6 +425,10 @@ async function handleServerRequest(client: LspClient, message: LspJsonRpcRequest
 		await handleConfigurationRequest(client, message);
 		return;
 	}
+	if (message.method === "workspace/workspaceFolders") {
+		await handleWorkspaceFoldersRequest(client, message);
+		return;
+	}
 	if (message.method === "workspace/applyEdit") {
 		await handleApplyEditRequest(client, message);
 		return;
@@ -412,7 +477,66 @@ async function sendResponse(
 /** Timeout for warmup initialize requests (5 seconds) */
 export const WARMUP_TIMEOUT_MS = 5000;
 
-/** Max time to wait for the server to report project loading completion via $/progress */
+/** Max time to poll rust-analyzer after progress ends but before Cargo workspaces are ready. */
+const RUST_ANALYZER_WORKSPACE_READY_TIMEOUT_MS = 5_000;
+const RUST_ANALYZER_WORKSPACE_READY_POLL_MS = 100;
+const RUST_ANALYZER_WORKSPACE_READY_SETTLE_MS = 2_000;
+const RUST_ANALYZER_STATUS_REQUEST_TIMEOUT_MS = 1_000;
+const rustAnalyzerReadyClients = new WeakSet<LspClient>();
+
+function commandBasename(command: string): string {
+	const slash = command.lastIndexOf("/");
+	const backslash = command.lastIndexOf("\\");
+	const separator = Math.max(slash, backslash);
+	return separator === -1 ? command : command.slice(separator + 1);
+}
+
+function isRustAnalyzerClient(client: LspClient): boolean {
+	return (
+		commandBasename(client.config.command) === "rust-analyzer" ||
+		(client.config.resolvedCommand ? commandBasename(client.config.resolvedCommand) === "rust-analyzer" : false)
+	);
+}
+
+function isRustAnalyzerStatusTimeout(err: unknown): boolean {
+	return err instanceof Error && err.message.startsWith("LSP request rust-analyzer/analyzerStatus timed out after ");
+}
+
+async function waitForRustAnalyzerWorkspace(client: LspClient, signal?: AbortSignal): Promise<void> {
+	if (rustAnalyzerReadyClients.has(client)) {
+		return;
+	}
+	const timings = client.config.workspaceReadyTimings;
+	const timeoutMs = timings?.timeoutMs ?? RUST_ANALYZER_WORKSPACE_READY_TIMEOUT_MS;
+	const pollMs = timings?.pollMs ?? RUST_ANALYZER_WORKSPACE_READY_POLL_MS;
+	const settleMs = timings?.settleMs ?? RUST_ANALYZER_WORKSPACE_READY_SETTLE_MS;
+	const statusRequestTimeoutMs = timings?.statusRequestTimeoutMs ?? RUST_ANALYZER_STATUS_REQUEST_TIMEOUT_MS;
+	const started = Date.now();
+	const deadline = started + timeoutMs;
+	while (true) {
+		throwIfAborted(signal);
+		let status: unknown;
+		try {
+			status = await sendRequest(client, "rust-analyzer/analyzerStatus", {}, signal, statusRequestTimeoutMs);
+		} catch (err) {
+			if (!isRustAnalyzerStatusTimeout(err) || Date.now() >= deadline) {
+				return;
+			}
+			await Bun.sleep(pollMs);
+			continue;
+		}
+		const ready = typeof status === "string" && !status.startsWith("No workspaces");
+		if (ready && Date.now() - started >= settleMs) {
+			rustAnalyzerReadyClients.add(client);
+			return;
+		}
+		if (Date.now() >= deadline) {
+			return;
+		}
+		await Bun.sleep(pollMs);
+	}
+}
+
 const PROJECT_LOAD_TIMEOUT_MS = 15_000;
 
 /** Max time to wait for graceful LSP shutdown and process exit. */
@@ -530,7 +654,7 @@ export async function getOrCreateClient(config: ServerConfig, cwd: string, initT
 					rootPath: cwd,
 					capabilities: CLIENT_CAPABILITIES,
 					initializationOptions: config.initOptions ?? {},
-					workspaceFolders: [{ uri: fileToUri(cwd), name: cwd.split("/").pop() ?? "workspace" }],
+					workspaceFolders: currentWorkspaceFolders(client),
 				},
 				undefined, // signal
 				initTimeoutMs,
@@ -635,6 +759,9 @@ export async function waitForProjectLoaded(client: LspClient, signal?: AbortSign
 			? [new Promise<void>(resolve => signal.addEventListener("abort", () => resolve(), { once: true }))]
 			: []),
 	]);
+	if (isRustAnalyzerClient(client)) {
+		await waitForRustAnalyzerWorkspace(client, signal);
+	}
 }
 
 /**
diff --git a/packages/coding-agent/src/lsp/index.ts b/packages/coding-agent/src/lsp/index.ts
index 0a4326502..e45866355 100644
--- a/packages/coding-agent/src/lsp/index.ts
+++ b/packages/coding-agent/src/lsp/index.ts
@@ -304,6 +304,32 @@ const SINGLE_DIAGNOSTICS_WAIT_TIMEOUT_MS = 3000;
 const BATCH_DIAGNOSTICS_WAIT_TIMEOUT_MS = 400;
 const MAX_GLOB_DIAGNOSTIC_TARGETS = 20;
 const WORKSPACE_SYMBOL_LIMIT = 200;
+const PROJECT_INDEXED_ACTIONS: ReadonlySet<string> = new Set([
+	"definition",
+	"type_definition",
+	"implementation",
+	"references",
+	"rename",
+	"hover",
+]);
+
+const RUST_WORKSPACE_MARKERS = ["Cargo.toml", "rust-analyzer.toml"] as const;
+
+function hasRustWorkspaceAncestor(filePath: string): boolean {
+	let dir = path.dirname(filePath);
+	while (true) {
+		for (const marker of RUST_WORKSPACE_MARKERS) {
+			if (fs.existsSync(path.join(dir, marker))) {
+				return true;
+			}
+		}
+		const parent = path.dirname(dir);
+		if (parent === dir) {
+			return false;
+		}
+		dir = parent;
+	}
+}
 
 function limitDiagnosticMessages(messages: string[]): string[] {
 	if (messages.length <= DIAGNOSTIC_MESSAGE_LIMIT) {
@@ -1940,10 +1966,21 @@ export class LspTool implements AgentTool<typeof lspSchema, LspToolDetails, Them
 		try {
 			const client = await getOrCreateClient(serverConfig, this.session.cwd);
 			const targetFile = resolvedFile;
+			const isRustAnalyzerServer =
+				serverName === "rust-analyzer" ||
+				path.basename(serverConfig.command) === "rust-analyzer" ||
+				(serverConfig.resolvedCommand ? path.basename(serverConfig.resolvedCommand) === "rust-analyzer" : false);
+			const needsProjectIndex =
+				targetFile !== null && PROJECT_INDEXED_ACTIONS.has(action) && isProjectAwareLspServer(serverConfig);
+			const rustWorkspaceWait =
+				needsProjectIndex && isRustAnalyzerServer && targetFile !== null && hasRustWorkspaceAncestor(targetFile);
 
 			if (targetFile) {
 				await ensureFileOpen(client, targetFile, signal);
 			}
+			if (rustWorkspaceWait) {
+				await waitForProjectLoaded(client, signal);
+			}
 
 			// For project-aware servers, references/rename/definition without a `symbol`
 			// silently falls back to the first non-whitespace column on the line, which
@@ -1968,10 +2005,7 @@ export class LspTool implements AgentTool<typeof lspSchema, LspToolDetails, Them
 
 			let output: string;
 
-			// Wait for project loading to complete before cross-file operations
-			// to ensure the server has indexed all project files.
-			const crossFileActions = new Set(["definition", "type_definition", "implementation", "references", "rename"]);
-			if (crossFileActions.has(action)) {
+			if (needsProjectIndex && !isRustAnalyzerServer) {
 				await waitForProjectLoaded(client, signal);
 			}
 
diff --git a/packages/coding-agent/src/lsp/render.ts b/packages/coding-agent/src/lsp/render.ts
index e49e356f3..f62a9e0dd 100644
--- a/packages/coding-agent/src/lsp/render.ts
+++ b/packages/coding-agent/src/lsp/render.ts
@@ -21,7 +21,7 @@ import {
 	truncateToWidth,
 } from "../tools/render-utils";
 import { renderStatusLine } from "../tui";
-import { CachedOutputBlock } from "../tui/output-block";
+import { CachedOutputBlock, markFramedBlockComponent } from "../tui/output-block";
 import type { LspParams, LspToolDetails } from "./types";
 
 // =============================================================================
@@ -138,7 +138,7 @@ export function renderResult(
 
 	const outputBlock = new CachedOutputBlock();
 
-	return {
+	return markFramedBlockComponent({
 		render(width: number): string[] {
 			// Read mutable state at render time
 			const { expanded, isPartial, spinnerFrame } = options;
@@ -194,7 +194,7 @@ export function renderResult(
 		invalidate() {
 			outputBlock.invalidate();
 		},
-	};
+	});
 }
 
 // =============================================================================
diff --git a/packages/coding-agent/src/lsp/types.ts b/packages/coding-agent/src/lsp/types.ts
index 96b6a1f6d..42028047a 100644
--- a/packages/coding-agent/src/lsp/types.ts
+++ b/packages/coding-agent/src/lsp/types.ts
@@ -356,6 +356,16 @@ export interface ServerConfig {
 	disabled?: boolean;
 	/** Per-server warmup timeout in milliseconds. Overrides the global WARMUP_TIMEOUT_MS for this server during startup. */
 	warmupTimeoutMs?: number;
+	/**
+	 * Per-server overrides for rust-analyzer workspace-ready polling. When omitted, the module
+	 * defaults are used. Primarily a tuning/test seam to bound the multi-second settle window.
+	 */
+	workspaceReadyTimings?: {
+		timeoutMs?: number;
+		pollMs?: number;
+		settleMs?: number;
+		statusRequestTimeoutMs?: number;
+	};
 	capabilities?: ServerCapabilities;
 	/** If true, this is a linter/formatter server (e.g., Biome) - used only for diagnostics/actions, not type intelligence */
 	isLinter?: boolean;
diff --git a/packages/coding-agent/src/lsp/utils.ts b/packages/coding-agent/src/lsp/utils.ts
index a2678f975..768d706c2 100644
--- a/packages/coding-agent/src/lsp/utils.ts
+++ b/packages/coding-agent/src/lsp/utils.ts
@@ -153,9 +153,10 @@ export function formatDiagnostic(diagnostic: Diagnostic, filePath: string): stri
 const DIAG_PATH_RE = /^(.+?):(\d+:\d+\s+.*)$/;
 
 /**
- * Reformat pre-formatted diagnostic messages into grep-style directory/file groups.
+ * Reformat pre-formatted diagnostic messages into a multi-level, prefix-folded
+ * directory/file grouping (see `formatGroupedFiles`).
  * Input:  ["path:line:col [sev] msg", ...]
- * Output: "# dir/\n## file.ts\n  line:col [sev] msg"
+ * Output: "# pkg/src/\n## file.ts\n  line:col [sev] msg"
  *
  * Messages that don't match the expected format are appended ungrouped at the end.
  */
diff --git a/packages/coding-agent/src/main.ts b/packages/coding-agent/src/main.ts
index 5cdc3a8d8..8d4d30b46 100644
--- a/packages/coding-agent/src/main.ts
+++ b/packages/coding-agent/src/main.ts
@@ -40,19 +40,13 @@ import {
 	resolveActiveProjectRegistryPath,
 } from "./discovery/helpers";
 import { injectOmpExtensionCliRoots } from "./discovery/omp-extension-roots";
-import { exportFromFile } from "./export/html";
 import { ExtensionRunner } from "./extensibility/extensions/runner";
 import type { ExtensionUIContext } from "./extensibility/extensions/types";
-import {
-	getInstalledPluginsRegistryPath,
-	getMarketplacesCacheDir,
-	getMarketplacesRegistryPath,
-	getPluginsCacheDir,
-	MarketplaceManager,
-} from "./extensibility/plugins/marketplace";
+import { scheduleMarketplaceAutoUpdate } from "./extensibility/plugins/marketplace-auto-update";
 import type { MCPManager } from "./mcp";
-import { InteractiveMode, runAcpMode, runPrintMode, runRpcMode } from "./modes";
-import { ALL_SCENES, runSetupWizard, selectSetupScenes } from "./modes/setup-wizard";
+import { InteractiveMode } from "./modes/interactive-mode";
+import type { PrintModeOptions } from "./modes/print-mode";
+import { CURRENT_SETUP_VERSION } from "./modes/setup-version";
 import { initTheme, stopThemeWatcher } from "./modes/theme/theme";
 import type { SubmittedUserInput } from "./modes/types";
 import {
@@ -72,6 +66,13 @@ import type { LspStartupServerInfo } from "./tools";
 import { getChangelogPath, getNewEntries, parseChangelog } from "./utils/changelog";
 import { EventBus } from "./utils/event-bus";
 
+type RunAcpMode = (createSession: AcpSessionFactory) => Promise<never>;
+type RunPrintMode = (session: AgentSession, options: PrintModeOptions) => Promise<void>;
+type RunRpcMode = (
+	session: AgentSession,
+	setToolUIContext?: (uiContext: ExtensionUIContext, hasUI: boolean) => void,
+) => Promise<never>;
+
 async function checkForNewVersion(currentVersion: string): Promise<string | undefined> {
 	if (!settings.get("startup.checkUpdate")) {
 		return;
@@ -261,17 +262,29 @@ async function runInteractiveMode(
 		eventBus,
 	);
 
-	const setupScenes = await selectSetupScenes(settings.get("setupVersion"), ALL_SCENES, mode, {
-		resuming,
-		isTTY: process.stdin.isTTY && process.stdout.isTTY,
-		setupWizardEnabled: settings.get("startup.setupWizard"),
-		force: forceSetupWizard,
+	// Cold-launch gate: the full setup wizard (every scene + the overlay and
+	// their TUI/OAuth/search/theme deps) is heavy, yet the common case only needs
+	// to know whether the stored setup version is current. Lazy-load the wizard
+	// barrel only when setup is stale or forced; otherwise skip it entirely.
+	const storedSetupVersion = settings.get("setupVersion");
+	const setupWizard =
+		forceSetupWizard || storedSetupVersion < CURRENT_SETUP_VERSION ? await import("./modes/setup-wizard") : undefined;
+	const setupScenes = setupWizard
+		? await setupWizard.selectSetupScenes(storedSetupVersion, setupWizard.ALL_SCENES, mode, {
+				resuming,
+				isTTY: process.stdin.isTTY && process.stdout.isTTY,
+				setupWizardEnabled: settings.get("startup.setupWizard"),
+				force: forceSetupWizard,
+			})
+		: [];
+
+	await mode.init({
+		suppressWelcomeIntro: resuming || setupScenes.length > 0,
+		clearInitialTerminalHistory: true,
 	});
 
-	await mode.init({ suppressWelcomeIntro: setupScenes.length > 0 });
-
-	if (setupScenes.length > 0) {
-		await runSetupWizard(mode, setupScenes);
+	if (setupWizard && setupScenes.length > 0) {
+		await setupWizard.runSetupWizard(mode, setupScenes);
 	}
 
 	versionCheckPromise
@@ -285,12 +298,11 @@ async function runInteractiveMode(
 		})
 		.catch(() => {});
 
-	// Cold-launch cleanup: wipe the terminal scrollback before painting the
-	// resumed/new transcript. The TUI's initial paint deliberately preserves
-	// native scrollback (prior shell content), but on `omp`/`omp -c` that leaves
-	// the previous run's welcome + transcript stacked above the fresh one. Every
-	// in-process session load already clears via `clearTerminalHistory`; the cold
-	// launch is the lone path that did not.
+	// Cold-launch cleanup: the first paint already clears native history, and this
+	// replay replaces the welcome/startup frame with the resumed/new transcript.
+	// Every in-process session load also uses `clearTerminalHistory`; cold launch
+	// follows the same clean-cutover path instead of preserving a previous run's
+	// transcript above the fresh one.
 	mode.renderInitialMessages(undefined, { preserveExistingChat: true, clearTerminalHistory: true });
 
 	for (const notify of notifs) {
@@ -716,7 +728,7 @@ async function buildSessionOptions(
 interface RunRootCommandDependencies {
 	createAgentSession?: typeof createAgentSession;
 	discoverAuthStorage?: typeof discoverAuthStorage;
-	runAcpMode?: typeof runAcpMode;
+	runAcpMode?: RunAcpMode;
 	settings?: Settings;
 	forceSetupWizard?: boolean;
 }
@@ -773,6 +785,7 @@ export async function runRootCommand(
 		let result: string;
 		try {
 			const outputPath = parsedArgs.messages.length > 0 ? parsedArgs.messages[0] : undefined;
+			const { exportFromFile } = await import("./export/html");
 			result = await exportFromFile(parsedArgs.export, outputPath);
 		} catch (error: unknown) {
 			const message = error instanceof Error ? error.message : "Failed to export session";
@@ -940,33 +953,11 @@ export async function runRootCommand(
 
 	await pluginPreloadPromise;
 
-	// Background marketplace auto-update — never blocks startup.
-	const autoUpdate = settingsInstance.get("marketplace.autoUpdate");
-	if (autoUpdate !== "off") {
-		void (async () => {
-			try {
-				const mgr = new MarketplaceManager({
-					marketplacesRegistryPath: getMarketplacesRegistryPath(),
-					installedRegistryPath: getInstalledPluginsRegistryPath(),
-					projectInstalledRegistryPath: (await resolveActiveProjectRegistryPath(getProjectDir())) ?? undefined,
-					marketplacesCacheDir: getMarketplacesCacheDir(),
-					pluginsCacheDir: getPluginsCacheDir(),
-					clearPluginRootsCache: clearPluginRootsAndCaches,
-				});
-				await mgr.refreshStaleMarketplaces();
-				const updates = await mgr.checkForUpdates();
-				if (updates.length === 0) return;
-				if (autoUpdate === "auto") {
-					await mgr.upgradeAllPlugins();
-					logger.debug(`Auto-upgraded ${updates.length} marketplace plugin(s)`);
-				} else {
-					logger.debug(`${updates.length} marketplace plugin update(s) available — /marketplace upgrade`);
-				}
-			} catch {
-				// Silently ignore — network failure, corrupt data, offline.
-			}
-		})();
-	}
+	scheduleMarketplaceAutoUpdate({
+		autoUpdate: settingsInstance.get("marketplace.autoUpdate"),
+		resolveActiveProjectRegistryPath,
+		clearPluginRootsCache: clearPluginRootsAndCaches,
+	});
 
 	const { options: sessionOptions } = await logger.time(
 		"buildSessionOptions",
@@ -988,7 +979,7 @@ export async function runRootCommand(
 	// Both are no-ops when OTEL_EXPORTER_OTLP_ENDPOINT is unset. An empty config
 	// is enough to enable telemetry — content capture is governed by the
 	// standard OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT env var.
-	initTelemetryExport();
+	await initTelemetryExport();
 	if (isTelemetryExportEnabled()) {
 		sessionOptions.telemetry = {};
 	}
@@ -1027,7 +1018,9 @@ export async function runRootCommand(
 			rawArgs,
 			createSession,
 		});
-		await (deps.runAcpMode ?? runAcpMode)(createAcpSession);
+		// Branch-only protocol runner: keep ACP server code out of normal interactive startup.
+		const runAcpMode = deps.runAcpMode ?? (await import("./modes/acp/acp-mode")).runAcpMode;
+		await runAcpMode(createAcpSession);
 	} else {
 		// Resolve extension-registered CLI flags before creating the session so a
 		// bad `@file` fails fast WITHOUT leaving a junk session/breadcrumb
@@ -1091,6 +1084,8 @@ export async function runRootCommand(
 		}
 
 		if (mode === "rpc" || mode === "rpc-ui") {
+			// Branch-only protocol runner: keep RPC host code out of normal interactive startup.
+			const runRpcMode: RunRpcMode = (await import("./modes/rpc/rpc-mode")).runRpcMode;
 			await runRpcMode(session, mode === "rpc-ui" ? setToolUIContext : undefined);
 		} else if (isInteractive) {
 			const versionCheckPromise = checkForNewVersion(VERSION).catch(() => undefined);
@@ -1109,7 +1104,7 @@ export async function runRootCommand(
 
 			if ($env.PI_TIMING) {
 				logger.printTimings();
-				if ($env.PI_TIMING === "x") {
+				if (logger.shouldExitAfterTimings()) {
 					process.exit(0);
 				}
 			}
@@ -1132,6 +1127,8 @@ export async function runRootCommand(
 				initialImages,
 			);
 		} else {
+			// Branch-only single-shot runner: keep print-mode code out of normal interactive startup.
+			const runPrintMode: RunPrintMode = (await import("./modes/print-mode")).runPrintMode;
 			await runPrintMode(session, {
 				mode,
 				messages: initialArgs.messages,
diff --git a/packages/coding-agent/src/memories/index.ts b/packages/coding-agent/src/memories/index.ts
index ae27abbe1..38205c5e2 100644
--- a/packages/coding-agent/src/memories/index.ts
+++ b/packages/coding-agent/src/memories/index.ts
@@ -3,8 +3,9 @@ import type * as fsNode from "node:fs";
 import * as fs from "node:fs/promises";
 import * as path from "node:path";
 import type { AgentMessage } from "@oh-my-pi/pi-agent-core";
-import { clampThinkingLevelForModel, completeSimple, Effort, type Model } from "@oh-my-pi/pi-ai";
+import { type ApiKey, clampThinkingLevelForModel, completeSimple, Effort, type Model } from "@oh-my-pi/pi-ai";
 import { getAgentDbPath, getMemoriesDir, logger, parseJsonlLenient, prompt } from "@oh-my-pi/pi-utils";
+
 import type { ModelRegistry } from "../config/model-registry";
 import { resolveModelRoleValue } from "../config/model-resolver";
 import type { Settings } from "../config/settings";
@@ -271,7 +272,10 @@ async function runPhase1(options: {
 			const result = await runStage1Job({
 				claim,
 				model: phase1Model,
-				apiKey: phase1ApiKey,
+				apiKey: modelRegistry.resolver(phase1Model.provider, {
+					sessionId: session.sessionId,
+					baseUrl: phase1Model.baseUrl,
+				}),
 				modelMaxTokens: computeModelTokenBudget(phase1Model, config),
 				config,
 				metadata: session.agent?.metadataForProvider(phase1Model.provider),
@@ -428,7 +432,10 @@ async function runPhase2(options: {
 			const consolidated = await runConsolidationModel({
 				memoryRoot,
 				model: phase2Model,
-				apiKey: phase2ApiKey,
+				apiKey: modelRegistry.resolver(phase2Model.provider, {
+					sessionId: session.sessionId,
+					baseUrl: phase2Model.baseUrl,
+				}),
 				metadata: session.agent?.metadataForProvider(phase2Model.provider),
 			});
 			await applyConsolidation(memoryRoot, consolidated);
@@ -574,7 +581,7 @@ function extractPersistableMessages(payload: string): AgentMessage[] {
 async function runStage1Job(options: {
 	claim: Stage1Claim;
 	model: Model;
-	apiKey: string;
+	apiKey: ApiKey;
 	modelMaxTokens: number;
 	config: MemoryRuntimeConfig;
 	metadata?: Record<string, unknown>;
@@ -718,7 +725,7 @@ async function readRolloutSummaries(memoryRoot: string): Promise<string> {
 async function runConsolidationModel(options: {
 	memoryRoot: string;
 	model: Model;
-	apiKey: string;
+	apiKey: ApiKey;
 	metadata?: Record<string, unknown>;
 }): Promise<{
 	memoryMd: string;
diff --git a/packages/coding-agent/src/memory-backend/index.ts b/packages/coding-agent/src/memory-backend/index.ts
index 2c2c678a3..62f504ee4 100644
--- a/packages/coding-agent/src/memory-backend/index.ts
+++ b/packages/coding-agent/src/memory-backend/index.ts
@@ -1,4 +1,16 @@
-export * from "../mnemopi";
+export type {
+	MnemopiBackendConfig,
+	MnemopiLlmMode,
+	MnemopiProviderOptions,
+	MnemopiScoping,
+} from "../mnemopi/config";
+export type {
+	MnemopiMemoryEditOperation,
+	MnemopiMemoryEditOptions,
+	MnemopiMemoryEditResult,
+	MnemopiSessionState,
+	MnemopiSessionStateOptions,
+} from "../mnemopi/state";
 export * from "./local-backend";
 export * from "./off-backend";
 export * from "./resolve";
diff --git a/packages/coding-agent/src/memory-backend/resolve.ts b/packages/coding-agent/src/memory-backend/resolve.ts
index a0066d7f9..638aabec1 100644
--- a/packages/coding-agent/src/memory-backend/resolve.ts
+++ b/packages/coding-agent/src/memory-backend/resolve.ts
@@ -1,6 +1,4 @@
 import type { Settings } from "../config/settings";
-import { hindsightBackend } from "../hindsight";
-import { mnemopiBackend } from "../mnemopi";
 import { localBackend } from "./local-backend";
 import { offBackend } from "./off-backend";
 import type { MemoryBackend } from "./types";
@@ -18,10 +16,10 @@ import type { MemoryBackend } from "./types";
  * `memories.enabled` remains accepted only as a legacy migration input. Once
  * a config is loaded, `memory.backend` is the sole runtime selector.
  */
-export function resolveMemoryBackend(settings: Settings): MemoryBackend {
+export async function resolveMemoryBackend(settings: Settings): Promise<MemoryBackend> {
 	const id = settings.get("memory.backend");
-	if (id === "hindsight") return hindsightBackend;
-	if (id === "mnemopi") return mnemopiBackend;
+	if (id === "hindsight") return (await import("../hindsight/backend")).hindsightBackend;
+	if (id === "mnemopi") return (await import("../mnemopi/backend")).mnemopiBackend;
 	if (id === "local") return localBackend;
 	return offBackend;
 }
diff --git a/packages/coding-agent/src/memory-backend/types.ts b/packages/coding-agent/src/memory-backend/types.ts
index 3d72976ea..8b2e5cd15 100644
--- a/packages/coding-agent/src/memory-backend/types.ts
+++ b/packages/coding-agent/src/memory-backend/types.ts
@@ -1,7 +1,7 @@
 /**
  * Memory backend abstraction.
  *
- * Backends are mutually exclusive — `resolveMemoryBackend(settings)` returns
+ * Backends are mutually exclusive — `await resolveMemoryBackend(settings)` resolves
  * exactly one. Implementations MUST be self-contained: they own the per-session
  * state they create in `start()` and tear it down on `clear()`.
  */
diff --git a/packages/coding-agent/src/mnemopi/backend.ts b/packages/coding-agent/src/mnemopi/backend.ts
index 167c37250..7eb6d531d 100644
--- a/packages/coding-agent/src/mnemopi/backend.ts
+++ b/packages/coding-agent/src/mnemopi/backend.ts
@@ -5,6 +5,7 @@ import { Mnemopi } from "@oh-my-pi/pi-mnemopi";
 import { BankManager } from "@oh-my-pi/pi-mnemopi/core";
 import { type DiagnosticSummary, inspectDatabase } from "@oh-my-pi/pi-mnemopi/diagnose";
 import { logger } from "@oh-my-pi/pi-utils";
+
 import type { ModelRegistry } from "../config/model-registry";
 import { resolveRoleSelection } from "../config/model-resolver";
 import type { MemoryBackend, MemoryBackendStartOptions } from "../memory-backend/types";
@@ -334,7 +335,10 @@ async function resolveMnemopiProviderOptions(
 						messages: [{ role: "user", content: prompt, timestamp: Date.now() }],
 					},
 					{
-						apiKey,
+						apiKey: modelRegistry.resolver(model.provider, {
+							sessionId,
+							baseUrl: model.baseUrl,
+						}),
 						maxTokens: opts?.maxTokens,
 						temperature: opts?.temperature,
 					},
diff --git a/packages/coding-agent/src/modes/acp/acp-agent.ts b/packages/coding-agent/src/modes/acp/acp-agent.ts
index 4bd991d51..efa8bc99c 100644
--- a/packages/coding-agent/src/modes/acp/acp-agent.ts
+++ b/packages/coding-agent/src/modes/acp/acp-agent.ts
@@ -1,3 +1,4 @@
+import * as fs from "node:fs/promises";
 import * as path from "node:path";
 import {
 	type Agent,
@@ -62,7 +63,7 @@ import { MCPManager } from "../../mcp/manager";
 import type { MCPServerConfig } from "../../mcp/types";
 import { loadAllExtensions } from "../../modes/components/extensions/state-manager";
 import { theme } from "../../modes/theme/theme";
-import { type PlanApprovalDetails, renameApprovedPlanFile, resolvePlanTitle } from "../../plan-mode/approved-plan";
+import { type PlanApprovalDetails, resolveApprovedPlan } from "../../plan-mode/approved-plan";
 import type { AgentSession, AgentSessionEvent } from "../../session/agent-session";
 import { isSilentAbort, SKILL_PROMPT_MESSAGE_TYPE } from "../../session/messages";
 import {
@@ -1425,24 +1426,16 @@ export class AcpAgent implements Agent {
 				if (!state?.enabled) {
 					throw new ToolError("Plan mode is not active.");
 				}
-				const planFilePath = state.planFilePath;
-				const planContent = await this.#readAcpPlanFile(session, planFilePath);
-				if (planContent === null) {
-					throw new ToolError(
-						`Plan file not found at ${planFilePath}. Write the finalized plan to ${planFilePath} before requesting approval.`,
-					);
-				}
-				const normalized = resolvePlanTitle({
+				const { planFilePath, planContent, title } = await resolveApprovedPlan({
 					suppliedTitle: extra?.title,
-					planContent,
-					planFilePath,
+					statePlanFilePath: state.planFilePath,
+					readPlan: url => this.#readAcpPlanFile(session, url),
+					listPlanFiles: () => this.#listAcpLocalPlanFiles(session),
 				});
-				const finalPlanFilePath = `local://${normalized.fileName}`;
-				const approved = await this.#requestAcpPlanApprovalChoice(session.sessionId, normalized.title, planContent);
+				const approved = await this.#requestAcpPlanApprovalChoice(session.sessionId, title, planContent);
 				const details: PlanApprovalDetails = {
 					planFilePath,
-					finalPlanFilePath,
-					title: normalized.title,
+					title,
 					planExists: true,
 				};
 				if (!approved) {
@@ -1458,16 +1451,10 @@ export class AcpAgent implements Agent {
 						details,
 					};
 				}
-				// Approved. Rename plan to its titled filename, set the plan
-				// reference so the next turn injects the plan content as
-				// context, then exit plan mode so the agent regains full tools.
-				await renameApprovedPlanFile({
-					planFilePath,
-					finalPlanFilePath,
-					getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
-					getSessionId: () => session.sessionManager.getSessionId(),
-				});
-				session.setPlanReferencePath(finalPlanFilePath);
+				// Approved. Set the plan reference so the next turn injects the plan
+				// content as context (the file keeps its agent-chosen name — no
+				// rename), then exit plan mode so the agent regains full tools.
+				session.setPlanReferencePath(planFilePath);
 				session.setStandingResolveHandler?.(null);
 				session.setPlanModeState(undefined);
 				try {
@@ -1486,7 +1473,7 @@ export class AcpAgent implements Agent {
 					content: [
 						{
 							type: "text" as const,
-							text: `Plan approved at ${finalPlanFilePath}. Plan mode exited; proceed with the implementation.`,
+							text: `Plan approved at ${planFilePath}. Plan mode exited; proceed with the implementation.`,
 						},
 					],
 					details,
@@ -1518,6 +1505,26 @@ export class AcpAgent implements Agent {
 		}
 	}
 
+	/** `local://` URLs of plan files in the session-local root, newest first —
+	 *  the `resolveApprovedPlan` fallback for a dropped `extra.title`. */
+	async #listAcpLocalPlanFiles(session: AgentSession): Promise<string[]> {
+		const localRoot = this.#resolveAcpPlanFilePath(session, "local://");
+		try {
+			const entries = await fs.readdir(localRoot, { withFileTypes: true });
+			const plans = await Promise.all(
+				entries
+					.filter(entry => entry.isFile() && /plan\.md$/i.test(entry.name))
+					.map(async entry => {
+						const stat = await fs.stat(path.join(localRoot, entry.name)).catch(() => null);
+						return { url: `local://${entry.name}`, mtime: stat?.mtimeMs ?? 0 };
+					}),
+			);
+			return plans.sort((a, b) => b.mtime - a.mtime).map(plan => plan.url);
+		} catch {
+			return [];
+		}
+	}
+
 	/**
 	 * Ask the ACP client to confirm plan approval. Returns `true` only on an
 	 * explicit `APPROVE_OPTION` selection. Refine, dismissal (`undefined`), or
diff --git a/packages/coding-agent/src/modes/components/agent-dashboard.ts b/packages/coding-agent/src/modes/components/agent-dashboard.ts
index 158841d12..0520e041b 100644
--- a/packages/coding-agent/src/modes/components/agent-dashboard.ts
+++ b/packages/coding-agent/src/modes/components/agent-dashboard.ts
@@ -27,6 +27,7 @@ import {
 	matchesKey,
 	padding,
 	replaceTabs,
+	ScrollView,
 	Spacer,
 	Text,
 	truncateToWidth,
@@ -205,9 +206,12 @@ class AgentListPane implements Component {
 			return lines;
 		}
 
+		const overflow = this.agents.length > this.maxVisible;
+		const rowWidth = Math.max(0, width - (overflow ? 1 : 0));
 		const start = this.scrollOffset;
 		const end = Math.min(start + this.maxVisible, this.agents.length);
 
+		const rows: string[] = [];
 		for (let i = start; i < end; i++) {
 			const agent = this.agents[i];
 			const selected = i === this.selectedIndex;
@@ -224,12 +228,17 @@ class AgentListPane implements Component {
 				line = theme.fg("dim", line);
 			}
 
-			lines.push(truncateToWidth(line, width));
+			rows.push(truncateToWidth(line, rowWidth));
 		}
 
-		if (this.agents.length > this.maxVisible) {
-			lines.push(theme.fg("muted", `  (${this.selectedIndex + 1}/${this.agents.length})`));
-		}
+		const sv = new ScrollView(rows, {
+			height: rows.length,
+			scrollbar: "auto",
+			totalRows: this.agents.length,
+			theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+		});
+		sv.setScrollOffset(this.scrollOffset);
+		lines.push(...sv.render(width));
 
 		return lines;
 	}
diff --git a/packages/coding-agent/src/modes/components/assistant-message.ts b/packages/coding-agent/src/modes/components/assistant-message.ts
index 17e9ab0b8..5fac18c12 100644
--- a/packages/coding-agent/src/modes/components/assistant-message.ts
+++ b/packages/coding-agent/src/modes/components/assistant-message.ts
@@ -4,7 +4,7 @@ import { formatNumber } from "@oh-my-pi/pi-utils";
 import { settings } from "../../config/settings";
 import type { AssistantThinkingRenderer } from "../../extensibility/extensions/types";
 import { getMarkdownTheme, theme } from "../../modes/theme/theme";
-import { isSilentAbort } from "../../session/messages";
+import { isSilentAbort, resolveAbortLabel } from "../../session/messages";
 import { resolveImageOptions } from "../../tools/render-utils";
 
 /**
@@ -18,6 +18,15 @@ export class AssistantMessageComponent extends Container {
 	#convertedKittyImages = new Map<string, ImageContent>();
 	#kittyConversionsInFlight = new Set<string>();
 	#transcriptBlockFinalized: boolean;
+	/**
+	 * When true, the turn-ending `Error: …` line for `stopReason === "error"` is
+	 * suppressed because the same error is currently shown in the pinned banner
+	 * above the editor (see `EventController` + `ErrorBannerComponent`). Avoids
+	 * rendering the identical error twice (inline + banner) at the error moment.
+	 * Restored to `false` when the banner is cleared at the next turn so the
+	 * transcript keeps the error in history.
+	 */
+	#errorPinned = false;
 
 	constructor(
 		message?: AssistantMessage,
@@ -49,6 +58,18 @@ export class AssistantMessageComponent extends Container {
 		this.hideThinkingBlock = hide;
 	}
 
+	/**
+	 * Toggle suppression of the inline `Error: …` line while the same error is
+	 * pinned in the banner above the editor. Re-renders so the change is visible.
+	 */
+	setErrorPinned(pinned: boolean): void {
+		if (this.#errorPinned === pinned) return;
+		this.#errorPinned = pinned;
+		if (this.#lastMessage) {
+			this.updateContent(this.#lastMessage);
+		}
+	}
+
 	isTranscriptBlockFinalized(): boolean {
 		return this.#transcriptBlockFinalized;
 	}
@@ -187,10 +208,6 @@ export class AssistantMessageComponent extends Container {
 			c => (c.type === "text" && c.text.trim()) || (c.type === "thinking" && c.thinking.trim()),
 		);
 
-		if (hasVisibleContent) {
-			this.#contentContainer.addChild(new Spacer(1));
-		}
-
 		// Render content in order
 		let thinkingIndex = 0;
 		for (let i = 0; i < message.content.length; i++) {
@@ -236,17 +253,14 @@ export class AssistantMessageComponent extends Container {
 		const hasToolCalls = message.content.some(c => c.type === "toolCall");
 		if (!hasToolCalls) {
 			if (message.stopReason === "aborted" && !isSilentAbort(message.errorMessage)) {
-				const abortMessage =
-					message.errorMessage && message.errorMessage !== "Request was aborted"
-						? message.errorMessage
-						: "Operation aborted";
+				const abortMessage = resolveAbortLabel(message.errorMessage);
 				if (hasVisibleContent) {
 					this.#contentContainer.addChild(new Spacer(1));
 				} else {
 					this.#contentContainer.addChild(new Spacer(1));
 				}
 				this.#contentContainer.addChild(new Text(theme.fg("error", abortMessage), 1, 0));
-			} else if (message.stopReason === "error") {
+			} else if (message.stopReason === "error" && !this.#errorPinned) {
 				const errorMsg = message.errorMessage || "Unknown error";
 				this.#contentContainer.addChild(new Spacer(1));
 				this.#contentContainer.addChild(new Text(theme.fg("error", `Error: ${errorMsg}`), 1, 0));
diff --git a/packages/coding-agent/src/modes/components/chat-block.ts b/packages/coding-agent/src/modes/components/chat-block.ts
new file mode 100644
index 000000000..da79c2041
--- /dev/null
+++ b/packages/coding-agent/src/modes/components/chat-block.ts
@@ -0,0 +1,111 @@
+import { Container } from "@oh-my-pi/pi-tui";
+
+/**
+ * Capabilities a mounted {@link ChatBlock} may use against its host transcript.
+ * Kept minimal so blocks never reach into the full TUI/InteractiveMode surface.
+ */
+export interface ChatBlockHost {
+	/** Schedule a repaint of the transcript. */
+	requestRender(): void;
+}
+
+/**
+ * Lifecycle-aware transcript block — the "return a block, let the host mount it"
+ * primitive, modelled on React/Svelte component lifecycles.
+ *
+ * Producers build and return a `ChatBlock` instead of poking `chatContainer` and
+ * `ui.requestRender()` directly. The host (`ctx.present`) appends it and calls
+ * {@link mount}, which runs {@link onMount}; effects started there register
+ * teardown via {@link onCleanup}. The block repaints through {@link requestRender}
+ * — never touching the TUI — and tears down exactly once on {@link finish}
+ * (self-complete: stop the animation, keep the final frame in the transcript) or
+ * {@link dispose} (host discards it, e.g. a transcript reset).
+ *
+ * While mounted and unfinished a block reports `isTranscriptBlockFinalized() ===
+ * false` so {@link "../components/transcript-container".TranscriptContainer}
+ * keeps it in the live, repaintable region on ED3-risk terminals; after
+ * `finish()`/`dispose()` it reports `true` and freezes at its final content.
+ */
+export abstract class ChatBlock extends Container {
+	#host: ChatBlockHost | undefined;
+	#cleanups: Array<() => void> = [];
+	#active = false;
+	#disposed = false;
+
+	/**
+	 * Run setup after the block is in the transcript: start timers/subscriptions
+	 * and register their teardown with {@link onCleanup}. Default: no-op (a block
+	 * whose content is fixed at construction needs no mount work).
+	 */
+	protected onMount(): void {}
+
+	/**
+	 * Register a teardown to run on {@link finish}/{@link dispose}, à la a
+	 * `useEffect` cleanup. If the block is already disposed the cleanup runs
+	 * immediately so callers never leak.
+	 */
+	protected onCleanup(cleanup: () => void): void {
+		if (this.#disposed) {
+			cleanup();
+			return;
+		}
+		this.#cleanups.push(cleanup);
+	}
+
+	/** Ask the host to repaint. No-op before mount or after dispose. */
+	protected requestRender(): void {
+		this.#host?.requestRender();
+	}
+
+	/** True between {@link mount} and {@link finish}/{@link dispose}. */
+	protected get active(): boolean {
+		return this.#active;
+	}
+
+	/**
+	 * Host-only: attach the host and run {@link onMount}. Idempotent — a second
+	 * call (e.g. a transcript rebuild that re-presents the same instance) is a
+	 * no-op.
+	 */
+	mount(host: ChatBlockHost): void {
+		if (this.#host || this.#disposed) return;
+		this.#host = host;
+		this.#active = true;
+		this.onMount();
+	}
+
+	/**
+	 * Self-complete: stop ongoing effects and freeze the block at its current
+	 * content, leaving it rendered in the transcript. Use when the operation the
+	 * block represents finishes (connection resolved, download done).
+	 */
+	finish(): void {
+		if (!this.#active) return;
+		this.#active = false;
+		this.#runCleanups();
+		this.requestRender();
+	}
+
+	/**
+	 * Host-only teardown: release everything and propagate to children. Called
+	 * when the host permanently discards the block (transcript reset). Idempotent.
+	 */
+	override dispose(): void {
+		if (this.#disposed) return;
+		this.#disposed = true;
+		this.#active = false;
+		this.#runCleanups();
+		super.dispose();
+		this.#host = undefined;
+	}
+
+	/** Live blocks stay repaintable; finished/disposed ones may freeze. */
+	isTranscriptBlockFinalized(): boolean {
+		return !this.#active;
+	}
+
+	#runCleanups(): void {
+		const cleanups = this.#cleanups.splice(0);
+		for (const cleanup of cleanups) cleanup();
+	}
+}
diff --git a/packages/coding-agent/src/modes/components/copy-selector.ts b/packages/coding-agent/src/modes/components/copy-selector.ts
new file mode 100644
index 000000000..ebc1f64d3
--- /dev/null
+++ b/packages/coding-agent/src/modes/components/copy-selector.ts
@@ -0,0 +1,206 @@
+import { type Component, matchesKey, padding, Text, truncateToWidth, visibleWidth } from "@oh-my-pi/pi-tui";
+import { replaceTabs } from "../../tools/render-utils";
+import { highlightCode, theme } from "../theme/theme";
+import type { CopyTarget } from "../utils/copy-targets";
+import {
+	matchesSelectCancel,
+	matchesSelectDown,
+	matchesSelectPageDown,
+	matchesSelectPageUp,
+	matchesSelectUp,
+} from "../utils/keybinding-matchers";
+import { keyHint, rawKeyHint } from "./keybinding-hints";
+import { bottomBorder, divider, row, topBorder } from "./overlay-box";
+
+/** Minimum rows reserved for the tree even on short terminals. */
+const MIN_TREE_ROWS = 3;
+/** Fixed chrome rows: top border, two dividers, footer, bottom border. */
+const CHROME_ROWS = 5;
+
+export interface CopySelectorCallbacks {
+	/** A copy target was chosen — copy its `content`. */
+	onPick: (target: CopyTarget) => void;
+	/** The picker was dismissed. */
+	onCancel: () => void;
+}
+
+interface FlatNode {
+	target: CopyTarget;
+	depth: number;
+	/** Last among its siblings (drives └─ vs ├─). */
+	isLast: boolean;
+	/** Per-ancestor flag: does ancestor at that level have a following sibling? */
+	ancestorHasNext: boolean[];
+}
+
+/** Render one tree connector as exactly three cells (e.g. "├─ ", "└─ ", "|--"). */
+function connectorCells(symbol: string): string {
+	const chars = Array.from(symbol);
+	return (chars[0] ?? " ") + (chars[1] ?? theme.tree.horizontal) + (chars[2] ?? " ");
+}
+
+/** The 3-cell ancestor gutter: a vertical guide when the ancestor continues. */
+function gutterCells(hasNext: boolean): string {
+	return `${hasNext ? theme.tree.vertical : " "}  `;
+}
+
+/**
+ * Fullscreen `/copy` picker rendered as a `/tree`-style tree inside one
+ * outlined box: a title, the tree of copy targets (recent assistant messages
+ * with their code blocks nested beneath), a live preview of the highlighted
+ * node, and a keybinding footer. Every node copies its `content` on Enter.
+ */
+export class CopySelectorComponent implements Component {
+	#roots: CopyTarget[];
+	#cursorId: string;
+	#treeRows = MIN_TREE_ROWS;
+	// Reused across renders to wrap preview content to the pane width.
+	#previewText = new Text("", 0, 0);
+
+	constructor(
+		roots: CopyTarget[],
+		private readonly callbacks: CopySelectorCallbacks,
+	) {
+		this.#roots = roots;
+		this.#cursorId = roots[0]?.id ?? "";
+	}
+
+	invalidate(): void {}
+
+	#flatten(): FlatNode[] {
+		const out: FlatNode[] = [];
+		const walk = (nodes: CopyTarget[], depth: number, ancestorHasNext: boolean[]) => {
+			nodes.forEach((target, i) => {
+				const isLast = i === nodes.length - 1;
+				out.push({ target, depth, isLast, ancestorHasNext });
+				if (target.children?.length) walk(target.children, depth + 1, [...ancestorHasNext, !isLast]);
+			});
+		};
+		walk(this.#roots, 0, []);
+		return out;
+	}
+
+	handleInput(keyData: string): void {
+		if (matchesSelectCancel(keyData)) {
+			this.callbacks.onCancel();
+			return;
+		}
+
+		const flat = this.#flatten();
+		if (flat.length === 0) return;
+		const idx = Math.max(
+			0,
+			flat.findIndex(n => n.target.id === this.#cursorId),
+		);
+
+		if (matchesSelectUp(keyData)) {
+			this.#cursorId = flat[idx === 0 ? flat.length - 1 : idx - 1]!.target.id;
+		} else if (matchesSelectDown(keyData)) {
+			this.#cursorId = flat[idx === flat.length - 1 ? 0 : idx + 1]!.target.id;
+		} else if (matchesSelectPageUp(keyData)) {
+			this.#cursorId = flat[Math.max(0, idx - this.#treeRows)]!.target.id;
+		} else if (matchesSelectPageDown(keyData)) {
+			this.#cursorId = flat[Math.min(flat.length - 1, idx + this.#treeRows)]!.target.id;
+		} else if (matchesKey(keyData, "enter") || matchesKey(keyData, "return") || keyData === "\n") {
+			const target = flat[idx]!.target;
+			if (target.content !== undefined) this.callbacks.onPick(target);
+		}
+	}
+
+	#renderTree(width: number, flat: FlatNode[], cursorIdx: number, rows: number): string[] {
+		const inner = Math.max(0, width - 4);
+		const start = Math.max(0, Math.min(cursorIdx - Math.floor(rows / 2), Math.max(0, flat.length - rows)));
+		const out: string[] = [];
+		for (let r = 0; r < rows; r++) {
+			const i = start + r;
+			const node = flat[i];
+			if (!node) {
+				out.push(row("", width));
+				continue;
+			}
+			const target = node.target;
+			const isSelected = i === cursorIdx;
+
+			let prefix = "";
+			for (let l = 0; l < node.depth - 1; l++) prefix += gutterCells(node.ancestorHasNext[l]!);
+			if (node.depth > 0) prefix += connectorCells(node.isLast ? theme.tree.last : theme.tree.branch);
+
+			const cursor = isSelected ? "❯ " : "  ";
+			const hint = target.hint ?? "";
+			const hintWidth = hint ? visibleWidth(hint) + 2 : 0;
+			const used = visibleWidth(cursor) + visibleWidth(prefix);
+			const labelPlain = truncateToWidth(target.label, Math.max(1, inner - used - hintWidth));
+			const left = isSelected
+				? theme.fg("accent", cursor) + theme.fg("dim", prefix) + theme.bold(theme.fg("accent", labelPlain))
+				: cursor + theme.fg("dim", prefix) + labelPlain;
+			const gap = Math.max(1, inner - used - visibleWidth(labelPlain) - visibleWidth(hint));
+			out.push(row(left + padding(gap) + (hint ? theme.fg("dim", hint) : ""), width));
+		}
+		return out;
+	}
+
+	#renderPreview(width: number, target: CopyTarget | undefined, rows: number): string[] {
+		const out: string[] = [];
+		const hint = target?.hint;
+		out.push(row(theme.fg("dim", `Preview${hint ? ` · ${hint}` : ""}`), width));
+
+		const contentRows = rows - 1;
+		if (!target || contentRows <= 0) {
+			while (out.length < rows) out.push(row("", width));
+			return out;
+		}
+
+		// Code/command previews are syntax-highlighted; everything else is shown
+		// as plain text. Both are wrapped (not hard-truncated) to the pane width.
+		const isCode = target.language !== undefined;
+		const source = isCode
+			? highlightCode(replaceTabs(target.preview), target.language).join("\n")
+			: replaceTabs(target.preview);
+		this.#previewText.setText(source);
+		const wrapped = this.#previewText.render(Math.max(1, width - 4));
+
+		const hasMore = wrapped.length > contentRows;
+		const visibleCount = hasMore ? contentRows - 1 : Math.min(wrapped.length, contentRows);
+		for (let k = 0; k < contentRows; k++) {
+			if (k < visibleCount) {
+				out.push(row(isCode ? wrapped[k]! : theme.fg("muted", wrapped[k]!), width));
+			} else if (k === visibleCount && hasMore) {
+				out.push(row(theme.fg("dim", `… ${wrapped.length - visibleCount} more lines`), width));
+			} else {
+				out.push(row("", width));
+			}
+		}
+		return out;
+	}
+
+	render(width: number): string[] {
+		const height = process.stdout.rows || 40;
+		const flat = this.#flatten();
+		const cursorIdx = Math.max(
+			0,
+			flat.findIndex(n => n.target.id === this.#cursorId),
+		);
+		const selected = flat[cursorIdx]?.target;
+
+		const available = Math.max(MIN_TREE_ROWS + 1, height - CHROME_ROWS);
+		const treeRows = Math.max(1, Math.min(flat.length, Math.floor(available / 2)));
+		this.#treeRows = treeRows;
+		const previewRows = Math.max(1, available - treeRows);
+
+		const footer = [
+			rawKeyHint("↑↓", "move"),
+			keyHint("tui.select.confirm", "copy"),
+			keyHint("tui.select.cancel", "quit"),
+		].join(theme.fg("dim", " · "));
+
+		return [
+			topBorder(width, "Copy to clipboard"),
+			...this.#renderTree(width, flat, cursorIdx, treeRows),
+			divider(width),
+			...this.#renderPreview(width, selected, previewRows),
+			divider(width),
+			row(footer, width),
+			bottomBorder(width),
+		];
+	}
+}
diff --git a/packages/coding-agent/src/modes/components/custom-editor.ts b/packages/coding-agent/src/modes/components/custom-editor.ts
index 5692e7537..835ea4ec3 100644
--- a/packages/coding-agent/src/modes/components/custom-editor.ts
+++ b/packages/coding-agent/src/modes/components/custom-editor.ts
@@ -10,6 +10,7 @@ type ConfigurableEditorAction = Extract<
 	| "app.clear"
 	| "app.exit"
 	| "app.suspend"
+	| "app.display.reset"
 	| "app.thinking.cycle"
 	| "app.model.cycleForward"
 	| "app.model.cycleBackward"
@@ -30,10 +31,11 @@ const DEFAULT_ACTION_KEYS: Record<ConfigurableEditorAction, KeyId[]> = {
 	"app.clear": ["ctrl+c"],
 	"app.exit": ["ctrl+d"],
 	"app.suspend": ["ctrl+z"],
+	"app.display.reset": ["ctrl+l"],
 	"app.thinking.cycle": ["shift+tab"],
 	"app.model.cycleForward": ["ctrl+p"],
 	"app.model.cycleBackward": ["shift+ctrl+p"],
-	"app.model.select": ["ctrl+l"],
+	"app.model.select": ["alt+m"],
 	"app.model.selectTemporary": ["alt+p"],
 	"app.tools.expand": ["ctrl+o"],
 	"app.thinking.toggle": ["ctrl+t"],
@@ -45,6 +47,21 @@ const DEFAULT_ACTION_KEYS: Record<ConfigurableEditorAction, KeyId[]> = {
 	"app.clipboard.copyPrompt": ["alt+shift+c"],
 };
 
+const BRACKETED_PASTE_START = "\x1b[200~";
+const BRACKETED_PASTE_END = "\x1b[201~";
+const BRACKETED_IMAGE_PATH_REGEX = /\.(?:png|jpe?g|gif|webp)$/i;
+
+export function extractBracketedImagePastePath(data: string): string | undefined {
+	if (!data.startsWith(BRACKETED_PASTE_START)) return undefined;
+	const endIndex = data.indexOf(BRACKETED_PASTE_END, BRACKETED_PASTE_START.length);
+	if (endIndex === -1 || endIndex + BRACKETED_PASTE_END.length !== data.length) return undefined;
+
+	const pasted = data.slice(BRACKETED_PASTE_START.length, endIndex).trim();
+	if (!pasted || /[\r\n]/.test(pasted)) return undefined;
+	if (!BRACKETED_IMAGE_PATH_REGEX.test(pasted)) return undefined;
+	return pasted;
+}
+
 /**
  * Custom editor that handles configurable app-level shortcuts for coding-agent.
  */
@@ -65,6 +82,7 @@ export class CustomEditor extends Editor {
 	onEscape?: () => void;
 	onClear?: () => void;
 	onExit?: () => void;
+	onDisplayReset?: () => void;
 	onCycleThinkingLevel?: () => void;
 	onCycleModelForward?: () => void;
 	onCycleModelBackward?: () => void;
@@ -79,6 +97,8 @@ export class CustomEditor extends Editor {
 	onCopyPrompt?: () => void;
 	/** Called when the configured image-paste shortcut is pressed. */
 	onPasteImage?: () => Promise<boolean>;
+	/** Called when a bracketed paste contains exactly one image-file path. */
+	onPasteImagePath?: (path: string) => void;
 	/** Called when the configured raw text-paste shortcut is pressed. */
 	onPasteTextRaw?: () => void;
 	/** Called when the configured dequeue shortcut is pressed. */
@@ -134,6 +154,12 @@ export class CustomEditor extends Editor {
 			return;
 		}
 
+		const pastedImagePath = extractBracketedImagePastePath(data);
+		if (pastedImagePath && this.onPasteImagePath) {
+			this.onPasteImagePath(pastedImagePath);
+			return;
+		}
+
 		// Intercept configured image paste (async - fires and handles result)
 		if (this.#matchesAction(data, "app.clipboard.pasteImage") && this.onPasteImage) {
 			void this.onPasteImage();
@@ -158,6 +184,12 @@ export class CustomEditor extends Editor {
 			return;
 		}
 
+		// Intercept configured display reset shortcut
+		if (this.#matchesAction(data, "app.display.reset") && this.onDisplayReset) {
+			this.onDisplayReset();
+			return;
+		}
+
 		// Intercept configured suspend shortcut
 		if (this.#matchesAction(data, "app.suspend") && this.onSuspend) {
 			this.onSuspend();
diff --git a/packages/coding-agent/src/modes/components/custom-message.ts b/packages/coding-agent/src/modes/components/custom-message.ts
index 1dcd8aa62..87dd43011 100644
--- a/packages/coding-agent/src/modes/components/custom-message.ts
+++ b/packages/coding-agent/src/modes/components/custom-message.ts
@@ -1,5 +1,5 @@
 import type { Component } from "@oh-my-pi/pi-tui";
-import { Box, Container, Spacer } from "@oh-my-pi/pi-tui";
+import { Box, Container } from "@oh-my-pi/pi-tui";
 import type { MessageRenderer } from "../../extensibility/extensions/types";
 import { theme } from "../../modes/theme/theme";
 import type { CustomMessage } from "../../session/messages";
@@ -20,8 +20,6 @@ export class CustomMessageComponent extends Container {
 	) {
 		super();
 
-		this.addChild(new Spacer(1));
-
 		// Create box with custom background (used for default rendering)
 		this.#box = new Box(1, 1, t => theme.bg("customMessageBg", t));
 
diff --git a/packages/coding-agent/src/modes/components/execution-shared.ts b/packages/coding-agent/src/modes/components/execution-shared.ts
index 0ad34a780..1f7d21135 100644
--- a/packages/coding-agent/src/modes/components/execution-shared.ts
+++ b/packages/coding-agent/src/modes/components/execution-shared.ts
@@ -7,7 +7,7 @@
  * stay in their respective files.
  */
 
-import { type Component, Container, Loader, Spacer, Text, type TUI } from "@oh-my-pi/pi-tui";
+import { type Component, Container, Loader, Text, type TUI } from "@oh-my-pi/pi-tui";
 import { getSymbolTheme, theme } from "../../modes/theme/theme";
 import { formatTruncationMetaNotice, type TruncationMeta } from "../../tools/output-meta";
 import { DynamicBorder } from "./dynamic-border";
@@ -31,7 +31,6 @@ export function buildExecutionFrame(
 ): { contentContainer: Container; loader: Loader } {
 	const borderColor = (str: string) => theme.fg(colorKey, str);
 
-	parent.addChild(new Spacer(1));
 	parent.addChild(new DynamicBorder(borderColor));
 
 	const contentContainer = new Container();
diff --git a/packages/coding-agent/src/modes/components/extensions/extension-list.ts b/packages/coding-agent/src/modes/components/extensions/extension-list.ts
index 045ca9e30..f813ff191 100644
--- a/packages/coding-agent/src/modes/components/extensions/extension-list.ts
+++ b/packages/coding-agent/src/modes/components/extensions/extension-list.ts
@@ -10,6 +10,7 @@ import {
 	extractPrintableText,
 	matchesKey,
 	padding,
+	ScrollView,
 	truncateToWidth,
 	visibleWidth,
 } from "@oh-my-pi/pi-tui";
@@ -134,25 +135,33 @@ export class ExtensionList implements Component {
 		const startIdx = this.#scrollOffset;
 		const endIdx = Math.min(startIdx + this.#maxVisible, this.#listItems.length);
 
+		// Reserve the rightmost column for the scrollbar when overflowing
+		const overflow = this.#listItems.length > this.#maxVisible;
+		const rowWidth = Math.max(0, width - (overflow ? 1 : 0));
+
 		// Render visible items
+		const rows: string[] = [];
 		for (let i = startIdx; i < endIdx; i++) {
 			const listItem = this.#listItems[i];
 			const isSelected = this.#focused && i === this.#selectedIndex;
 
 			if (listItem.type === "master") {
-				lines.push(this.#renderMasterSwitch(listItem, isSelected, width));
+				rows.push(this.#renderMasterSwitch(listItem, isSelected, rowWidth));
 			} else if (listItem.type === "kind-header") {
-				lines.push(this.#renderKindHeader(listItem, isSelected, width));
+				rows.push(this.#renderKindHeader(listItem, isSelected, rowWidth));
 			} else {
-				lines.push(this.#renderExtensionRow(listItem.item, isSelected, width, masterDisabled));
+				rows.push(this.#renderExtensionRow(listItem.item, isSelected, rowWidth, masterDisabled));
 			}
 		}
 
-		// Scroll indicator
-		if (this.#listItems.length > this.#maxVisible) {
-			const indicator = theme.fg("muted", `  (${this.#selectedIndex + 1}/${this.#listItems.length})`);
-			lines.push(indicator);
-		}
+		const sv = new ScrollView(rows, {
+			height: rows.length,
+			scrollbar: "auto",
+			totalRows: this.#listItems.length,
+			theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+		});
+		sv.setScrollOffset(this.#scrollOffset);
+		lines.push(...sv.render(width));
 
 		return lines;
 	}
diff --git a/packages/coding-agent/src/modes/components/history-search.ts b/packages/coding-agent/src/modes/components/history-search.ts
index 5ce3887d3..feb8de9bb 100644
--- a/packages/coding-agent/src/modes/components/history-search.ts
+++ b/packages/coding-agent/src/modes/components/history-search.ts
@@ -5,6 +5,7 @@ import {
 	Input,
 	matchesKey,
 	padding,
+	ScrollView,
 	Spacer,
 	Text,
 	truncateToWidth,
@@ -115,15 +116,19 @@ class HistoryResultsList implements Component {
 		);
 		const endIndex = Math.min(startIndex + this.#maxVisible, this.#results.length);
 
+		const overflow = this.#results.length > this.#maxVisible;
+		const rowWidth = Math.max(0, width - (overflow ? 1 : 0));
+		const rows: string[] = [];
+
 		for (let i = startIndex; i < endIndex; i++) {
 			const entry = this.#results[i];
 			const isSelected = i === this.#selectedIndex;
 
 			const timeStr = relativeTime(entry.created_at);
 			const timeWidth = visibleWidth(timeStr);
-			const showTime = width >= gutterWidth + 12 + timeWidth;
+			const showTime = rowWidth >= gutterWidth + 12 + timeWidth;
 
-			const promptBudget = Math.max(4, width - gutterWidth - (showTime ? timeWidth + 1 : 0));
+			const promptBudget = Math.max(4, rowWidth - gutterWidth - (showTime ? timeWidth + 1 : 0));
 			const normalized = entry.prompt.replace(/\s+/g, " ").trim();
 			const plain = truncateToWidth(normalized, promptBudget);
 			const highlighted = highlightTokens(plain, this.#tokens);
@@ -133,21 +138,24 @@ class HistoryResultsList implements Component {
 
 			if (showTime) {
 				// Pad the prompt region so the timestamp sits flush right with a one-cell gap.
-				line = `${truncateToWidth(line, width - timeWidth - 1, Ellipsis.Unicode, true)} ${theme.fg("dim", timeStr)}`;
+				line = `${truncateToWidth(line, rowWidth - timeWidth - 1, Ellipsis.Unicode, true)} ${theme.fg("dim", timeStr)}`;
 			}
 
-			lines.push(
+			rows.push(
 				isSelected
-					? theme.bg("selectedBg", truncateToWidth(line, width, Ellipsis.Omit, true))
-					: truncateToWidth(line, width),
+					? theme.bg("selectedBg", truncateToWidth(line, rowWidth, Ellipsis.Omit, true))
+					: truncateToWidth(line, rowWidth),
 			);
 		}
 
-		if (startIndex > 0 || endIndex < this.#results.length) {
-			const scrollText = `  ${this.#selectedIndex + 1}/${this.#results.length}`;
-			lines.push(theme.fg("muted", truncateToWidth(scrollText, width, Ellipsis.Omit)));
-		}
-
+		const sv = new ScrollView(rows, {
+			height: rows.length,
+			scrollbar: "auto",
+			totalRows: this.#results.length,
+			theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+		});
+		sv.setScrollOffset(startIndex);
+		lines.push(...sv.render(width));
 		return lines;
 	}
 }
diff --git a/packages/coding-agent/src/modes/components/hook-message.ts b/packages/coding-agent/src/modes/components/hook-message.ts
index dc19d306c..0e372f160 100644
--- a/packages/coding-agent/src/modes/components/hook-message.ts
+++ b/packages/coding-agent/src/modes/components/hook-message.ts
@@ -1,5 +1,5 @@
 import type { Component } from "@oh-my-pi/pi-tui";
-import { Box, Container, Spacer } from "@oh-my-pi/pi-tui";
+import { Box, Container } from "@oh-my-pi/pi-tui";
 import type { HookMessageRenderer } from "../../extensibility/hooks/types";
 import { theme } from "../../modes/theme/theme";
 import type { HookMessage } from "../../session/messages";
@@ -23,8 +23,6 @@ export class HookMessageComponent extends Container {
 	) {
 		super();
 
-		this.addChild(new Spacer(1));
-
 		// Create box with purple background (used for default rendering)
 		this.#box = new Box(1, 1, t => theme.bg("customMessageBg", t));
 
diff --git a/packages/coding-agent/src/modes/components/model-selector.ts b/packages/coding-agent/src/modes/components/model-selector.ts
index 8841720a8..ce6eeb83c 100644
--- a/packages/coding-agent/src/modes/components/model-selector.ts
+++ b/packages/coding-agent/src/modes/components/model-selector.ts
@@ -6,6 +6,7 @@ import {
 	getKeybindings,
 	Input,
 	matchesKey,
+	ScrollView,
 	Spacer,
 	type Tab,
 	TabBar,
@@ -13,6 +14,7 @@ import {
 	type TUI,
 	visibleWidth,
 } from "@oh-my-pi/pi-tui";
+import { formatNumber } from "@oh-my-pi/pi-utils";
 import type { ModelRegistry } from "../../config/model-registry";
 import { getKnownRoleIds, getRoleInfo, MODEL_ROLE_IDS, MODEL_ROLES } from "../../config/model-registry";
 import { resolveModelRoleValue } from "../../config/model-resolver";
@@ -147,6 +149,7 @@ export class ModelSelectorComponent extends Container {
 	#tui: TUI;
 	#scopedModels: ReadonlyArray<ScopedModelItem>;
 	#temporaryOnly: boolean;
+	#currentContextTokens: number;
 
 	#menuRoleActions: MenuRoleAction[] = [];
 
@@ -172,7 +175,7 @@ export class ModelSelectorComponent extends Container {
 		scopedModels: ReadonlyArray<ScopedModelItem>,
 		onSelect: RoleSelectCallback,
 		onCancel: () => void,
-		options?: { temporaryOnly?: boolean; initialSearchInput?: string },
+		options?: { temporaryOnly?: boolean; initialSearchInput?: string; currentContextTokens?: number },
 	) {
 		super();
 
@@ -183,6 +186,9 @@ export class ModelSelectorComponent extends Container {
 		this.#onSelectCallback = onSelect;
 		this.#onCancelCallback = onCancel;
 		this.#temporaryOnly = options?.temporaryOnly ?? false;
+		const currentContextTokens = options?.currentContextTokens ?? 0;
+		this.#currentContextTokens =
+			Number.isFinite(currentContextTokens) && currentContextTokens > 0 ? Math.floor(currentContextTokens) : 0;
 		const initialSearchInput = options?.initialSearchInput;
 
 		// Initialize menu role actions (built-in + custom from settings)
@@ -215,8 +221,8 @@ export class ModelSelectorComponent extends Container {
 			this.#searchInput.setValue(initialSearchInput);
 		}
 		this.#searchInput.onSubmit = () => {
-			// Enter on search input opens menu if we have a selection
-			if (this.#filteredModels[this.#selectedIndex]) {
+			// Enter on search input opens menu if we have an enabled selection
+			if (this.#getSelectedItem()) {
 				this.#openMenu();
 			}
 		};
@@ -460,7 +466,11 @@ export class ModelSelectorComponent extends Container {
 		this.#filteredModels = models;
 		this.#canonicalModels = canonicalModels;
 		this.#filteredCanonicalModels = canonicalModels;
-		this.#selectedIndex = Math.min(this.#selectedIndex, Math.max(0, models.length - 1));
+		const visibleModels = this.#isCanonicalTab() ? canonicalModels : models;
+		this.#selectedIndex = this.#coerceSelectedIndex(
+			Math.min(this.#selectedIndex, Math.max(0, visibleModels.length - 1)),
+			visibleModels,
+		);
 	}
 
 	async #loadModels(): Promise<void> {
@@ -626,6 +636,74 @@ export class ModelSelectorComponent extends Container {
 		return this.#getActiveTabId() === CANONICAL_TAB;
 	}
 
+	#isModelOverContextLimit(model: Model): boolean {
+		const contextWindow = model.contextWindow ?? 0;
+		return this.#currentContextTokens > 0 && contextWindow > 0 && this.#currentContextTokens > contextWindow;
+	}
+
+	#isItemDisabled(item: ModelItem | CanonicalModelItem): boolean {
+		return this.#isModelOverContextLimit(item.model);
+	}
+
+	#formatContextLimitSuffix(model: Model): string {
+		if (!this.#isModelOverContextLimit(model)) {
+			return "";
+		}
+		return ` ${theme.status.disabled} context>${formatNumber(model.contextWindow).toLowerCase()}`;
+	}
+
+	#getVisibleItems(): ReadonlyArray<ModelItem | CanonicalModelItem> {
+		return this.#isCanonicalTab() ? this.#filteredCanonicalModels : this.#filteredModels;
+	}
+
+	#coerceSelectedIndex(
+		index: number,
+		visibleItems: ReadonlyArray<ModelItem | CanonicalModelItem> = this.#getVisibleItems(),
+	): number {
+		const maxIndex = visibleItems.length - 1;
+		if (maxIndex < 0) {
+			return 0;
+		}
+		const clamped = Math.max(0, Math.min(index, maxIndex));
+		const clampedItem = visibleItems[clamped];
+		if (clampedItem && !this.#isItemDisabled(clampedItem)) {
+			return clamped;
+		}
+		for (let i = clamped + 1; i <= maxIndex; i++) {
+			const item = visibleItems[i];
+			if (item && !this.#isItemDisabled(item)) {
+				return i;
+			}
+		}
+		for (let i = clamped - 1; i >= 0; i--) {
+			const item = visibleItems[i];
+			if (item && !this.#isItemDisabled(item)) {
+				return i;
+			}
+		}
+		return clamped;
+	}
+
+	#moveSelection(delta: number): void {
+		const visibleItems = this.#getVisibleItems();
+		const count = visibleItems.length;
+		if (count === 0) {
+			return;
+		}
+		let index = this.#selectedIndex;
+		for (let step = 0; step < count; step++) {
+			index = (index + delta + count) % count;
+			const item = visibleItems[index];
+			if (item && !this.#isItemDisabled(item)) {
+				this.#selectedIndex = index;
+				this.#updateList();
+				return;
+			}
+		}
+		this.#selectedIndex = this.#coerceSelectedIndex(this.#selectedIndex, visibleItems);
+		this.#updateList();
+	}
+
 	#filterModels(query: string): void {
 		const activeTabId = this.#getActiveTabId();
 		const activeProviderId = this.#getActiveProviderId();
@@ -696,8 +774,11 @@ export class ModelSelectorComponent extends Container {
 			this.#filteredCanonicalModels = baseCanonicalModels;
 		}
 
-		const visibleCount = isCanonicalTab ? this.#filteredCanonicalModels.length : this.#filteredModels.length;
-		this.#selectedIndex = Math.min(this.#selectedIndex, Math.max(0, visibleCount - 1));
+		const visibleItems = isCanonicalTab ? this.#filteredCanonicalModels : this.#filteredModels;
+		this.#selectedIndex = this.#coerceSelectedIndex(
+			Math.min(this.#selectedIndex, Math.max(0, visibleItems.length - 1)),
+			visibleItems,
+		);
 		this.#updateList();
 	}
 
@@ -778,6 +859,7 @@ export class ModelSelectorComponent extends Container {
 
 		const showProvider = this.#getActiveTabId() === ALL_TAB;
 
+		const rows: string[] = [];
 		// Show visible slice of filtered models
 		for (let i = startIndex; i < endIndex; i++) {
 			const item = visibleItems[i];
@@ -786,6 +868,8 @@ export class ModelSelectorComponent extends Container {
 			const providerItem = isCanonicalTab ? undefined : (item as ModelItem);
 
 			const isSelected = i === this.#selectedIndex;
+			const isDisabled = this.#isItemDisabled(item);
+			const disabledSuffix = this.#formatContextLimitSuffix(item.model);
 
 			// Build role badges (inverted: color as background, black text)
 			const roleBadgeTokens: string[] = [];
@@ -815,34 +899,42 @@ export class ModelSelectorComponent extends Container {
 				if (isCanonicalTab) {
 					const variants = theme.fg("dim", ` [${canonicalItem?.variantCount ?? 0}]`);
 					const backing = theme.fg("dim", ` -> ${item.model.provider}/${item.model.id}`);
-					line = `${prefix}${theme.fg("accent", item.id)}${variants}${backing}${badgeText}`;
+					line = `${prefix}${theme.fg("accent", item.id)}${variants}${backing}${badgeText}${disabledSuffix}`;
 				} else if (showProvider) {
 					const providerPrefix = theme.fg("dim", `${providerItem?.provider ?? ""}/`);
-					line = `${prefix}${providerPrefix}${theme.fg("accent", providerItem?.id ?? item.id)}${badgeText}`;
+					line = `${prefix}${providerPrefix}${theme.fg("accent", providerItem?.id ?? item.id)}${badgeText}${disabledSuffix}`;
 				} else {
-					line = `${prefix}${theme.fg("accent", item.id)}${badgeText}`;
+					line = `${prefix}${theme.fg("accent", item.id)}${badgeText}${disabledSuffix}`;
 				}
 			} else {
 				const prefix = "  ";
 				if (isCanonicalTab) {
 					const variants = theme.fg("dim", ` [${canonicalItem?.variantCount ?? 0}]`);
 					const backing = theme.fg("dim", ` -> ${item.model.provider}/${item.model.id}`);
-					line = `${prefix}${item.id}${variants}${backing}${badgeText}`;
+					line = `${prefix}${item.id}${variants}${backing}${badgeText}${disabledSuffix}`;
 				} else if (showProvider) {
 					const providerPrefix = theme.fg("dim", `${providerItem?.provider ?? ""}/`);
-					line = `${prefix}${providerPrefix}${providerItem?.id ?? item.id}${badgeText}`;
+					line = `${prefix}${providerPrefix}${providerItem?.id ?? item.id}${badgeText}${disabledSuffix}`;
 				} else {
-					line = `${prefix}${item.id}${badgeText}`;
+					line = `${prefix}${item.id}${badgeText}${disabledSuffix}`;
 				}
 			}
 
-			this.#listContainer.addChild(new Text(line, 0, 0));
+			if (isDisabled) {
+				line = theme.fg("dim", Bun.stripANSI(line));
+			}
+			rows.push(line);
 		}
 
-		// Add scroll indicator if needed
-		if (startIndex > 0 || endIndex < visibleItems.length) {
-			const scrollInfo = theme.fg("muted", `  (${this.#selectedIndex + 1}/${visibleItems.length})`);
-			this.#listContainer.addChild(new Text(scrollInfo, 0, 0));
+		if (rows.length > 0) {
+			const sv = new ScrollView(rows, {
+				height: rows.length,
+				scrollbar: "auto",
+				totalRows: visibleItems.length,
+				theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+			});
+			sv.setScrollOffset(startIndex);
+			this.#listContainer.addChild(sv);
 		}
 
 		// Show error message or "no results" if empty
@@ -863,8 +955,14 @@ export class ModelSelectorComponent extends Container {
 			const suffix = isCanonicalTab
 				? ` (${selected.model.provider}/${selected.model.id}, ${(selected as CanonicalModelItem).variantCount} variants)`
 				: "";
+			const limitWarning = this.#isItemDisabled(selected)
+				? theme.fg(
+						"dim",
+						` — current context ${formatNumber(this.#currentContextTokens).toLowerCase()} > ${formatNumber(selected.model.contextWindow).toLowerCase()} limit`,
+					)
+				: "";
 			this.#listContainer.addChild(
-				new Text(theme.fg("muted", `  Model Name: ${selected.model.name}${suffix}`), 0, 0),
+				new Text(theme.fg("muted", `  Model Name: ${selected.model.name}${suffix}`) + limitWarning, 0, 0),
 			);
 		}
 	}
@@ -890,7 +988,8 @@ export class ModelSelectorComponent extends Container {
 	}
 
 	#openMenu(): void {
-		if (!this.#getSelectedItem()) return;
+		const selectedItem = this.#getSelectedItem();
+		if (!selectedItem || this.#isItemDisabled(selectedItem)) return;
 
 		this.#isMenuOpen = true;
 		this.#menuStep = "role";
@@ -978,26 +1077,20 @@ export class ModelSelectorComponent extends Container {
 
 		// Up arrow - navigate list (wrap to bottom when at top)
 		if (matchesSelectUp(keyData)) {
-			const itemCount = this.#isCanonicalTab() ? this.#filteredCanonicalModels.length : this.#filteredModels.length;
-			if (itemCount === 0) return;
-			this.#selectedIndex = this.#selectedIndex === 0 ? itemCount - 1 : this.#selectedIndex - 1;
-			this.#updateList();
+			this.#moveSelection(-1);
 			return;
 		}
 
 		// Down arrow - navigate list (wrap to top when at bottom)
 		if (matchesSelectDown(keyData)) {
-			const itemCount = this.#isCanonicalTab() ? this.#filteredCanonicalModels.length : this.#filteredModels.length;
-			if (itemCount === 0) return;
-			this.#selectedIndex = this.#selectedIndex === itemCount - 1 ? 0 : this.#selectedIndex + 1;
-			this.#updateList();
+			this.#moveSelection(1);
 			return;
 		}
 
 		// Enter - open context menu or select directly in temporary mode
 		if (matchesKey(keyData, "enter") || matchesKey(keyData, "return") || keyData === "\n") {
 			const selectedItem = this.#getSelectedItem();
-			if (selectedItem) {
+			if (selectedItem && !this.#isItemDisabled(selectedItem)) {
 				if (this.#temporaryOnly) {
 					// In temporary mode, skip menu and select directly
 					this.#handleSelect(selectedItem, null);
@@ -1020,7 +1113,7 @@ export class ModelSelectorComponent extends Container {
 	}
 	#handleMenuInput(keyData: string): void {
 		const selectedItem = this.#getSelectedItem();
-		if (!selectedItem) return;
+		if (!selectedItem || this.#isItemDisabled(selectedItem)) return;
 
 		const optionCount =
 			this.#menuStep === "thinking" && this.#menuSelectedRole !== null
@@ -1079,6 +1172,9 @@ export class ModelSelectorComponent extends Container {
 		role: string | null,
 		thinkingLevel?: ConfiguredThinkingLevel,
 	): void {
+		if (this.#isItemDisabled(item)) {
+			return;
+		}
 		// For temporary role, don't save to settings - just notify caller
 		if (role === null) {
 			this.#onSelectCallback(item.model, null, undefined, item.selector);
diff --git a/packages/coding-agent/src/modes/components/oauth-selector.ts b/packages/coding-agent/src/modes/components/oauth-selector.ts
index 3e302be10..e73e3f86e 100644
--- a/packages/coding-agent/src/modes/components/oauth-selector.ts
+++ b/packages/coding-agent/src/modes/components/oauth-selector.ts
@@ -1,6 +1,14 @@
 import { getOAuthProviders } from "@oh-my-pi/pi-ai/utils/oauth";
 import type { OAuthProviderInfo } from "@oh-my-pi/pi-ai/utils/oauth/types";
-import { Container, extractPrintableText, fuzzyFilter, matchesKey, Spacer, TruncatedText } from "@oh-my-pi/pi-tui";
+import {
+	Container,
+	extractPrintableText,
+	fuzzyFilter,
+	matchesKey,
+	ScrollView,
+	Spacer,
+	TruncatedText,
+} from "@oh-my-pi/pi-tui";
 import { theme } from "../../modes/theme/theme";
 import { matchesSelectCancel, matchesSelectDown, matchesSelectUp } from "../../modes/utils/keybinding-matchers";
 import type { AuthStorage } from "../../session/auth-storage";
@@ -162,14 +170,10 @@ export class OAuthSelectorComponent extends Container {
 		return this.#isSearchEnabled() || this.#searchQuery.length > 0;
 	}
 
-	#renderStatusLine(total: number): string {
-		const selectedCount = total === 0 ? 0 : this.#selectedIndex + 1;
-		const count =
-			this.#searchQuery.trim() && total !== this.#allProviders.length
-				? `${selectedCount}/${total} of ${this.#allProviders.length}`
-				: `${selectedCount}/${total}`;
-		const suffix = this.#searchQuery.trim() ? `  Search: ${this.#searchQuery}` : "  Type to search";
-		return theme.fg("muted", `  (${count})${suffix}`);
+	#renderStatusLine(_total: number): string {
+		const query = this.#searchQuery.trim();
+		const suffix = query ? `Search: ${this.#searchQuery}` : "Type to search";
+		return theme.fg("muted", `  ${suffix}`);
 	}
 
 	#getProviderSearchText(provider: OAuthProviderInfo): string {
@@ -223,6 +227,7 @@ export class OAuthSelectorComponent extends Container {
 				: Math.max(0, Math.min(this.#selectedIndex - Math.floor(maxVisible / 2), total - maxVisible));
 		const endIndex = Math.min(startIndex + maxVisible, total);
 
+		const rows: string[] = [];
 		for (let i = startIndex; i < endIndex; i++) {
 			const provider = this.#filteredProviders[i];
 			if (!provider) continue;
@@ -239,11 +244,22 @@ export class OAuthSelectorComponent extends Container {
 				const text = isAvailable ? `  ${provider.name}` : theme.fg("dim", `  ${provider.name}`);
 				line = text + statusIndicator;
 			}
-			this.#listContainer.addChild(new TruncatedText(line, 0, 0));
+			rows.push(line);
 		}
 
-		// Scroll/search indicator when list is windowed or searchable
-		if (startIndex > 0 || endIndex < total || this.#shouldRenderSearchStatus()) {
+		if (rows.length > 0) {
+			const sv = new ScrollView(rows, {
+				height: rows.length,
+				scrollbar: "auto",
+				totalRows: total,
+				theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+			});
+			sv.setScrollOffset(startIndex);
+			this.#listContainer.addChild(sv);
+		}
+
+		// Search status line (scrollbar covers overflow indication)
+		if (this.#shouldRenderSearchStatus()) {
 			this.#listContainer.addChild(new TruncatedText(this.#renderStatusLine(total), 0, 0));
 		}
 
diff --git a/packages/coding-agent/src/modes/components/overlay-box.ts b/packages/coding-agent/src/modes/components/overlay-box.ts
new file mode 100644
index 000000000..e427e6a25
--- /dev/null
+++ b/packages/coding-agent/src/modes/components/overlay-box.ts
@@ -0,0 +1,108 @@
+/**
+ * Shared box-drawing chrome for fullscreen overlays (the `/copy` picker, the
+ * plan-review overlay, …). Every helper paints with `theme.boxSharp` glyphs and
+ * the `border`/`accent` theme colors so all outlined overlays read identically.
+ */
+import { padding, truncateToWidth, visibleWidth } from "@oh-my-pi/pi-tui";
+import { theme } from "../theme/theme";
+
+/** Pad or truncate a (possibly ANSI-styled) string to exactly `width` columns. */
+export function fit(text: string, width: number): string {
+	if (width <= 0) return "";
+	const w = visibleWidth(text);
+	if (w === width) return text;
+	if (w < width) return text + padding(width - w);
+	const cut = truncateToWidth(text, width);
+	const cw = visibleWidth(cut);
+	return cw < width ? cut + padding(width - cw) : cut;
+}
+
+function paint(s: string): string {
+	return theme.fg("border", s);
+}
+
+/** Top border with an optional accent-colored title inset into the rule. */
+export function topBorder(width: number, title: string): string {
+	const box = theme.boxSharp;
+	const inner = Math.max(0, width - 2);
+	if (!title) return paint(box.topLeft + box.horizontal.repeat(inner) + box.topRight);
+	const shown = truncateToWidth(` ${title} `, Math.max(0, inner - 2));
+	const fillWidth = Math.max(0, inner - 1 - visibleWidth(shown));
+	return (
+		paint(box.topLeft + box.horizontal) +
+		theme.bold(theme.fg("accent", shown)) +
+		paint(box.horizontal.repeat(fillWidth) + box.topRight)
+	);
+}
+
+/** A horizontal rule with left/right tees, splitting overlay sections. */
+export function divider(width: number): string {
+	const box = theme.boxSharp;
+	return paint(box.teeRight + box.horizontal.repeat(Math.max(0, width - 2)) + box.teeLeft);
+}
+
+export function bottomBorder(width: number): string {
+	const box = theme.boxSharp;
+	return paint(box.bottomLeft + box.horizontal.repeat(Math.max(0, width - 2)) + box.bottomRight);
+}
+
+/** Wrap pre-styled content in vertical borders with single-column insets. */
+export function row(content: string, width: number): string {
+	const box = theme.boxSharp;
+	return `${paint(box.vertical)} ${fit(content, Math.max(0, width - 4))} ${paint(box.vertical)}`;
+}
+
+/**
+ * Column index (0-based) of the inner divider for a two-column layout whose
+ * sidebar content area is `sidebarWidth` columns wide. The layout is
+ * `│ sidebar │ body │` with a single-column inset on every side, so the divider
+ * vertical sits at `sidebarWidth + 3` and the body content area is
+ * {@link splitBodyWidth} columns.
+ */
+function splitDividerCol(sidebarWidth: number): number {
+	return sidebarWidth + 3;
+}
+
+/** Body content width for a two-column overlay of total `width`. */
+export function splitBodyWidth(width: number, sidebarWidth: number): number {
+	return Math.max(0, width - sidebarWidth - 7);
+}
+
+/** Top border carrying the title, split by a `┬` over the column divider. */
+export function topBorderSplit(width: number, title: string, sidebarWidth: number): string {
+	const box = theme.boxSharp;
+	const dividerCol = splitDividerCol(sidebarWidth);
+	const leftLen = Math.max(0, dividerCol - 1);
+	const rightLen = Math.max(0, width - 2 - dividerCol);
+	let left: string;
+	if (!title) {
+		left = paint(box.topLeft + box.horizontal.repeat(leftLen));
+	} else {
+		const shown = truncateToWidth(` ${title} `, Math.max(0, leftLen - 1));
+		const fillWidth = Math.max(0, leftLen - 1 - visibleWidth(shown));
+		left =
+			paint(box.topLeft + box.horizontal) +
+			theme.bold(theme.fg("accent", shown)) +
+			paint(box.horizontal.repeat(fillWidth));
+	}
+	return left + paint(box.teeDown + box.horizontal.repeat(rightLen) + box.topRight);
+}
+
+/** Section rule that closes the sidebar column with a `┴` over the divider. */
+export function dividerSplit(width: number, sidebarWidth: number): string {
+	const box = theme.boxSharp;
+	const dividerCol = splitDividerCol(sidebarWidth);
+	const leftLen = Math.max(0, dividerCol - 1);
+	const rightLen = Math.max(0, width - 2 - dividerCol);
+	return paint(
+		box.teeRight + box.horizontal.repeat(leftLen) + box.teeUp + box.horizontal.repeat(rightLen) + box.teeLeft,
+	);
+}
+
+/** A two-column content row: `│ sidebar │ body │`, each inset by one column. */
+export function splitRow(sidebar: string, body: string, width: number, sidebarWidth: number): string {
+	const box = theme.boxSharp;
+	const bodyWidth = splitBodyWidth(width, sidebarWidth);
+	const bar = paint(box.vertical);
+	return `${bar} ${fit(sidebar, sidebarWidth)} ${bar} ${fit(body, bodyWidth)} ${bar}`;
+}
diff --git a/packages/coding-agent/src/modes/components/plan-review-overlay.ts b/packages/coding-agent/src/modes/components/plan-review-overlay.ts
new file mode 100644
index 000000000..ef74ae7ea
--- /dev/null
+++ b/packages/coding-agent/src/modes/components/plan-review-overlay.ts
@@ -0,0 +1,799 @@
+/**
+ * Fullscreen plan-review overlay. The overlay owns its entire content: the plan
+ * is split into sections (preamble + one per heading), each rendered through its
+ * own {@link Markdown} and windowed by a {@link ScrollView}, while the approval
+ * options (plus the optional model-tier slider) sit beneath inside the same
+ * outlined box — one self-contained surface in the spirit of the `/copy` picker.
+ *
+ * When the terminal is wide enough and the plan has ≥2 headings, a Contents
+ * sidebar appears: it tracks the scrolled section with an accent "glow", and —
+ * when focused — lets the operator jump between sections, delete a section
+ * (with undo), and annotate sections with feedback that feeds the Refine loop.
+ *
+ * Focus regions (`toc`/`body`/`actions`) cycle with Tab/Shift+Tab; arrows move
+ * within the focused region and step left into the sidebar. The default focus is
+ * `actions`, so the muscle memory of the old single-target overlay carries over:
+ * ↑/↓ select options, Enter confirms, ←/→ drives the slider when there is no
+ * sidebar, g/G + PgUp/PgDn scroll, and the external-editor key opens the plan.
+ */
+import {
+	type Component,
+	Ellipsis,
+	Input,
+	Markdown,
+	type MarkdownTheme,
+	matchesKey,
+	ScrollView,
+	truncateToWidth,
+	visibleWidth,
+} from "@oh-my-pi/pi-tui";
+import { getMarkdownTheme, theme } from "../theme/theme";
+import {
+	matchesAppExternalEditor,
+	matchesSelectCancel,
+	matchesSelectDown,
+	matchesSelectUp,
+} from "../utils/keybinding-matchers";
+import type { HookSelectorSlider } from "./hook-selector";
+import {
+	bottomBorder,
+	divider,
+	dividerSplit,
+	fit,
+	row,
+	splitBodyWidth,
+	splitRow,
+	topBorder,
+	topBorderSplit,
+} from "./overlay-box";
+import { joinPlanSections, parsePlanSections, sectionDeletionSpan } from "./plan-toc";
+import { renderSegmentTrack } from "./segment-track";
+
+/** Title shown in the overlay's top border. */
+const OVERLAY_TITLE = "Plan Review";
+/** Minimum plan-body rows kept visible even on short terminals. */
+const MIN_BODY_ROWS = 3;
+/** Sidebar gates: enough headings, a wide terminal, and a usable body column. */
+const SIDEBAR_MIN_HEADINGS = 2;
+const SIDEBAR_MIN_TOTAL_WIDTH = 64;
+const SIDEBAR_MIN_BODY_WIDTH = 40;
+
+type Focus = "toc" | "body" | "actions";
+
+interface OverlaySection {
+	level: number;
+	title: string;
+	raw: string;
+	md: Markdown;
+	annotations: string[];
+}
+
+/** Undo snapshot: joined plan text, annotations aligned by section, and the
+ *  accumulated deleted-section feedback at the time of the snapshot. */
+interface UndoEntry {
+	text: string;
+	annotations: string[][];
+	deleted: string[];
+}
+
+export interface PlanReviewOverlayCallbacks {
+	/** Invoked with the chosen option label (never a disabled one). */
+	onPick: (label: string) => void;
+	/** Invoked on Esc / cancel. */
+	onCancel: () => void;
+	/** Invoked when the external-editor key is pressed (overlay stays open). */
+	onExternalEditor?: () => void;
+	/** Invoked with the new full plan text after an in-overlay delete/undo. */
+	onPlanEdited?: (content: string) => void;
+	/** Invoked with the Refine feedback markdown whenever annotations change. */
+	onFeedbackChange?: (feedback: string) => void;
+}
+
+export interface PlanReviewOverlayOptions {
+	/** Prompt rendered above the options (e.g. "Plan mode - next step"). */
+	promptTitle?: string;
+	options: string[];
+	/** Indices into `options` that render dimmed and cannot be selected. */
+	disabledIndices?: number[];
+	/** Trailing footer hint (cancel hint); the overlay prepends dynamic help. */
+	helpText?: string;
+	/** Initially highlighted option index. */
+	initialIndex?: number;
+	/** Optional model-tier slider rendered between the plan body and options. */
+	slider?: HookSelectorSlider;
+	/** Display label for the external-editor key, surfaced in the footer help. */
+	externalEditorLabel?: string;
+}
+
+/** Default trailing footer hint when the caller supplies none. */
+const DEFAULT_HELP_SUFFIX = "esc cancel";
+
+export class PlanReviewOverlay implements Component {
+	#mdTheme: MarkdownTheme;
+	#scrollView: ScrollView;
+	#sections: OverlaySection[] = [];
+	#toc: number[] = [];
+	/** Shallowest level among ToC entries, used to flatten indentation. */
+	#tocBaseLevel = 1;
+	#sectionOffsets: number[] = [];
+	#undo: UndoEntry[] = [];
+	/** Titles of sections deleted in the overlay, surfaced as Refine feedback. */
+	#deleted: string[] = [];
+
+	#options: string[];
+	#disabled: Set<number>;
+	#helpSuffix: string;
+	#externalEditorLabel: string | undefined;
+	#promptTitle: string | undefined;
+	#selectedIndex: number;
+	#slider: HookSelectorSlider | undefined;
+	#sliderIndex: number;
+
+	#focus: Focus = "actions";
+	#tocCursor = 0;
+	#sidebarShown = false;
+	#pendingScrollToToc = false;
+
+	// Click hit-testing, rebuilt every render. Keys are 0-based rendered-line
+	// indices (== screen rows, since the fullscreen overlay paints from row 0).
+	#optionClickRows = new Map<number, number>();
+	#tocClickRows = new Map<number, number>();
+	#bodyClickRows = new Set<number>();
+	/** 1-based column at/under which a region-row click targets the sidebar. */
+	#sidebarClickMaxCol = 0;
+
+	#annotating = false;
+	#input: Input;
+
+	constructor(
+		planContent: string,
+		options: PlanReviewOverlayOptions,
+		private readonly callbacks: PlanReviewOverlayCallbacks,
+	) {
+		this.#mdTheme = getMarkdownTheme();
+		this.#scrollView = new ScrollView([], {
+			height: MIN_BODY_ROWS,
+			scrollbar: "auto",
+			ellipsis: Ellipsis.Omit,
+			theme: { track: t => theme.fg("dim", t), thumb: t => theme.fg("accent", t) },
+		});
+		this.#options = options.options;
+		this.#disabled = new Set(
+			(options.disabledIndices ?? []).filter(i => Number.isInteger(i) && i >= 0 && i < this.#options.length),
+		);
+		this.#helpSuffix = options.helpText ?? DEFAULT_HELP_SUFFIX;
+		this.#externalEditorLabel = options.externalEditorLabel;
+		this.#promptTitle = options.promptTitle;
+		this.#selectedIndex = this.#coerceIndex(options.initialIndex ?? 0);
+		if (options.slider && options.slider.segments.length > 0) {
+			this.#slider = options.slider;
+			this.#sliderIndex = Math.max(0, Math.min(options.slider.index, options.slider.segments.length - 1));
+		} else {
+			this.#sliderIndex = 0;
+		}
+		this.#input = new Input();
+		this.#input.setUseTerminalCursor(false);
+		this.#input.onSubmit = value => this.#submitAnnotation(value);
+		this.#input.onEscape = () => this.#exitAnnotate();
+		this.#setSections(planContent);
+	}
+
+	invalidate(): void {
+		for (const section of this.#sections) section.md.invalidate();
+	}
+
+	/** Swap the displayed plan (e.g. after an external-editor round-trip) and
+	 *  reset scroll/focus so the operator starts at the top. Does not emit
+	 *  `onPlanEdited` (the editor round-trip already persisted the file). */
+	setPlanContent(planContent: string): void {
+		this.#setSections(planContent);
+		this.#scrollView.scrollToTop();
+		this.#tocCursor = 0;
+		// A wholesale external-editor swap supersedes prior in-overlay deletions.
+		this.#deleted = [];
+		this.#undo = [];
+		this.#recomputeFeedback();
+	}
+
+	#setSections(planContent: string): void {
+		this.#sections = parsePlanSections(planContent).map(section => ({
+			level: section.level,
+			title: section.title,
+			raw: section.raw,
+			md: new Markdown(section.raw, 1, 0, this.#mdTheme),
+			annotations: [] as string[],
+		}));
+		this.#rebuildToc();
+		this.#tocCursor = Math.min(this.#tocCursor, Math.max(0, this.#toc.length - 1));
+	}
+
+	#rebuildToc(): void {
+		const headings: number[] = [];
+		for (let i = 0; i < this.#sections.length; i++) {
+			if (this.#sections[i]!.level >= 1) headings.push(i);
+		}
+		// Drop the plan's title from the ToC: a single shallowest heading at the
+		// top of the document is the plan name itself ("we know it's the plan"),
+		// so listing it adds noise. Plans with several top-level sections keep
+		// them all.
+		let minLevel = Number.POSITIVE_INFINITY;
+		for (const i of headings) minLevel = Math.min(minLevel, this.#sections[i]!.level);
+		const topLevel = headings.filter(i => this.#sections[i]!.level === minLevel);
+		const titleIndex = topLevel.length === 1 && headings[0] === topLevel[0] ? topLevel[0] : -1;
+		this.#toc = headings.filter(i => i !== titleIndex);
+		this.#tocBaseLevel = this.#toc.length > 0 ? Math.min(...this.#toc.map(i => this.#sections[i]!.level)) : 1;
+	}
+
+	/** Clamp `index` to range, then walk to the nearest enabled option so the
+	 *  cursor never rests on a disabled row. */
+	#coerceIndex(index: number): number {
+		const max = this.#options.length - 1;
+		if (max < 0) return -1;
+		const clamped = Math.max(0, Math.min(index, max));
+		if (!this.#disabled.has(clamped)) return clamped;
+		for (let i = clamped + 1; i <= max; i++) if (!this.#disabled.has(i)) return i;
+		for (let i = clamped - 1; i >= 0; i--) if (!this.#disabled.has(i)) return i;
+		return clamped;
+	}
+
+	/** First enabled option index (or -1 when none), used to detect the "top". */
+	#firstEnabledIndex(): number {
+		for (let i = 0; i < this.#options.length; i++) if (!this.#disabled.has(i)) return i;
+		return -1;
+	}
+
+	/** Move the option cursor by `delta`, skipping disabled rows, stopping at the
+	 *  list edge. */
+	#moveSelection(delta: number): void {
+		const max = this.#options.length - 1;
+		if (max < 0) return;
+		let index = this.#selectedIndex;
+		while (true) {
+			const next = Math.max(0, Math.min(index + delta, max));
+			if (next === index) return;
+			index = next;
+			if (!this.#disabled.has(index)) {
+				this.#selectedIndex = index;
+				return;
+			}
+		}
+	}
+
+	/** Step the slider by `delta`, clamped to its edges (narrow-terminal mode). */
+	#moveSlider(delta: number): void {
+		const slider = this.#slider;
+		if (!slider) return;
+		const next = Math.max(0, Math.min(slider.segments.length - 1, this.#sliderIndex + delta));
+		if (next === this.#sliderIndex) return;
+		this.#sliderIndex = next;
+		slider.onChange?.(next);
+	}
+
+	#confirmSelection(): void {
+		const index = this.#selectedIndex;
+		if (index >= 0 && index < this.#options.length && !this.#disabled.has(index)) {
+			this.callbacks.onPick(this.#options[index]!);
+		}
+	}
+
+	handleInput(keyData: string): void {
+		if (keyData.startsWith("\x1b[<") && this.#handleMouse(keyData)) return;
+		if (this.#annotating) {
+			this.#input.handleInput(keyData);
+			return;
+		}
+		if (matchesSelectCancel(keyData)) {
+			this.callbacks.onCancel();
+			return;
+		}
+		if (this.callbacks.onExternalEditor && matchesAppExternalEditor(keyData)) {
+			this.callbacks.onExternalEditor();
+			return;
+		}
+		if (matchesKey(keyData, "tab") || keyData === "\t") {
+			this.#cycleRegion(1);
+			return;
+		}
+		if (matchesKey(keyData, "shift+tab") || keyData === "\x1b[Z") {
+			this.#cycleRegion(-1);
+			return;
+		}
+		switch (this.#focus) {
+			case "actions":
+				this.#handleActions(keyData);
+				return;
+			case "body":
+				this.#handleBody(keyData);
+				return;
+			case "toc":
+				this.#handleToc(keyData);
+				return;
+		}
+	}
+
+	/**
+	 * Hit-test an SGR mouse report (`\x1b[<b;x;yM/m`) against the click maps the
+	 * last render recorded. Returns true when consumed. The fullscreen overlay
+	 * paints from screen row 0, so a 1-based mouse row maps directly to the
+	 * rendered-line index. Wheel scrolls the body; a left click on an option
+	 * activates it (select + confirm), on a ToC row jumps to that section, and on
+	 * the body column focuses the body.
+	 */
+	#handleMouse(data: string): boolean {
+		const match = /^\x1b\[<(\d+);(\d+);(\d+)([Mm])$/.exec(data);
+		if (!match) return false;
+		const button = Number(match[1]);
+		const x = Number(match[2]);
+		const row = Number(match[3]) - 1;
+		if (button & 64) {
+			// Scroll wheel: low bit selects direction (64 up, 65 down).
+			this.#scrollView.scroll(button & 1 ? 3 : -3);
+			return true;
+		}
+		if (match[4] !== "M") return true; // release
+		if (button & 32) return true; // motion/drag
+		if ((button & 3) !== 0) return true; // not the left button
+		const optionIndex = this.#optionClickRows.get(row);
+		if (optionIndex !== undefined) {
+			if (!this.#disabled.has(optionIndex)) {
+				this.#focus = "actions";
+				this.#selectedIndex = optionIndex;
+				this.#confirmSelection();
+			}
+			return true;
+		}
+		const tocPos = this.#tocClickRows.get(row);
+		if (tocPos !== undefined && x <= this.#sidebarClickMaxCol) {
+			this.#focus = "toc";
+			this.#tocCursor = tocPos;
+			this.#scrubBodyToToc();
+			return true;
+		}
+		if (this.#bodyClickRows.has(row)) {
+			this.#setFocus("body");
+		}
+		return true;
+	}
+
+	#cycleRegion(direction: number): void {
+		// Sidebar is skipped from the cycle when it is not shown.
+		const regions: Focus[] = this.#sidebarShown ? ["toc", "body", "actions"] : ["body", "actions"];
+		const current = regions.indexOf(this.#focus);
+		const base = current < 0 ? regions.length - 1 : current;
+		this.#setFocus(regions[(base + direction + regions.length) % regions.length]!);
+	}
+
+	#setFocus(focus: Focus): void {
+		this.#focus = focus;
+		if (focus === "toc") this.#tocCursor = this.#deriveTocCursorFromScroll();
+	}
+
+	#handleActions(data: string): void {
+		// Left/right always drive the slider. The sidebar sits beside the body
+		// (above this row), not the slider, so stealing left for it would strand
+		// the operator unable to step the model tier back — reach the ToC via Tab.
+		const isLeft = matchesKey(data, "left") || (this.#slider !== undefined && data === "h");
+		const isRight = matchesKey(data, "right") || (this.#slider !== undefined && data === "l");
+		if (isLeft) {
+			this.#moveSlider(-1);
+			return;
+		}
+		if (isRight) {
+			this.#moveSlider(1);
+			return;
+		}
+		if (matchesSelectUp(data) || data === "k") {
+			if (this.#selectedIndex === this.#firstEnabledIndex()) this.#setFocus("body");
+			else this.#moveSelection(-1);
+			return;
+		}
+		if (matchesSelectDown(data) || data === "j") {
+			this.#moveSelection(1);
+			return;
+		}
+		if (matchesKey(data, "enter") || matchesKey(data, "return") || data === "\n") {
+			this.#confirmSelection();
+			return;
+		}
+		this.#handleBodyScroll(data);
+	}
+
+	#handleBody(data: string): void {
+		if (matchesKey(data, "left") || data === "h") {
+			if (this.#sidebarShown) this.#setFocus("toc");
+			return;
+		}
+		if (
+			matchesKey(data, "right") ||
+			data === "l" ||
+			matchesKey(data, "enter") ||
+			matchesKey(data, "return") ||
+			data === "\n"
+		) {
+			this.#setFocus("actions");
+			return;
+		}
+		// Vertical nav flows between regions at the edges: scrolling off the bottom
+		// drops into the actions ("next step"); scrolling off the top steps back up
+		// to the ToC.
+		if (matchesSelectUp(data) || data === "k") {
+			if (this.#scrollView.getScrollOffset() <= 0 && this.#sidebarShown) this.#setFocus("toc");
+			else this.#scrollView.scroll(-1);
+			return;
+		}
+		if (matchesSelectDown(data) || data === "j") {
+			if (this.#scrollView.getScrollOffset() >= this.#scrollView.getMaxScrollOffset()) this.#setFocus("actions");
+			else this.#scrollView.scroll(1);
+			return;
+		}
+		this.#handleBodyScroll(data);
+	}
+
+	/**
+	 * Shared scroll dispatch for body + actions focus. Delegates standard keys
+	 * (Arrows, Shift+Arrow fast-scroll, PgUp/PgDn, Home/End) to the ScrollView,
+	 * then adds the vim g/G jumps. Plain Arrow/k/j are consumed by the callers
+	 * before this runs, so here it only ever sees the paging/fast keys.
+	 */
+	#handleBodyScroll(data: string): void {
+		if (this.#scrollView.handleScrollKey(data)) return;
+		if (data === "g") this.#scrollView.scrollToTop();
+		else if (data === "G") this.#scrollView.scrollToBottom();
+	}
+
+	#handleToc(data: string): void {
+		if (matchesSelectUp(data) || data === "k") {
+			this.#moveTocCursor(-1);
+			return;
+		}
+		if (matchesSelectDown(data) || data === "j") {
+			// Past the last section, fall through to the actions ("next step").
+			if (this.#tocCursor >= this.#toc.length - 1) this.#setFocus("actions");
+			else this.#moveTocCursor(1);
+			return;
+		}
+		if (
+			matchesKey(data, "right") ||
+			data === "l" ||
+			matchesKey(data, "enter") ||
+			matchesKey(data, "return") ||
+			data === "\n"
+		) {
+			this.#setFocus("body");
+			return;
+		}
+		if (data === "d" || matchesKey(data, "delete")) {
+			this.#deleteSelectedSection();
+			return;
+		}
+		if (data === "a") {
+			this.#startAnnotate();
+			return;
+		}
+		if (data === "u") {
+			this.#undoLast();
+			return;
+		}
+	}
+
+	#moveTocCursor(delta: number): void {
+		if (this.#toc.length === 0) return;
+		const next = Math.max(0, Math.min(this.#toc.length - 1, this.#tocCursor + delta));
+		if (next === this.#tocCursor) return;
+		this.#tocCursor = next;
+		this.#scrubBodyToToc();
+	}
+
+	/** Scroll the body so the selected ToC section's heading sits at the top. */
+	#scrubBodyToToc(): void {
+		const sectionIndex = this.#toc[this.#tocCursor];
+		if (sectionIndex === undefined) return;
+		const offset = this.#sectionOffsets[sectionIndex];
+		if (offset !== undefined) this.#scrollView.setScrollOffset(offset);
+	}
+
+	/** Greatest ToC position whose section starts at or above the scroll offset. */
+	#deriveTocCursorFromScroll(): number {
+		if (this.#toc.length === 0) return 0;
+		const scrollOffset = this.#scrollView.getScrollOffset();
+		let current = 0;
+		for (let i = 0; i < this.#sections.length; i++) {
+			if ((this.#sectionOffsets[i] ?? 0) <= scrollOffset) current = i;
+			else break;
+		}
+		let pos = 0;
+		for (let p = 0; p < this.#toc.length; p++) {
+			if (this.#toc[p]! <= current) pos = p;
+			else break;
+		}
+		return pos;
+	}
+
+	#pushUndo(): void {
+		this.#undo.push({
+			text: joinPlanSections(this.#sections),
+			annotations: this.#sections.map(section => [...section.annotations]),
+			deleted: [...this.#deleted],
+		});
+	}
+
+	#deleteSelectedSection(): void {
+		const sectionIndex = this.#toc[this.#tocCursor];
+		if (sectionIndex === undefined) return;
+		const span = sectionDeletionSpan(this.#sections, sectionIndex);
+		if (span.length === 0) return;
+		this.#pushUndo();
+		// Record the removed headings so the Refine feedback can ask the model to
+		// drop them, then splice from the bottom up so earlier indices stay valid.
+		for (const i of span) {
+			const section = this.#sections[i]!;
+			if (section.level >= 1 && section.title) this.#deleted.push(section.title);
+		}
+		for (let i = span.length - 1; i >= 0; i--) this.#sections.splice(span[i]!, 1);
+		this.#rebuildToc();
+		this.#tocCursor = Math.min(this.#tocCursor, Math.max(0, this.#toc.length - 1));
+		this.#pendingScrollToToc = true;
+		this.callbacks.onPlanEdited?.(joinPlanSections(this.#sections));
+		this.#recomputeFeedback();
+	}
+
+	#undoLast(): void {
+		const entry = this.#undo.pop();
+		if (!entry) return;
+		this.#setSections(entry.text);
+		for (let i = 0; i < this.#sections.length; i++) {
+			this.#sections[i]!.annotations = entry.annotations[i] ? [...entry.annotations[i]!] : [];
+		}
+		this.#deleted = [...entry.deleted];
+		this.#tocCursor = Math.min(this.#tocCursor, Math.max(0, this.#toc.length - 1));
+		this.#pendingScrollToToc = true;
+		this.callbacks.onPlanEdited?.(joinPlanSections(this.#sections));
+		this.#recomputeFeedback();
+	}
+
+	#startAnnotate(): void {
+		if (this.#toc[this.#tocCursor] === undefined) return;
+		this.#annotating = true;
+		this.#input.setValue("");
+	}
+
+	#submitAnnotation(value: string): void {
+		this.#annotating = false;
+		const note = value.trim();
+		const sectionIndex = this.#toc[this.#tocCursor];
+		if (note && sectionIndex !== undefined) {
+			this.#pushUndo();
+			this.#sections[sectionIndex]!.annotations.push(note);
+			this.#recomputeFeedback();
+		}
+		this.#input.setValue("");
+	}
+
+	#exitAnnotate(): void {
+		this.#annotating = false;
+		this.#input.setValue("");
+	}
+
+	#recomputeFeedback(): void {
+		const annotated = this.#sections.filter(section => section.level >= 1 && section.annotations.length > 0);
+		if (annotated.length === 0 && this.#deleted.length === 0) {
+			this.callbacks.onFeedbackChange?.("");
+			return;
+		}
+		let feedback = "Refinement feedback on the plan:\n";
+		if (this.#deleted.length > 0) {
+			feedback += "\nRemove these sections:\n";
+			for (const title of this.#deleted) feedback += `- ${title}\n`;
+		}
+		for (const section of annotated) {
+			feedback += `\n## ${section.title}\n`;
+			for (const note of section.annotations) feedback += `- ${note}\n`;
+		}
+		this.callbacks.onFeedbackChange?.(feedback);
+	}
+
+	#renderSliderLines(): string[] {
+		const slider = this.#slider;
+		if (!slider) return [];
+		const active = this.#sliderIndex;
+		const track = renderSegmentTrack(slider.segments, active);
+		const leftArrow = theme.fg(active > 0 ? "accent" : "dim", "◂");
+		const rightArrow = theme.fg(active < slider.segments.length - 1 ? "accent" : "dim", "▸");
+		const caption = slider.caption ? `${theme.fg("dim", slider.caption)}  ` : "";
+		const trackLine = `${caption}${leftArrow}  ${track}  ${rightArrow}`;
+		const detail = slider.segments[active]?.detail;
+		if (!detail) return [trackLine];
+		return [trackLine, `  ${theme.fg("dim", "↳")} ${theme.fg("muted", detail)}`];
+	}
+
+	#renderOptionLines(): string[] {
+		const active = this.#focus === "actions";
+		return this.#options.map((label, i) => {
+			const selected = i === this.#selectedIndex;
+			const isDisabled = this.#disabled.has(i);
+			// The cursor marks the selected option; it dims when actions are not the
+			// focused region so the active region's highlight stays unambiguous.
+			const cursor = selected ? theme.fg(active ? "accent" : "dim", `${theme.nav.cursor} `) : "  ";
+			const text = isDisabled
+				? theme.fg("dim", label)
+				: selected && active
+					? theme.bold(theme.fg("accent", label))
+					: theme.fg("text", label);
+			return cursor + text;
+		});
+	}
+
+	#buildHelp(): string {
+		const sep = " · ";
+		const parts: string[] = [];
+		switch (this.#focus) {
+			case "actions":
+				parts.push("↑↓ select", "⏎ confirm");
+				if (this.#slider) parts.push("◂▸ model");
+				break;
+			case "toc":
+				parts.push("↑↓ section", "⏎ open", "a annotate", "d delete", "u undo");
+				break;
+			case "body":
+				parts.push("↑↓ scroll", "⇧ faster", "pgup/pgdn", "g/G ends");
+				break;
+		}
+		parts.push("tab regions");
+		if (this.#externalEditorLabel && this.#focus !== "toc") parts.push(`${this.#externalEditorLabel} editor`);
+		parts.push(this.#helpSuffix);
+		return parts.join(sep);
+	}
+
+	/** Build the concatenated body lines and record each section's start row. */
+	#buildBody(bodyContentWidth: number): string[] {
+		const lines: string[] = [];
+		const offsets: number[] = new Array(this.#sections.length);
+		for (let i = 0; i < this.#sections.length; i++) {
+			const section = this.#sections[i]!;
+			offsets[i] = lines.length;
+			const rendered = section.md.render(bodyContentWidth);
+			if (section.level >= 1 && section.annotations.length > 0 && rendered.length > 0) {
+				lines.push(rendered[0]!);
+				for (const note of section.annotations) {
+					lines.push(`${theme.fg("warning", "▎ ")}${theme.fg("dim", "note: ")}${theme.fg("accent", note)}`);
+				}
+				for (let k = 1; k < rendered.length; k++) lines.push(rendered[k]!);
+			} else {
+				for (const line of rendered) lines.push(line);
+			}
+		}
+		this.#sectionOffsets = offsets;
+		return lines;
+	}
+
+	#sidebarWidthFor(width: number): number {
+		return Math.max(18, Math.min(30, Math.round(width * 0.24)));
+	}
+
+	#sidebarVisible(width: number): boolean {
+		if (this.#toc.length < SIDEBAR_MIN_HEADINGS) return false;
+		if (width < SIDEBAR_MIN_TOTAL_WIDTH) return false;
+		return splitBodyWidth(width, this.#sidebarWidthFor(width)) >= SIDEBAR_MIN_BODY_WIDTH;
+	}
+
+	/** Sidebar lines plus, per row, the ToC position shown there (for clicks). */
+	#renderSidebarLines(
+		regionRows: number,
+		sidebarWidth: number,
+	): { lines: string[]; posForRow: (number | undefined)[] } {
+		// No "Contents" label and no plan-title entry: the box title already says
+		// "Plan Review", so the sidebar is just the bare list of sections, VS
+		// Code-style. Window the entries around the cursor.
+		const lines: string[] = [];
+		const posForRow: (number | undefined)[] = [];
+		const slots = Math.max(0, regionRows);
+		const total = this.#toc.length;
+		let start = 0;
+		if (total > slots) {
+			start = Math.max(0, Math.min(this.#tocCursor - Math.floor(slots / 2), total - slots));
+		}
+		for (let r = 0; r < slots; r++) {
+			const p = start + r;
+			lines.push(p < total ? this.#renderTocEntry(p, sidebarWidth) : "");
+			posForRow.push(p < total ? p : undefined);
+		}
+		return { lines, posForRow };
+	}
+
+	#renderTocEntry(p: number, width: number): string {
+		const section = this.#sections[this.#toc[p]!]!;
+		const highlighted = p === this.#tocCursor;
+		const selected = highlighted && this.#focus === "toc";
+		const glow = highlighted && this.#focus !== "toc";
+		// Compact, VS Code-like rows: a single-column gutter, one space of indent
+		// per nesting level, then the title and an annotation marker.
+		const indent = " ".repeat(Math.max(0, section.level - this.#tocBaseLevel));
+		const ann = section.annotations.length > 0 ? " ✎" : "";
+		const avail = Math.max(0, width - 1 - indent.length - visibleWidth(ann));
+		const title = truncateToWidth(section.title || "(untitled)", avail, Ellipsis.Unicode);
+		const body = indent + title + ann;
+		// Single-column gutter glyph: a cursor `›` on the focused selection, an
+		// accent bar `▎` on the current scrolled section, otherwise blank. The
+		// glyph keeps the cursor legible even where the selection background is
+		// subtle; the focused row also gets the full-row highlight.
+		const gutter = selected ? "›" : glow ? "▎" : " ";
+		const line = gutter + body;
+		if (selected) return theme.bg("selectedBg", theme.bold(fit(line, width)));
+		if (glow) return theme.fg("accent", line);
+		return theme.fg("muted", line);
+	}
+
+	#renderFooterLines(innerWidth: number): string[] {
+		if (this.#annotating) {
+			const section = this.#sections[this.#toc[this.#tocCursor]!];
+			const title = section?.title ?? "";
+			const caption = `${theme.fg("dim", "Annotate")} ${theme.fg("accent", `‹${title}›`)}`;
+			return [caption, this.#input.render(innerWidth)[0] ?? ""];
+		}
+		return [theme.fg("dim", this.#buildHelp())];
+	}
+
+	render(width: number): string[] {
+		const termHeight = process.stdout.rows || 40;
+		const sidebarShown = this.#sidebarVisible(width);
+		this.#sidebarShown = sidebarShown;
+		const sidebarWidth = sidebarShown ? this.#sidebarWidthFor(width) : 0;
+		const innerWidth = Math.max(1, width - 4);
+		const bodyContentWidth = sidebarShown ? splitBodyWidth(width, sidebarWidth) : innerWidth;
+
+		const sliderLines = this.#renderSliderLines();
+		const optionLines = this.#renderOptionLines();
+		const promptLines = this.#promptTitle ? [theme.bold(theme.fg("accent", this.#promptTitle))] : [];
+		const footerLines = this.#renderFooterLines(innerWidth);
+
+		// Chrome rows: top border, two dividers, bottom border, plus the
+		// prompt/slider/option/footer rows between them.
+		const chrome = 4 + promptLines.length + sliderLines.length + optionLines.length + footerLines.length;
+		const regionRows = Math.max(MIN_BODY_ROWS, termHeight - chrome);
+
+		const bodyLines = this.#buildBody(bodyContentWidth);
+		this.#scrollView.setLines(bodyLines);
+		this.#scrollView.setHeight(regionRows);
+		if (this.#pendingScrollToToc) {
+			this.#pendingScrollToToc = false;
+			this.#scrubBodyToToc();
+		}
+		if (this.#focus !== "toc") this.#tocCursor = this.#deriveTocCursorFromScroll();
+		const body = this.#scrollView.render(bodyContentWidth);
+
+		this.#optionClickRows.clear();
+		this.#tocClickRows.clear();
+		this.#bodyClickRows.clear();
+		this.#sidebarClickMaxCol = sidebarShown ? sidebarWidth + 3 : 0;
+
+		const out: string[] = [];
+		if (sidebarShown) {
+			const { lines: sidebar, posForRow } = this.#renderSidebarLines(regionRows, sidebarWidth);
+			out.push(topBorderSplit(width, OVERLAY_TITLE, sidebarWidth));
+			for (let i = 0; i < regionRows; i++) {
+				const pos = posForRow[i];
+				if (pos !== undefined) this.#tocClickRows.set(out.length, pos);
+				this.#bodyClickRows.add(out.length);
+				out.push(splitRow(sidebar[i] ?? "", body[i] ?? "", width, sidebarWidth));
+			}
+			out.push(dividerSplit(width, sidebarWidth));
+		} else {
+			out.push(topBorder(width, OVERLAY_TITLE));
+			for (const line of body) {
+				this.#bodyClickRows.add(out.length);
+				out.push(row(line, width));
+			}
+			out.push(divider(width));
+		}
+		for (const line of promptLines) out.push(row(line, width));
+		for (const line of sliderLines) out.push(row(line, width));
+		for (let i = 0; i < optionLines.length; i++) {
+			this.#optionClickRows.set(out.length, i);
+			out.push(row(optionLines[i]!, width));
+		}
+		out.push(divider(width));
+		for (const line of footerLines) out.push(row(line, width));
+		out.push(bottomBorder(width));
+		return out;
+	}
+}
diff --git a/packages/coding-agent/src/modes/components/plan-toc.ts b/packages/coding-agent/src/modes/components/plan-toc.ts
new file mode 100644
index 000000000..03ab43c0c
--- /dev/null
+++ b/packages/coding-agent/src/modes/components/plan-toc.ts
@@ -0,0 +1,138 @@
+/**
+ * Pure heading/section parser for the plan-review overlay. It splits a plan's
+ * markdown into a flat list of sections — a leading preamble (text before the
+ * first heading) followed by one entry per ATX heading — preserving the exact
+ * source bytes of each section so the overlay can render, reorder-free delete,
+ * and round-trip the document without a full markdown re-render.
+ *
+ * No TUI dependencies: this module is unit-tested in isolation.
+ */
+
+/** ATX heading: 1-6 `#`, required whitespace, a title, optional closing `#`s. */
+const HEADING_RE = /^(#{1,6})[ \t]+(.+?)[ \t]*#*[ \t]*$/;
+/** Opening/closing code fence run (``` or ~~~), allowing up to 3 lead spaces. */
+const FENCE_RE = /^ {0,3}(`{3,}|~{3,})(.*)$/;
+
+export interface PlanSection {
+	/** `0` = preamble (no heading, no ToC entry); `1..6` = heading depth. */
+	level: number;
+	/** Plain-text heading label with inline markdown lightly stripped. */
+	title: string;
+	/** Exact source slice for this section, including its trailing newline(s). */
+	raw: string;
+}
+
+/**
+ * Collapse inline markdown emphasis/link/code syntax to readable text. This is
+ * a deliberately light strip (not a full markdown render) just so ToC entries
+ * read cleanly — `**Goal** & [docs](x)` becomes `Goal & docs`.
+ */
+export function stripInlineMarkdown(text: string): string {
+	let out = text;
+	// Images first (so the link pass below does not eat the `(url)`), then links.
+	out = out.replace(/!\[([^\]]*)\]\([^)]*\)/g, "$1");
+	out = out.replace(/\[([^\]]*)\]\([^)]*\)/g, "$1");
+	out = out.replace(/\[([^\]]*)\]\[[^\]]*\]/g, "$1");
+	// Autolinks `<https://…>` keep their URL as the readable text.
+	out = out.replace(/<([^>\s]+)>/g, "$1");
+	// Inline code, then bold/italic/strikethrough emphasis runs.
+	out = out.replace(/`([^`]+)`/g, "$1");
+	out = out.replace(/(\*\*|__)(.+?)\1/g, "$2");
+	out = out.replace(/(\*|_)(.+?)\1/g, "$2");
+	out = out.replace(/~~(.+?)~~/g, "$1");
+	return out.replace(/\s+/g, " ").trim();
+}
+
+/**
+ * Split `text` into preamble + heading sections. `#` characters inside fenced
+ * code blocks are never treated as headings. Concatenating every section's
+ * `raw` reproduces the original text exactly.
+ */
+export function parsePlanSections(text: string): PlanSection[] {
+	const lines = text.split("\n");
+	// Character offset of each line start so section `raw` can slice the source.
+	const offsets: number[] = new Array(lines.length);
+	let cursor = 0;
+	for (let i = 0; i < lines.length; i++) {
+		offsets[i] = cursor;
+		cursor += lines[i]!.length + 1; // +1 for the "\n" join separator
+	}
+
+	// Heading line indices (start of each heading section), with metadata.
+	const heads: { line: number; level: number; title: string }[] = [];
+	let fenceChar: string | null = null;
+	let fenceLen = 0;
+	for (let i = 0; i < lines.length; i++) {
+		const line = lines[i]!;
+		const fence = FENCE_RE.exec(line);
+		if (fenceChar === null) {
+			if (fence) {
+				fenceChar = fence[1]![0]!;
+				fenceLen = fence[1]!.length;
+			}
+			// Opening-fence lines are body, not headings.
+			if (fence) continue;
+		} else {
+			// Inside a fence: only a matching-or-longer run of the same char closes.
+			if (fence && fence[1]![0] === fenceChar && fence[1]!.length >= fenceLen && fence[2]!.trim() === "") {
+				fenceChar = null;
+				fenceLen = 0;
+			}
+			continue;
+		}
+		const heading = HEADING_RE.exec(line);
+		if (heading) {
+			heads.push({ line: i, level: heading[1]!.length, title: stripInlineMarkdown(heading[2]!) });
+		}
+	}
+
+	const sections: PlanSection[] = [];
+	const sliceRaw = (startLine: number, endLine: number): string => {
+		const startOffset = offsets[startLine]!;
+		const endOffset = endLine < lines.length ? offsets[endLine]! : text.length;
+		return text.slice(startOffset, endOffset);
+	};
+
+	// Preamble: everything before the first heading (only when non-empty).
+	const firstHeadLine = heads.length > 0 ? heads[0]!.line : lines.length;
+	if (firstHeadLine > 0) {
+		const raw = sliceRaw(0, firstHeadLine);
+		if (raw.length > 0) sections.push({ level: 0, title: "", raw });
+	}
+
+	for (let h = 0; h < heads.length; h++) {
+		const head = heads[h]!;
+		const endLine = h + 1 < heads.length ? heads[h + 1]!.line : lines.length;
+		sections.push({ level: head.level, title: head.title, raw: sliceRaw(head.line, endLine) });
+	}
+
+	return sections;
+}
+
+/**
+ * Concatenate every section's `raw` back into a single document and guarantee a
+ * single trailing newline. Inverse of {@link parsePlanSections} for any input
+ * that already ends with a newline.
+ */
+export function joinPlanSections(sections: readonly PlanSection[]): string {
+	let joined = "";
+	for (const section of sections) joined += section.raw;
+	if (joined.length === 0) return "";
+	return joined.endsWith("\n") ? joined : `${joined}\n`;
+}
+
+/**
+ * Indices to remove when deleting `sections[index]`: the heading itself plus
+ * every following section nested deeper than it (its sub-headings). The
+ * preamble (level 0) is never a deletion target and yields an empty span.
+ */
+export function sectionDeletionSpan(sections: readonly PlanSection[], index: number): number[] {
+	const target = sections[index];
+	if (!target || target.level === 0) return [];
+	const span = [index];
+	for (let j = index + 1; j < sections.length; j++) {
+		if (sections[j]!.level > target.level) span.push(j);
+		else break;
+	}
+	return span;
+}
diff --git a/packages/coding-agent/src/modes/components/read-tool-group.ts b/packages/coding-agent/src/modes/components/read-tool-group.ts
index 5a236cb65..1af94f4ba 100644
--- a/packages/coding-agent/src/modes/components/read-tool-group.ts
+++ b/packages/coding-agent/src/modes/components/read-tool-group.ts
@@ -81,6 +81,14 @@ export class ReadToolGroupComponent extends Container implements ToolExecutionHa
 	#text: Text;
 	#expanded = false;
 	#showContentPreview: boolean;
+	// A read group accretes entries across multiple assistant completions for as
+	// long as the run of reads is uninterrupted. While it is the active group it
+	// must stay in the transcript's repaintable live region — its header line
+	// re-layouts from `Read <path>` to `Read (N)` + tree as entries arrive, so a
+	// frozen snapshot taken on a risk terminal would strand the single-entry form
+	// (see TranscriptContainer / NativeScrollbackLiveRegion). The controller calls
+	// `finalize()` once the run breaks so the block can commit to native scrollback.
+	#finalized = false;
 
 	constructor(options: ReadToolGroupOptions = {}) {
 		super();
@@ -90,6 +98,14 @@ export class ReadToolGroupComponent extends Container implements ToolExecutionHa
 		this.#updateDisplay();
 	}
 
+	isTranscriptBlockFinalized(): boolean {
+		return this.#finalized;
+	}
+
+	finalize(): void {
+		this.#finalized = true;
+	}
+
 	updateArgs(args: ReadRenderArgs, toolCallId?: string): void {
 		if (!toolCallId) return;
 		const basePath = args.file_path || args.path || "";
@@ -181,9 +197,9 @@ export class ReadToolGroupComponent extends Container implements ToolExecutionHa
 		const total = entriesWithoutPreview.length;
 		for (const [index, entry] of entriesWithoutPreview.entries()) {
 			const connector = index === total - 1 ? theme.tree.last : theme.tree.branch;
-			const statusSymbol = this.#formatStatus(entry.status);
+			const statusPrefix = entry.status === "success" ? "" : `${this.#formatStatus(entry.status)} `;
 			const pathDisplay = this.#formatPath(entry);
-			lines.push(`   ${theme.fg("dim", connector)} ${statusSymbol} ${pathDisplay}`.trimEnd());
+			lines.push(`   ${theme.fg("dim", connector)} ${statusPrefix}${pathDisplay}`.trimEnd());
 		}
 
 		this.#text.setText(lines.join("\n"));
@@ -198,7 +214,7 @@ export class ReadToolGroupComponent extends Container implements ToolExecutionHa
 
 	/**
 	 * Add a code-cell content preview below the entry summary.
-	 * When collapsed: shows first COLLAPSED_PREVIEW_LINES lines with "… N more lines (Ctrl+O for more)" hint.
+	 * When collapsed: shows first COLLAPSED_PREVIEW_LINES lines with a "… N more lines ⟨<key>: Expand⟩" hint.
 	 * When expanded: shows full content.
 	 */
 	#addContentPreview(entry: ReadEntry): void {
@@ -254,7 +270,7 @@ export class ReadToolGroupComponent extends Container implements ToolExecutionHa
 
 	#formatStatus(status: ReadEntry["status"]): string {
 		if (status === "success") {
-			return theme.fg("success", theme.status.success);
+			return theme.fg("text", theme.status.enabled);
 		}
 		if (status === "warning") {
 			return theme.fg("warning", theme.status.warning);
diff --git a/packages/coding-agent/src/modes/components/session-observer-overlay.ts b/packages/coding-agent/src/modes/components/session-observer-overlay.ts
index 8f62b4673..c484cf593 100644
--- a/packages/coding-agent/src/modes/components/session-observer-overlay.ts
+++ b/packages/coding-agent/src/modes/components/session-observer-overlay.ts
@@ -15,7 +15,7 @@
  *   - Enter on main session -> close overlay (jump back)
  */
 import type { ToolResultMessage } from "@oh-my-pi/pi-ai";
-import { Container, Markdown, type MarkdownTheme, matchesKey } from "@oh-my-pi/pi-tui";
+import { Container, Markdown, type MarkdownTheme, matchesKey, ScrollView } from "@oh-my-pi/pi-tui";
 import { formatDuration, formatNumber, logger } from "@oh-my-pi/pi-utils";
 import type { KeyId } from "../../config/keybindings";
 import { isSilentAbort } from "../../session/messages";
@@ -230,23 +230,21 @@ export class SessionObserverOverlayComponent extends Container {
 		lines.push(...new DynamicBorder().render(width));
 
 		// --- Scrolled content viewport ---
-		const visibleLines = this.#renderedLines.slice(this.#scrollOffset, this.#scrollOffset + this.#viewportHeight);
-		for (const vl of visibleLines) {
-			lines.push(` ${vl}`);
-		}
-		// Pad to fill viewport if content is shorter
-		const pad = this.#viewportHeight - visibleLines.length;
-		for (let i = 0; i < pad; i++) {
-			lines.push("");
-		}
+		const sv = new ScrollView(
+			this.#renderedLines.slice(this.#scrollOffset, this.#scrollOffset + this.#viewportHeight),
+			{
+				height: this.#viewportHeight,
+				scrollbar: "auto",
+				totalRows: this.#renderedLines.length,
+				theme: { track: t => theme.fg("dim", t), thumb: t => theme.fg("accent", t) },
+			},
+		);
+		sv.setScrollOffset(this.#scrollOffset);
+		for (const row of sv.render(Math.max(1, width - 1))) lines.push(` ${row}`);
 
 		// --- Footer ---
-		const scrollInfo =
-			this.#renderedLines.length > this.#viewportHeight
-				? ` ${theme.fg("dim", `[${this.#scrollOffset + 1}-${Math.min(this.#scrollOffset + this.#viewportHeight, this.#renderedLines.length)}/${this.#renderedLines.length}]`)}`
-				: "";
 		lines.push("");
-		lines.push(` ${this.#viewerFooterLines[0] ?? ""}${scrollInfo}`);
+		lines.push(` ${this.#viewerFooterLines[0] ?? ""}`);
 		for (let i = 1; i < this.#viewerFooterLines.length; i++) {
 			lines.push(` ${this.#viewerFooterLines[i]}`);
 		}
diff --git a/packages/coding-agent/src/modes/components/session-selector.ts b/packages/coding-agent/src/modes/components/session-selector.ts
index 87a7185e0..520a5dbcc 100644
--- a/packages/coding-agent/src/modes/components/session-selector.ts
+++ b/packages/coding-agent/src/modes/components/session-selector.ts
@@ -6,6 +6,7 @@ import {
 	matchesKey,
 	padding,
 	replaceTabs,
+	ScrollView,
 	Spacer,
 	Text,
 	truncateToWidth,
@@ -247,6 +248,11 @@ class SessionList implements Component {
 		const endIndex = Math.min(startIndex + maxVisible, this.#filteredSessions.length);
 
 		// Render visible sessions (3 lines, or 4 when a title adds a preview line).
+		// Each session block is built into sessionLines, then wrapped by ScrollView
+		// so the right-edge scrollbar is proportional at the physical-line level.
+		const sessionLines: string[] = [];
+		const overflow = this.#filteredSessions.length > maxVisible;
+		const rowWidth = Math.max(0, width - (overflow ? 1 : 0));
 		for (let i = startIndex; i < endIndex; i++) {
 			const session = this.#filteredSessions[i];
 			const isSelected = i === this.#selectedIndex;
@@ -258,22 +264,22 @@ class SessionList implements Component {
 			const cursorSymbol = `${theme.nav.cursor} `;
 			const cursorWidth = visibleWidth(cursorSymbol);
 			const cursor = isSelected ? theme.fg("accent", cursorSymbol) : padding(cursorWidth);
-			const maxWidth = width - cursorWidth; // Account for cursor width
+			const maxWidth = rowWidth - cursorWidth; // Account for cursor width
 
 			if (session.title) {
 				// Has title: show title on first line, dimmed first message on second line
 				const truncatedTitle = truncateToWidth(session.title, maxWidth);
 				const titleLine = cursor + (isSelected ? theme.bold(truncatedTitle) : truncatedTitle);
-				lines.push(titleLine);
+				sessionLines.push(titleLine);
 
 				// Second line: dimmed first message preview
 				const truncatedPreview = truncateToWidth(normalizedMessage, maxWidth);
-				lines.push(`  ${theme.fg("dim", truncatedPreview)}`);
+				sessionLines.push(`  ${theme.fg("dim", truncatedPreview)}`);
 			} else {
 				// No title: show first message as main line
 				const truncatedMsg = truncateToWidth(normalizedMessage, maxWidth);
 				const messageLine = cursor + (isSelected ? theme.bold(truncatedMsg) : truncatedMsg);
-				lines.push(messageLine);
+				sessionLines.push(messageLine);
 			}
 
 			// Metadata line: date + file size + lifecycle status (+ project dir in
@@ -290,18 +296,23 @@ class SessionList implements Component {
 			if (this.#showCwd && session.cwd) {
 				metadata += ` ${dot} ${dim(shortenPath(session.cwd))}`;
 			}
-			const metadataLine = truncateToWidth(metadata, width);
+			const metadataLine = truncateToWidth(metadata, rowWidth);
 
-			lines.push(metadataLine);
-			lines.push(""); // Blank line between sessions
+			sessionLines.push(metadataLine);
+			sessionLines.push(""); // Blank line between sessions
 		}
 
-		// Add scroll indicator if needed
-		if (startIndex > 0 || endIndex < this.#filteredSessions.length) {
-			const scrollText = `  (${this.#selectedIndex + 1}/${this.#filteredSessions.length})`;
-			const scrollInfo = theme.fg("muted", truncateToWidth(scrollText, width));
-			lines.push(scrollInfo);
-		}
+		// Wrap the rendered window in a ScrollView for a proportional right-edge bar.
+		const visibleCount = endIndex - startIndex;
+		const linesPerItem = visibleCount > 0 ? sessionLines.length / visibleCount : 1;
+		const sv = new ScrollView(sessionLines, {
+			height: sessionLines.length,
+			scrollbar: "auto",
+			totalRows: Math.round(this.#filteredSessions.length * linesPerItem),
+			theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+		});
+		sv.setScrollOffset(Math.round(startIndex * linesPerItem));
+		lines.push(...sv.render(width));
 
 		// Add keybinding hint
 		lines.push("");
diff --git a/packages/coding-agent/src/modes/components/skill-message.ts b/packages/coding-agent/src/modes/components/skill-message.ts
index 3187b30da..f2aeb41d0 100644
--- a/packages/coding-agent/src/modes/components/skill-message.ts
+++ b/packages/coding-agent/src/modes/components/skill-message.ts
@@ -11,7 +11,6 @@ export class SkillMessageComponent extends Container {
 
 	constructor(private readonly message: CustomMessage<SkillPromptDetails>) {
 		super();
-		this.addChild(new Spacer(1));
 
 		this.#box = new Box(1, 1, t => theme.bg("customMessageBg", t));
 		this.#rebuild();
diff --git a/packages/coding-agent/src/modes/components/status-line.ts b/packages/coding-agent/src/modes/components/status-line.ts
index 26f6e10f4..6c8b338d4 100644
--- a/packages/coding-agent/src/modes/components/status-line.ts
+++ b/packages/coding-agent/src/modes/components/status-line.ts
@@ -546,7 +546,7 @@ export class StatusLineComponent implements Component {
 		return `${modelId}|${sp.length}:${sp[0]?.length ?? 0}|${tools.length}|${skills.length}`;
 	}
 
-	#buildSegmentContext(width: number): SegmentContext {
+	#buildSegmentContext(width: number, segmentOptions: StatusLineSettings["segmentOptions"]): SegmentContext {
 		const state = this.session.state;
 
 		// Trigger background fetch (5-min TTL); render uses cached value
@@ -575,7 +575,7 @@ export class StatusLineComponent implements Component {
 		return {
 			session: this.session,
 			width,
-			options: this.#resolveSettings().segmentOptions ?? {},
+			options: segmentOptions ?? {},
 			planMode: this.#planModeStatus,
 			loopMode: this.#loopModeStatus,
 			goalMode: this.#goalModeStatus,
@@ -632,8 +632,8 @@ export class StatusLineComponent implements Component {
 	}
 
 	#buildStatusLine(width: number): string {
-		const ctx = this.#buildSegmentContext(width);
 		const effectiveSettings = this.#resolveSettings();
+		const ctx = this.#buildSegmentContext(width, effectiveSettings.segmentOptions);
 		const separatorDef = getSeparator(effectiveSettings.separator ?? "powerline-thin", theme);
 
 		const bgAnsi = theme.getBgAnsi("statusLineBg");
@@ -759,8 +759,6 @@ export class StatusLineComponent implements Component {
 			return leftGroup + (leftGroup && rightGroup ? " " : "") + rightGroup;
 		}
 
-		leftWidth = groupWidth(left, leftCapWidth, leftSepWidth);
-		rightWidth = groupWidth(right, rightCapWidth, rightSepWidth);
 		const gapWidth = Math.max(1, topFillWidth - leftWidth - rightWidth);
 		const sessionName =
 			effectiveSettings.sessionAccent !== false ? this.session.sessionManager?.getSessionName() : undefined;
diff --git a/packages/coding-agent/src/modes/components/tips.txt b/packages/coding-agent/src/modes/components/tips.txt
index e685aaf0c..85c89573f 100644
--- a/packages/coding-agent/src/modes/components/tips.txt
+++ b/packages/coding-agent/src/modes/components/tips.txt
@@ -1,5 +1,6 @@
 Tired of typing "keep going"? Just send a '.'
 You can /btw to ask a side question
+Use /tan to fork the current conversation into a background agent
 Ctrl+D can be used to exit, but with your draft saved!
 Find out which model you emotionally abuse the most with `omp stats`
 Try task isolation to create CoW worktrees
diff --git a/packages/coding-agent/src/modes/components/todo-reminder.ts b/packages/coding-agent/src/modes/components/todo-reminder.ts
index 0a536d73d..2d8bcf173 100644
--- a/packages/coding-agent/src/modes/components/todo-reminder.ts
+++ b/packages/coding-agent/src/modes/components/todo-reminder.ts
@@ -16,8 +16,6 @@ export class TodoReminderComponent extends Container {
 	) {
 		super();
 
-		this.addChild(new Spacer(1));
-
 		this.#box = new Box(1, 1, t => theme.inverse(theme.fg("warning", t)));
 		this.addChild(this.#box);
 
diff --git a/packages/coding-agent/src/modes/components/tool-execution.ts b/packages/coding-agent/src/modes/components/tool-execution.ts
index 5c89070bd..aee73cc06 100644
--- a/packages/coding-agent/src/modes/components/tool-execution.ts
+++ b/packages/coding-agent/src/modes/components/tool-execution.ts
@@ -31,20 +31,12 @@ import {
 	renderJsonTreeLines,
 } from "../../tools/json-tree";
 import { formatExpandHint, replaceTabs, resolveImageOptions, truncateToWidth } from "../../tools/render-utils";
-import { toolRenderers } from "../../tools/renderers";
+import { type ToolRenderer, toolRenderers } from "../../tools/renderers";
 import { TODO_STRIKE_TOTAL_FRAMES } from "../../tools/todo";
-import { renderStatusLine } from "../../tui";
+import { isFramedBlockComponent, renderStatusLine } from "../../tui";
 import { sanitizeWithOptionalSixelPassthrough } from "../../utils/sixel";
 import { renderDiff } from "./diff";
 
-function ensureInvalidate(component: unknown): Component {
-	const c = component as { render: Component["render"]; invalidate?: () => void };
-	if (!c.invalidate) {
-		c.invalidate = () => {};
-	}
-	return c as Component;
-}
-
 /**
  * Drop trailing removal/hunk-header lines that appear in a streaming diff
  * before the matching `+added` lines have arrived. Without this, a partial
@@ -107,7 +99,7 @@ function rawTextInputFromPartialJson(partialJson: unknown): string | undefined {
 	// Function-tool arguments stream as JSON. Custom/free-form tools stream raw
 	// text in the same transport field; only the raw form is a valid fallback for
 	// the conventional `input` parameter.
-	if (first === "{" || first === "[" || first === '"') return undefined;
+	if (first === "{" || first === '"') return undefined;
 	return partialJson;
 }
 
@@ -141,12 +133,12 @@ export interface ToolExecutionHandle {
 	setExpanded(expanded: boolean): void;
 }
 
-/** Drive pending-tool redraws at ~60fps so the animated border sweep is smooth.
- * The TUI already throttles at its 16ms `MIN_RENDER_INTERVAL_MS`, so this is the
- * natural upper bound and static frames diff to a no-op redraw at ~zero cost. */
-const SPINNER_RENDER_INTERVAL_MS = 16;
+/** Drive pending-tool redraws at 30fps so the animated border sweep stays
+ * smooth without spending twice the frame budget. The TUI throttles at the same
+ * cadence, and static frames diff to a no-op redraw at ~zero cost. */
+const SPINNER_RENDER_INTERVAL_MS = 1000 / 30;
 /** Advance the spinner glyph at its classic ~12.5fps step, decoupled from the
- * 60fps render cadence (mirrors `Loader`). */
+ * render cadence (mirrors `Loader`). */
 const SPINNER_GLYPH_ADVANCE_MS = 80;
 
 // Stable per-instance counter so each tool execution's inline images get a
@@ -233,11 +225,13 @@ export class ToolExecutionComponent extends Container {
 		this.#cwd = cwd;
 		this.#args = args;
 
-		this.addChild(new Spacer(1));
-
-		// Always create both - contentBox for custom tools/bash/tools with renderers, contentText for other built-ins
-		this.#contentBox = new Box(1, 1, (text: string) => theme.bg("toolPendingBg", text));
-		this.#contentText = new Text("", 1, 1, (text: string) => theme.bg("toolPendingBg", text));
+		// Always create both - contentBox for custom tools/bash/tools with renderers, contentText for other built-ins.
+		// paddingY is 1 so background-tinted blocks (custom/extension tools and the
+		// generic fallback) get top/bottom breathing room. TranscriptContainer
+		// strips PLAIN-blank edges, so framed/minimal blocks (no bg set) drop these
+		// lines and keep their tight spacing — only tinted lines survive.
+		this.#contentBox = new Box(0, 1);
+		this.#contentText = new Text("", 1, 1);
 
 		// Use Box for custom tools or built-in tools that have renderers
 		const hasRenderer = toolName in toolRenderers;
@@ -427,26 +421,43 @@ export class ToolExecutionComponent extends Container {
 	#updateSpinnerAnimation(): void {
 		// Spinner for: task tool with partial result, or edit/write while args streaming
 		const isStreamingArgs = !this.#argsComplete && (isEditLikeToolName(this.#toolName) || this.#toolName === "write");
-		const isBackgroundAsyncTask =
-			this.#toolName === "task" &&
+		const isBackgroundAsyncRunning =
 			(this.#result?.details as { async?: { state?: string } } | undefined)?.async?.state === "running";
+		const isBackgroundAsyncTask = this.#toolName === "task" && isBackgroundAsyncRunning;
 		const isPartialTask = this.#isPartial && this.#toolName === "task" && !isBackgroundAsyncTask;
-		// Sweep the border of bash/eval execution blocks while they're pending.
+		// Sweep the border of bash/eval execution blocks while they're pending — but
+		// not once they've been backgrounded: a backgrounded job's block gets
+		// committed to scrollback and finalizes later via the async update path, so a
+		// mid-sweep frame would freeze a stray dark "bar" segment into the border.
 		const isPendingExecBlock =
-			this.#isPartial && shimmerEnabled() && (this.#toolName === "bash" || this.#toolName === "eval");
+			this.#isPartial &&
+			shimmerEnabled() &&
+			(this.#toolName === "bash" || this.#toolName === "eval") &&
+			!isBackgroundAsyncRunning;
 		const needsSpinner = isStreamingArgs || isPartialTask || isPendingExecBlock;
 		if (needsSpinner && !this.#spinnerInterval) {
-			this.#lastSpinnerAdvanceAt = performance.now();
+			const now = performance.now();
+			const frameCount = theme.spinnerFrames.length;
+			this.#lastSpinnerAdvanceAt = now;
+			if (frameCount > 0 && this.#spinnerFrame === undefined) {
+				this.#spinnerFrame = 0;
+				this.#renderState.spinnerFrame = 0;
+			}
 			this.#spinnerInterval = setInterval(() => {
 				const now = performance.now();
 				const frameCount = theme.spinnerFrames.length;
-				// Redraw at ~60fps for a smooth border sweep, but only step the spinner
-				// glyph at its classic ~12.5fps cadence. The TUI throttles renders at
-				// 16ms and the differ drops no-op redraws, so the extra ticks are free.
-				if (frameCount > 0 && now - this.#lastSpinnerAdvanceAt >= SPINNER_GLYPH_ADVANCE_MS) {
-					this.#spinnerFrame = ((this.#spinnerFrame ?? -1) + 1) % frameCount;
-					this.#renderState.spinnerFrame = this.#spinnerFrame;
-					this.#lastSpinnerAdvanceAt = now;
+				// Redraw at 30fps for a smooth border sweep, but keep the spinner
+				// glyph phase-locked to its classic ~12.5fps cadence. Advancing the
+				// anchor by elapsed frames instead of resetting to `now` avoids the
+				// 30fps timer quantizing the glyph down to one step every three ticks.
+				if (frameCount > 0) {
+					const elapsed = now - this.#lastSpinnerAdvanceAt;
+					if (elapsed >= SPINNER_GLYPH_ADVANCE_MS) {
+						const steps = Math.floor(elapsed / SPINNER_GLYPH_ADVANCE_MS);
+						this.#spinnerFrame = ((this.#spinnerFrame ?? 0) + steps) % frameCount;
+						this.#renderState.spinnerFrame = this.#spinnerFrame;
+						this.#lastSpinnerAdvanceAt += steps * SPINNER_GLYPH_ADVANCE_MS;
+					}
 				}
 				this.#ui.requestRender();
 			}, SPINNER_RENDER_INTERVAL_MS);
@@ -518,6 +529,39 @@ export class ToolExecutionComponent extends Container {
 		return (this.#result.details as { async?: { state?: string } } | undefined)?.async?.state === "running";
 	}
 
+	/**
+	 * While a tool's preview is still streaming, a block whose preview is
+	 * append-only (rows only grow at the bottom, never re-layout) lets the
+	 * renderer commit the scrolled-off head of an over-tall preview to native
+	 * scrollback instead of dropping it — the same anti-yank path a streaming
+	 * assistant reply uses (see {@link TranscriptContainer} +
+	 * `NativeScrollbackLiveRegion`). Covers both phases: a pre-result call preview
+	 * (a `write` whose content streams in) and a partial-result preview that
+	 * streams output below fixed input (an `eval`/`bash` whose stdout grows under
+	 * its code cell). Gated on {@link isTranscriptBlockFinalized} so the boundary
+	 * closes the instant the block reaches a terminal state — a final result that
+	 * may collapse to a compact view, a backgrounded async tool, or a seal — and
+	 * the renderer decides whether its current preview shape qualifies via
+	 * `isStreamingPreviewAppendOnly` (typically: only the expanded full view,
+	 * which is top-anchored; the collapsed tail window re-layouts but is bounded
+	 * so it never overflows anyway).
+	 */
+	isTranscriptBlockAppendOnly(): boolean {
+		// A finalized block's preview can collapse/re-layout; only a live,
+		// still-streaming block is a candidate.
+		if (this.isTranscriptBlockFinalized()) return false;
+		const predicate =
+			(this.#tool as { isStreamingPreviewAppendOnly?: ToolRenderer["isStreamingPreviewAppendOnly"] } | undefined)
+				?.isStreamingPreviewAppendOnly ?? toolRenderers[this.#toolName]?.isStreamingPreviewAppendOnly;
+		if (!predicate) return false;
+		try {
+			return predicate(this.#getCallArgsForRender(), this.#renderState, this.#result);
+		} catch (err) {
+			logger.warn("Tool append-only predicate failed", { tool: this.#toolName, error: String(err) });
+			return false;
+		}
+	}
+
 	/**
 	 * Mark the tool terminal even though no result arrived (the turn aborted or
 	 * abandoned it) and stop animating, so it can freeze and stops pinning the
@@ -562,25 +606,23 @@ export class ToolExecutionComponent extends Container {
 	}
 
 	#updateDisplay(): void {
-		// Set background based on state
-		const bgFn = this.#isPartial
-			? (text: string) => theme.bg("toolPendingBg", text)
-			: this.#result?.isError
-				? (text: string) => theme.bg("toolErrorBg", text)
-				: (text: string) => theme.bg("toolSuccessBg", text);
-
 		// Sync shared mutable render state for component closures
 		this.#renderState.expanded = this.#expanded;
 		this.#renderState.isPartial = this.#isPartial;
 		this.#renderState.spinnerFrame = this.#spinnerFrame;
 
+		// Non-self-framing tools (custom/extension renderers and the generic
+		// fallback) get a padded, state-tinted block — built-ins that draw their
+		// own frame opt out below via the framed-component mark.
+		const stateBgKey = this.#isPartial ? "toolPendingBg" : this.#result?.isError ? "toolErrorBg" : "toolSuccessBg";
+		const stateBgFn = (t: string) => theme.bg(stateBgKey, t);
+
 		// Check for custom tool rendering
 		if (this.#tool && (this.#tool.renderCall || this.#tool.renderResult)) {
 			const tool = this.#tool;
 			const mergeCallAndResult = Boolean((tool as { mergeCallAndResult?: boolean }).mergeCallAndResult);
 			// Custom tools use Box for flexible component rendering
-			const inline = Boolean((tool as { inline?: boolean }).inline);
-			this.#contentBox.setBgFn(inline ? undefined : bgFn);
+			this.#contentBox.setBgFn(undefined);
 			this.#contentBox.clear();
 			// Mirror the built-in renderer branch so custom renderers (notably the
 			// task tool, whose live instance routes through here) receive the same
@@ -588,22 +630,25 @@ export class ToolExecutionComponent extends Container {
 			// call preview once result lines exist.
 			this.#renderState.renderContext = this.#buildRenderContext();
 
-			// Render call component
+			// Render call component. The fallback label only stands in for a
+			// missing `renderCall`; when the call is intentionally suppressed
+			// (mergeCallAndResult once a result exists) we render nothing here so
+			// the result component isn't preceded by a redundant tool-name line.
 			const shouldRenderCall = !this.#result || !mergeCallAndResult;
-			if (shouldRenderCall && tool.renderCall) {
-				try {
-					const callComponent = tool.renderCall(this.#getCallArgsForRender(), this.#renderState, theme);
-					if (callComponent) {
-						this.#contentBox.addChild(ensureInvalidate(callComponent));
+			if (shouldRenderCall) {
+				if (tool.renderCall) {
+					try {
+						const callComponent = tool.renderCall(this.#getCallArgsForRender(), this.#renderState, theme);
+						if (callComponent) this.#contentBox.addChild(callComponent as Component);
+					} catch (err) {
+						logger.warn("Tool renderer failed", { tool: this.#toolName, error: String(err) });
+						// Fall back to default on error
+						this.#contentBox.addChild(new Text(theme.fg("toolTitle", theme.bold(this.#toolLabel)), 0, 0));
 					}
-				} catch (err) {
-					logger.warn("Tool renderer failed", { tool: this.#toolName, error: String(err) });
-					// Fall back to default on error
+				} else {
+					// No custom renderCall, show tool name
 					this.#contentBox.addChild(new Text(theme.fg("toolTitle", theme.bold(this.#toolLabel)), 0, 0));
 				}
-			} else {
-				// No custom renderCall, show tool name
-				this.#contentBox.addChild(new Text(theme.fg("toolTitle", theme.bold(this.#toolLabel)), 0, 0));
 			}
 
 			// Render result component if we have a result
@@ -625,9 +670,7 @@ export class ToolExecutionComponent extends Container {
 						theme,
 						this.#args,
 					);
-					if (resultComponent) {
-						this.#contentBox.addChild(ensureInvalidate(resultComponent));
-					}
+					if (resultComponent) this.#contentBox.addChild(resultComponent);
 				} catch (err) {
 					logger.warn("Tool renderer failed", { tool: this.#toolName, error: String(err) });
 					// Fall back to showing raw output on error
@@ -643,6 +686,11 @@ export class ToolExecutionComponent extends Container {
 					this.#contentBox.addChild(new Text(theme.fg("toolOutput", replaceTabs(output)), 0, 0));
 				}
 			}
+			// Custom tools that draw their own frame (task) render flush; plain
+			// extension renderers get the padded, state-tinted block back.
+			const customFramed = this.#contentBox.children.some(isFramedBlockComponent);
+			this.#contentBox.setPaddingX(customFramed ? 0 : 1);
+			this.#contentBox.setBgFn(customFramed ? undefined : stateBgFn);
 		} else if (this.#toolName in toolRenderers) {
 			// Built-in tools with renderers
 			const renderer = toolRenderers[this.#toolName];
@@ -672,19 +720,14 @@ export class ToolExecutionComponent extends Container {
 						this.#multiFileBoxes.push(spacer);
 						this.addChild(spacer);
 					}
-					const fileBgFn = fileResult.isError
-						? (text: string) => theme.bg("toolErrorBg", text)
-						: (text: string) => theme.bg("toolSuccessBg", text);
-					const fileBox = new Box(1, 1, fileBgFn);
+					const fileBox = new Box(0, 0);
 					try {
 						const resultComponent = renderer.renderResult(
 							{ content: [], details: fileResult, isError: fileResult.isError },
 							this.#renderState,
 							theme,
 						);
-						if (resultComponent) {
-							fileBox.addChild(ensureInvalidate(resultComponent));
-						}
+						if (resultComponent) fileBox.addChild(resultComponent);
 					} catch (err) {
 						logger.warn("Tool renderer failed", { tool: this.#toolName, error: String(err) });
 					}
@@ -701,7 +744,7 @@ export class ToolExecutionComponent extends Container {
 					const pendingSpacer = new Spacer(1);
 					this.#multiFileBoxes.push(pendingSpacer);
 					this.addChild(pendingSpacer);
-					const pendingBox = new Box(1, 1, (text: string) => theme.bg("toolPendingBg", text));
+					const pendingBox = new Box(0, 0);
 					const pendingText = renderStatusLine(
 						{
 							icon: "pending",
@@ -717,7 +760,7 @@ export class ToolExecutionComponent extends Container {
 			} else {
 				// Single-file or no result: standard rendering
 				// Inline renderers skip background styling
-				this.#contentBox.setBgFn(renderer.inline ? undefined : bgFn);
+				this.#contentBox.setBgFn(undefined);
 				this.#contentBox.clear();
 
 				const renderContext = this.#buildRenderContext();
@@ -728,9 +771,7 @@ export class ToolExecutionComponent extends Container {
 					// Render call component
 					try {
 						const callComponent = renderer.renderCall(this.#getCallArgsForRender(), this.#renderState, theme);
-						if (callComponent) {
-							this.#contentBox.addChild(ensureInvalidate(callComponent));
-						}
+						if (callComponent) this.#contentBox.addChild(callComponent);
 					} catch (err) {
 						logger.warn("Tool renderer failed", { tool: this.#toolName, error: String(err) });
 						// Fall back to default on error
@@ -751,9 +792,7 @@ export class ToolExecutionComponent extends Container {
 							theme,
 							this.#getCallArgsForRender(),
 						);
-						if (resultComponent) {
-							this.#contentBox.addChild(ensureInvalidate(resultComponent));
-						}
+						if (resultComponent) this.#contentBox.addChild(resultComponent);
 					} catch (err) {
 						logger.warn("Tool renderer failed", { tool: this.#toolName, error: String(err) });
 						// Fall back to showing raw output on error
@@ -766,7 +805,7 @@ export class ToolExecutionComponent extends Container {
 			}
 		} else {
 			// Other built-in tools: use Text directly with caching
-			this.#contentText.setCustomBgFn(bgFn);
+			this.#contentText.setCustomBgFn(stateBgFn);
 			this.#contentText.setText(this.#formatToolExecution());
 		}
 
diff --git a/packages/coding-agent/src/modes/components/transcript-container.ts b/packages/coding-agent/src/modes/components/transcript-container.ts
index b33a818bd..c3d7cf545 100644
--- a/packages/coding-agent/src/modes/components/transcript-container.ts
+++ b/packages/coding-agent/src/modes/components/transcript-container.ts
@@ -39,6 +39,26 @@ function isBlockAppendOnly(child: Component): boolean {
 	return fn ? fn.call(child) : false;
 }
 
+// A "plain blank" row is empty or whitespace-only with no ANSI bytes. It marks
+// separation padding (a `Spacer`, or a no-background `paddingY` row) as opposed
+// to a background-colored padding row, whose escape sequences contain `\S` and
+// are therefore preserved as part of a block's visual design.
+const NON_WHITESPACE = /\S/;
+function isPlainBlank(line: string): boolean {
+	return !NON_WHITESPACE.test(line);
+}
+
+// Strip leading/trailing plain-blank rows so each block contributes only its
+// visible body; the container owns the gaps between blocks. Returns the input
+// array unchanged when there is nothing to trim (no allocation on the hot path).
+function stripPlainBlankEdges(lines: string[]): string[] {
+	let start = 0;
+	let end = lines.length;
+	while (start < end && isPlainBlank(lines[start]!)) start++;
+	while (end > start && isPlainBlank(lines[end - 1]!)) end--;
+	return start === 0 && end === lines.length ? lines : lines.slice(start, end);
+}
+
 /**
  * Transcript container that freezes the rendered output of every block except
  * the bottom-most (live) one on terminals where committed native scrollback is
@@ -118,9 +138,13 @@ export class TranscriptContainer extends Container implements NativeScrollbackLi
 		width = Math.max(1, width);
 		this.#nativeScrollbackLiveRegionStart = undefined;
 		this.#nativeScrollbackCommitSafeEnd = undefined;
-		if (!TERMINAL.eagerEraseScrollbackRisk) return super.render(width);
 
+		// Freezing/snapshotting only applies on ED3-risk terminals; elsewhere every
+		// block renders live. Inter-block spacing applies on BOTH paths so the gap
+		// between blocks is identical regardless of terminal.
+		const risk = TERMINAL.eagerEraseScrollbackRisk;
 		const count = this.children.length;
+
 		// The live region spans from the earliest still-mutating block through the
 		// bottom. A block that has not finalized must stay repaintable: out-of-band
 		// inserts (TTSR/todo cards) can append a finalized block *below* a tool that
@@ -137,45 +161,81 @@ export class TranscriptContainer extends Container implements NativeScrollbackLi
 		// recompute them so they freeze at their final content. Everything below
 		// the lower of the two cutoffs was already frozen last frame and replays.
 		const replayCutoff = Math.min(liveStartIndex, this.#prevLiveStartIndex);
-		this.#prevLiveStartIndex = liveStartIndex;
+		if (risk) this.#prevLiveStartIndex = liveStartIndex;
 
 		const lines: string[] = [];
 		// Tracks whether we are still inside the leading run of append-only live
-		// blocks. The first non-append-only live block (or a finalized block below
-		// the live region's start, which cannot happen for a leading run) closes it.
+		// blocks. The first non-append-only live block closes it.
 		let commitSafeOpen = true;
+		// The live-region start is recorded at the first visible row at/after the
+		// cutoff; empty leading blocks (or a separator) must not claim it early.
+		let liveRecorded = false;
 		for (let i = 0; i < count; i++) {
 			const child = this.children[i]! as Component & SnapshotCarrier;
-			if (i >= liveStartIndex) {
-				if (i === liveStartIndex) this.#nativeScrollbackLiveRegionStart = lines.length;
-			} else {
+
+			// Resolve this child's contribution — its visible body with plain-blank
+			// top/bottom edges stripped (the container owns inter-block gaps). On
+			// ED3-risk terminals a frozen, scrolled-off block replays its snapshot
+			// instead of recomputing; a stale generation (post-thaw) or width
+			// mismatch (resize) recomputes, as does a block still live last frame.
+			let contribution: string[] | undefined;
+			if (risk && i < liveStartIndex && i < replayCutoff) {
 				const snapshot = child[kSnapshot];
-				// Replay a frozen block's last live render. A stale generation
-				// (post-thaw) or width mismatch (resize, explicit rebuild) recomputes
-				// instead, as does a block that was still live last frame (i >= cutoff).
-				if (i < replayCutoff && snapshot && snapshot.generation === this.#generation && snapshot.width === width) {
-					lines.push(...snapshot.lines);
-					continue;
+				if (snapshot && snapshot.generation === this.#generation && snapshot.width === width) {
+					contribution = snapshot.lines;
 				}
 			}
-			const rendered = child.render(width);
+			if (contribution === undefined) {
+				const rendered = child.render(width);
+				contribution = stripPlainBlankEdges(rendered);
+				// Cache every block's latest contribution. While a block is in the
+				// live region this keeps its snapshot current; on the frame it crosses
+				// out, the recompute above refreshes it before it freezes.
+				if (risk) child[kSnapshot] = { width, lines: contribution, generation: this.#generation };
+			}
+
+			// Empty (or stripped-to-nothing) children contribute nothing and never
+			// affect spacing or the live-region offsets.
+			if (contribution.length === 0) continue;
+
+			// Every block is separated from preceding visible content by exactly one
+			// blank row — skipped when it opens the transcript or the prior row is
+			// already a plain blank (a fragment's own trailing pad), never doubling.
+			const sep = lines.length > 0 && !isPlainBlank(lines[lines.length - 1]!) ? 1 : 0;
+
+			// The separator before the first live block stays in the committed prefix
+			// (it is deterministic and never changes once the prior block is frozen),
+			// so the live region begins at the block's first content row.
+			if (risk && !liveRecorded && i >= liveStartIndex) {
+				this.#nativeScrollbackLiveRegionStart = lines.length + sep;
+				liveRecorded = true;
+			}
+
+			if (sep) lines.push("");
+			for (let j = 0; j < contribution.length; j++) lines.push(contribution[j]!);
+
 			// Extend the commit-safe boundary through each leading append-only live
-			// block. `lines.length` here is this block's start offset; the boundary
-			// runs to the end of its rendered rows. The first volatile live block
-			// closes the run so its mutable rows stay deferred.
-			if (i >= liveStartIndex && commitSafeOpen) {
+			// block. The first volatile live block closes the run so its mutable
+			// rows stay deferred.
+			if (risk && i >= liveStartIndex && commitSafeOpen) {
 				if (isBlockAppendOnly(child)) {
-					this.#nativeScrollbackCommitSafeEnd = lines.length + rendered.length;
+					this.#nativeScrollbackCommitSafeEnd = lines.length;
 				} else {
 					commitSafeOpen = false;
 				}
 			}
-			// Cache every block's latest render. While a block is in the live region
-			// this keeps its snapshot current; on the frame it crosses out, the
-			// recompute above refreshes it to the final state before it freezes.
-			child[kSnapshot] = { width, lines: rendered, generation: this.#generation };
-			lines.push(...rendered);
 		}
 		return lines;
 	}
 }
+
+/**
+ * Groups a run of sibling rows (an IRC card's header + body, a file-mention
+ * list, a bordered command/version panel) into a single transcript child so the
+ * container spaces it as one block — one blank line above, none injected between
+ * its rows. Without this wrapper the rows would be top-level children and the
+ * container would put a blank line between each (and inside any border box).
+ * It is a plain {@link Container}; the named subclass documents intent and makes
+ * every manual block grouping greppable.
+ */
+export class TranscriptBlock extends Container {}
diff --git a/packages/coding-agent/src/modes/components/tree-selector.ts b/packages/coding-agent/src/modes/components/tree-selector.ts
index c015f8ca4..362ad2308 100644
--- a/packages/coding-agent/src/modes/components/tree-selector.ts
+++ b/packages/coding-agent/src/modes/components/tree-selector.ts
@@ -6,6 +6,7 @@ import {
 	fuzzyMatch,
 	Input,
 	matchesKey,
+	ScrollView,
 	Spacer,
 	Text,
 	TruncatedText,
@@ -492,6 +493,10 @@ class TreeList implements Component {
 		const contentReserve = Math.max(MIN_CONTENT_COLS, Math.floor(width / 2));
 		const maxIndentLevels = Math.max(1, Math.floor((width - contentReserve - OVERHEAD_COLS) / 3));
 
+		const overflow = this.#filteredNodes.length > this.maxVisibleLines;
+		const rowWidth = Math.max(0, width - (overflow ? 1 : 0));
+		const rows: string[] = [];
+
 		for (let i = startIndex; i < endIndex; i++) {
 			const flatNode = this.#filteredNodes[i];
 			const entry = flatNode.node.entry;
@@ -560,15 +565,22 @@ class TreeList implements Component {
 			if (isSelected) {
 				line = theme.bg("selectedBg", line);
 			}
-			lines.push(truncateToWidth(line, width));
+			rows.push(truncateToWidth(line, rowWidth));
 		}
 
-		lines.push(
-			truncateToWidth(
-				theme.fg("muted", `  (${this.#selectedIndex + 1}/${this.#filteredNodes.length})${this.#getFilterLabel()}`),
-				width,
-			),
-		);
+		const sv = new ScrollView(rows, {
+			height: rows.length,
+			scrollbar: "auto",
+			totalRows: this.#filteredNodes.length,
+			theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+		});
+		sv.setScrollOffset(startIndex);
+		lines.push(...sv.render(width));
+
+		const filterLabel = this.#getFilterLabel();
+		if (filterLabel) {
+			lines.push(truncateToWidth(theme.fg("muted", `  ${filterLabel.trim()}`), width));
+		}
 
 		return lines;
 	}
diff --git a/packages/coding-agent/src/modes/components/user-message-selector.ts b/packages/coding-agent/src/modes/components/user-message-selector.ts
index d67aabe38..a1845eb7c 100644
--- a/packages/coding-agent/src/modes/components/user-message-selector.ts
+++ b/packages/coding-agent/src/modes/components/user-message-selector.ts
@@ -4,6 +4,7 @@ import {
 	extractPrintableText,
 	fuzzyFilter,
 	matchesKey,
+	ScrollView,
 	Spacer,
 	Text,
 	truncateToWidth,
@@ -48,14 +49,10 @@ class UserMessageList implements Component {
 		return this.#isSearchEnabled() || this.#searchQuery.length > 0;
 	}
 
-	#renderStatusLine(total: number): string {
-		const selectedCount = total === 0 ? 0 : this.#selectedIndex + 1;
-		const count =
-			this.#searchQuery.trim() && total !== this.messages.length
-				? `${selectedCount}/${total} of ${this.messages.length}`
-				: `${selectedCount}/${total}`;
-		const suffix = this.#searchQuery.trim() ? `  Search: ${this.#searchQuery}` : "  Type to search";
-		return theme.fg("muted", `  (${count})${suffix}`);
+	#renderStatusLine(_total: number): string {
+		const query = this.#searchQuery.trim();
+		const suffix = query ? `Search: ${this.#searchQuery}` : "Type to search";
+		return theme.fg("muted", `  ${suffix}`);
 	}
 
 	#setSearchQuery(query: string): void {
@@ -103,6 +100,9 @@ class UserMessageList implements Component {
 		const endIndex = Math.min(startIndex + this.#maxVisible, total);
 
 		// Render visible messages (2 lines per message + blank line)
+		const overflow = total > this.#maxVisible;
+		const rowWidth = Math.max(0, width - (overflow ? 1 : 0));
+		const messageLines: string[] = [];
 		for (let i = startIndex; i < endIndex; i++) {
 			const message = this.#filteredMessages[i];
 			if (!message) continue;
@@ -113,26 +113,37 @@ class UserMessageList implements Component {
 
 			// First line: cursor + message
 			const cursor = isSelected ? theme.fg("accent", "› ") : "  ";
-			const maxMsgWidth = width - 2; // Account for cursor (2 chars)
+			const maxMsgWidth = rowWidth - 2; // Account for cursor (2 chars)
 			const truncatedMsg = truncateToWidth(normalizedMessage, maxMsgWidth);
 			const messageLine = cursor + (isSelected ? theme.bold(truncatedMsg) : truncatedMsg);
 
-			lines.push(messageLine);
+			messageLines.push(messageLine);
 
 			// Second line: metadata (position in history)
 			const position = this.messages.indexOf(message) + 1;
 			const metadata = `  Message ${position} of ${this.messages.length}`;
 			const metadataLine = theme.fg("muted", metadata);
-			lines.push(metadataLine);
-			lines.push(""); // Blank line between messages
+			messageLines.push(metadataLine);
+			messageLines.push(""); // Blank line between messages
 		}
 
 		if (total === 0) {
 			lines.push(theme.fg("muted", "  No matching messages"));
+		} else {
+			const visibleCount = endIndex - startIndex;
+			const linesPerItem = visibleCount > 0 ? messageLines.length / visibleCount : 1;
+			const sv = new ScrollView(messageLines, {
+				height: messageLines.length,
+				scrollbar: "auto",
+				totalRows: Math.round(total * linesPerItem),
+				theme: { track: t => theme.fg("muted", t), thumb: t => theme.fg("accent", t) },
+			});
+			sv.setScrollOffset(Math.round(startIndex * linesPerItem));
+			lines.push(...sv.render(width));
 		}
 
-		// Add scroll/search indicator if needed
-		if (startIndex > 0 || endIndex < total || this.#shouldRenderSearchStatus()) {
+		// Add search indicator if needed
+		if (this.#shouldRenderSearchStatus()) {
 			lines.push(this.#renderStatusLine(total));
 		}
 
diff --git a/packages/coding-agent/src/modes/components/user-message.ts b/packages/coding-agent/src/modes/components/user-message.ts
index c03eb7e73..e393a4ef3 100644
--- a/packages/coding-agent/src/modes/components/user-message.ts
+++ b/packages/coding-agent/src/modes/components/user-message.ts
@@ -1,4 +1,4 @@
-import { Container, Markdown, Spacer } from "@oh-my-pi/pi-tui";
+import { Container, Markdown } from "@oh-my-pi/pi-tui";
 import { getMarkdownTheme, theme } from "../../modes/theme/theme";
 import { imageReferenceHyperlink, renderImageReferences } from "../image-references";
 import { highlightMagicKeywords } from "../magic-keywords";
@@ -30,7 +30,6 @@ export class UserMessageComponent extends Container {
 				renderText: baseText,
 				renderReference: (label, index) => imageReferenceHyperlink(label, index, imageLinks, imageLabel),
 			});
-		this.addChild(new Spacer(1));
 		this.addChild(
 			new Markdown(text, 1, 1, getMarkdownTheme(), {
 				bgColor,
diff --git a/packages/coding-agent/src/modes/controllers/command-controller-shared.ts b/packages/coding-agent/src/modes/controllers/command-controller-shared.ts
index ba119d512..2e3eb6471 100644
--- a/packages/coding-agent/src/modes/controllers/command-controller-shared.ts
+++ b/packages/coding-agent/src/modes/controllers/command-controller-shared.ts
@@ -7,10 +7,11 @@
  * wording, and add-flow logic stay in the per-controller files because they
  * diverge in workflow.
  */
-import { Spacer, Text } from "@oh-my-pi/pi-tui";
+import { Text } from "@oh-my-pi/pi-tui";
 import type { SourceMeta } from "../../capability/types";
 import { shortenPath } from "../../tools/render-utils";
 import { DynamicBorder } from "../components/dynamic-border";
+import { TranscriptBlock } from "../components/transcript-container";
 import { parseCommandArgs } from "../shared";
 import type { InteractiveModeContext } from "../types";
 
@@ -100,9 +101,9 @@ export function* groupBySource<T>(
  * container and request a render.
  */
 export function showCommandMessage(ctx: InteractiveModeContext, text: string): void {
-	ctx.chatContainer.addChild(new Spacer(1));
-	ctx.chatContainer.addChild(new DynamicBorder());
-	ctx.chatContainer.addChild(new Text(text, 1, 1));
-	ctx.chatContainer.addChild(new DynamicBorder());
-	ctx.ui.requestRender();
+	const block = new TranscriptBlock();
+	block.addChild(new DynamicBorder());
+	block.addChild(new Text(text, 1, 1));
+	block.addChild(new DynamicBorder());
+	ctx.present(block);
 }
diff --git a/packages/coding-agent/src/modes/controllers/command-controller.ts b/packages/coding-agent/src/modes/controllers/command-controller.ts
index 6cfce5c9e..59a4b89de 100644
--- a/packages/coding-agent/src/modes/controllers/command-controller.ts
+++ b/packages/coding-agent/src/modes/controllers/command-controller.ts
@@ -6,7 +6,6 @@ import {
 	getEnvApiKey,
 	getProviderDetails,
 	type ProviderDetails,
-	type ToolCall,
 	type UsageLimit,
 	type UsageReport,
 } from "@oh-my-pi/pi-ai";
@@ -14,7 +13,6 @@ import { Loader, Markdown, padding, Spacer, Text, visibleWidth } from "@oh-my-pi
 import { formatDuration, Snowflake } from "@oh-my-pi/pi-utils";
 import { $ } from "bun";
 import { shouldEnableAppendOnlyContext } from "../../config/append-only-context-mode";
-import { loadCustomShare } from "../../export/custom-share";
 import type { CompactOptions } from "../../extensibility/extensions/types";
 import {
 	diffMentalModelContent,
@@ -30,6 +28,7 @@ import { BashExecutionComponent } from "../../modes/components/bash-execution";
 import { BorderedLoader } from "../../modes/components/bordered-loader";
 import { DynamicBorder } from "../../modes/components/dynamic-border";
 import { EvalExecutionComponent } from "../../modes/components/eval-execution";
+import { TranscriptBlock } from "../../modes/components/transcript-container";
 import { getMarkdownTheme, getSymbolTheme, theme } from "../../modes/theme/theme";
 import type { InteractiveModeContext } from "../../modes/types";
 import { computeContextBreakdown, renderContextUsage } from "../../modes/utils/context-usage";
@@ -48,13 +47,13 @@ import { openPath } from "../../utils/open";
 import { setSessionTerminalTitle } from "../../utils/title-generator";
 
 function showMarkdownPanel(ctx: InteractiveModeContext, title: string, markdown: string): void {
-	ctx.chatContainer.addChild(new Spacer(1));
-	ctx.chatContainer.addChild(new DynamicBorder());
-	ctx.chatContainer.addChild(new Text(theme.bold(theme.fg("accent", title)), 1, 0));
-	ctx.chatContainer.addChild(new Spacer(1));
-	ctx.chatContainer.addChild(new Markdown(markdown.trim(), 1, 1, getMarkdownTheme()));
-	ctx.chatContainer.addChild(new DynamicBorder());
-	ctx.ui.requestRender();
+	const block = new TranscriptBlock();
+	block.addChild(new DynamicBorder());
+	block.addChild(new Text(theme.bold(theme.fg("accent", title)), 1, 0));
+	block.addChild(new Spacer(1));
+	block.addChild(new Markdown(markdown.trim(), 1, 1, getMarkdownTheme()));
+	block.addChild(new DynamicBorder());
+	ctx.present(block);
 }
 
 export class CommandController {
@@ -132,6 +131,7 @@ export class CommandController {
 		}
 
 		try {
+			const { loadCustomShare } = await import("../../export/custom-share");
 			const customShare = await loadCustomShare();
 			if (customShare) {
 				const loader = new BorderedLoader(this.ctx.ui, theme, "Sharing...");
@@ -239,121 +239,6 @@ export class CommandController {
 		}
 	}
 
-	handleCopyCommand(sub?: string) {
-		switch (sub) {
-			case "code":
-				return this.#copyCode();
-			case "all":
-				return this.#copyAllCode();
-			case "cmd":
-				return this.#copyLastCommand();
-			case "last":
-			case undefined:
-				return this.#copyLastMessage();
-			default:
-				this.ctx.showError(`Unknown subcommand: ${sub}. Use code, all, cmd, or last.`);
-		}
-	}
-
-	#copyLastMessage() {
-		const assistantText = this.ctx.session.getLastAssistantText();
-		if (assistantText) {
-			this.#doCopy(assistantText, "Copied last agent message to clipboard");
-			return;
-		}
-
-		if (!this.ctx.session.hasCopyCandidateAssistantMessage()) {
-			const handoffText = this.ctx.session.getLastVisibleHandoffText();
-			if (handoffText) {
-				this.#doCopy(handoffText, "Copied handoff context to clipboard");
-				return;
-			}
-		}
-
-		this.ctx.showError("No agent messages to copy yet.");
-	}
-
-	#copyCode() {
-		const text = this.ctx.session.getLastAssistantText();
-		if (!text) {
-			this.ctx.showError("No agent messages to copy yet.");
-			return;
-		}
-		const matches = [...text.matchAll(/^```[^\n]*\n([\s\S]*?)^```/gm)];
-		const lastMatch = matches.at(-1);
-		if (!lastMatch) {
-			this.ctx.showWarning("No code block found in the last agent message.");
-			return;
-		}
-		this.#doCopy(lastMatch[1].replace(/\n$/, ""), "Copied last code block to clipboard");
-	}
-
-	#copyAllCode() {
-		const text = this.ctx.session.getLastAssistantText();
-		if (!text) {
-			this.ctx.showError("No agent messages to copy yet.");
-			return;
-		}
-		const matches = [...text.matchAll(/^```[^\n]*\n([\s\S]*?)^```/gm)];
-		if (matches.length === 0) {
-			this.ctx.showWarning("No code blocks found in the last agent message.");
-			return;
-		}
-		const combined = matches.map(m => m[1].replace(/\n$/, "")).join("\n\n");
-		this.#doCopy(combined, `Copied ${matches.length} code block${matches.length > 1 ? "s" : ""} to clipboard`);
-	}
-
-	#extractEvalCode(args: unknown): string | undefined {
-		if (!args || typeof args !== "object") return undefined;
-		const cells = (args as { cells?: unknown }).cells;
-		if (!Array.isArray(cells)) return undefined;
-
-		const codeBlocks: string[] = [];
-		for (const cell of cells) {
-			if (!cell || typeof cell !== "object") continue;
-			const code = (cell as { code?: unknown }).code;
-			if (typeof code === "string" && code.length > 0) {
-				codeBlocks.push(code);
-			}
-		}
-
-		return codeBlocks.length > 0 ? codeBlocks.join("\n\n") : undefined;
-	}
-
-	#copyLastCommand() {
-		const messages = this.ctx.session.messages;
-		// Walk backwards to find the last bash/eval tool call
-		for (let i = messages.length - 1; i >= 0; i--) {
-			const msg = messages[i];
-			if (msg.role !== "assistant") continue;
-			const toolCalls = msg.content.filter((c): c is ToolCall => c.type === "toolCall");
-			for (let j = toolCalls.length - 1; j >= 0; j--) {
-				const tc = toolCalls[j];
-				if (tc.name === "bash" && typeof tc.arguments.command === "string") {
-					this.#doCopy(tc.arguments.command, "Copied last bash command to clipboard");
-					return;
-				}
-				if (tc.name === "eval") {
-					const code = this.#extractEvalCode(tc.arguments);
-					if (code) {
-						this.#doCopy(code, "Copied last eval code to clipboard");
-						return;
-					}
-				}
-			}
-		}
-		this.ctx.showWarning("No bash or eval command found in the conversation.");
-	}
-
-	#doCopy(content: string, label: string) {
-		try {
-			copyToClipboard(content);
-			this.ctx.showStatus(label);
-		} catch (error) {
-			this.ctx.showError(error instanceof Error ? error.message : String(error));
-		}
-	}
-
 	async handleSessionCommand(): Promise<void> {
 		const stats = this.ctx.session.getSessionStats();
 		const premiumRequests =
@@ -450,9 +335,7 @@ export class CommandController {
 			}
 		}
 
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Text(info, 1, 0));
-		this.ctx.ui.requestRender();
+		this.ctx.present([new Spacer(1), new Text(info, 1, 0)]);
 	}
 
 	async handleJobsCommand(): Promise<void> {
@@ -469,9 +352,7 @@ export class CommandController {
 
 		if (snapshot.running.length === 0 && snapshot.recent.length === 0) {
 			info += `\n${theme.fg("dim", "No async jobs yet.")}\n`;
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(new Text(info, 1, 0));
-			this.ctx.ui.requestRender();
+			this.ctx.present([new Spacer(1), new Text(info, 1, 0)]);
 			return;
 		}
 
@@ -491,9 +372,7 @@ export class CommandController {
 			}
 		}
 
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Text(info.trimEnd(), 1, 0));
-		this.ctx.ui.requestRender();
+		this.ctx.present([new Spacer(1), new Text(info.trimEnd(), 1, 0)]);
 	}
 
 	async handleUsageCommand(reports?: UsageReport[] | null): Promise<void> {
@@ -519,9 +398,7 @@ export class CommandController {
 
 		const availableWidth = Math.max(40, (this.ctx.ui.terminal.columns ?? 100) - 2);
 		const output = renderUsageReports(usageReports, theme, Date.now(), availableWidth);
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Text(output, 1, 0));
-		this.ctx.ui.requestRender();
+		this.ctx.present([new Spacer(1), new Text(output, 1, 0)]);
 	}
 
 	async handleChangelogCommand(showFull = false): Promise<void> {
@@ -542,13 +419,13 @@ export class CommandController {
 			? ""
 			: `\n\n${theme.fg("dim", "Use")} ${theme.bold("/changelog full")} ${theme.fg("dim", "to view the complete changelog.")}`;
 
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new DynamicBorder());
-		this.ctx.chatContainer.addChild(new Text(theme.bold(theme.fg("accent", title)), 1, 0));
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Markdown(changelogMarkdown + hint, 1, 1, getMarkdownTheme()));
-		this.ctx.chatContainer.addChild(new DynamicBorder());
-		this.ctx.ui.requestRender();
+		const block = new TranscriptBlock();
+		block.addChild(new DynamicBorder());
+		block.addChild(new Text(theme.bold(theme.fg("accent", title)), 1, 0));
+		block.addChild(new Spacer(1));
+		block.addChild(new Markdown(changelogMarkdown + hint, 1, 1, getMarkdownTheme()));
+		block.addChild(new DynamicBorder());
+		this.ctx.present(block);
 	}
 
 	handleHotkeysCommand(): void {
@@ -568,20 +445,20 @@ export class CommandController {
 			return;
 		}
 		const output = renderContextUsage(breakdown, theme);
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new DynamicBorder());
-		this.ctx.chatContainer.addChild(new Text(theme.bold(theme.fg("accent", "Context Usage")), 1, 0));
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Text(output, 1, 0));
-		this.ctx.chatContainer.addChild(new DynamicBorder());
-		this.ctx.ui.requestRender();
+		const block = new TranscriptBlock();
+		block.addChild(new DynamicBorder());
+		block.addChild(new Text(theme.bold(theme.fg("accent", "Context Usage")), 1, 0));
+		block.addChild(new Spacer(1));
+		block.addChild(new Text(output, 1, 0));
+		block.addChild(new DynamicBorder());
+		this.ctx.present(block);
 	}
 
 	async handleMemoryCommand(text: string): Promise<void> {
 		const argumentText = text.slice(7).trim();
 		const action = argumentText.split(/\s+/, 1)[0]?.toLowerCase() || "view";
 		const agentDir = this.ctx.settings.getAgentDir();
-		const backend = resolveMemoryBackend(this.ctx.settings);
+		const backend = await resolveMemoryBackend(this.ctx.settings);
 
 		if (action === "view") {
 			const payload = await backend.buildDeveloperInstructions(agentDir, this.ctx.settings, this.ctx.session);
@@ -589,13 +466,13 @@ export class CommandController {
 				this.ctx.showWarning("Memory payload is empty (memory backend off, disabled, or no memory available).");
 				return;
 			}
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(new DynamicBorder());
-			this.ctx.chatContainer.addChild(new Text(theme.bold(theme.fg("accent", "Memory Injection Payload")), 1, 0));
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(new Markdown(payload, 1, 1, getMarkdownTheme()));
-			this.ctx.chatContainer.addChild(new DynamicBorder());
-			this.ctx.ui.requestRender();
+			const block = new TranscriptBlock();
+			block.addChild(new DynamicBorder());
+			block.addChild(new Text(theme.bold(theme.fg("accent", "Memory Injection Payload")), 1, 0));
+			block.addChild(new Spacer(1));
+			block.addChild(new Markdown(payload, 1, 1, getMarkdownTheme()));
+			block.addChild(new DynamicBorder());
+			this.ctx.present(block);
 			return;
 		}
 
@@ -916,8 +793,7 @@ export class CommandController {
 		this.ctx.streamingMessage = undefined;
 		this.ctx.pendingTools.clear();
 
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Text(`${theme.fg("accent", `${theme.status.success} ${label}`)}`, 1, 1));
+		this.ctx.present([new Spacer(1), new Text(`${theme.fg("accent", `${theme.status.success} ${label}`)}`, 1, 1)]);
 		await this.ctx.reloadTodos();
 		this.ctx.ui.requestRender(true, { clearScrollback: true });
 	}
@@ -926,6 +802,18 @@ export class CommandController {
 		await this.#runNewSessionFlow();
 	}
 
+	async handleFreshCommand(): Promise<void> {
+		const result = this.ctx.session.freshSession();
+		if (!result) {
+			this.ctx.showWarning("Wait for the current response to finish or abort it before refreshing provider state.");
+			return;
+		}
+		const stateLabel = result.closedProviderSessions === 1 ? "provider state" : "provider states";
+		this.ctx.statusLine.invalidate();
+		this.ctx.updateEditorTopBorder();
+		this.ctx.showStatus(`Fresh provider session started (${result.closedProviderSessions} ${stateLabel} pruned).`);
+	}
+
 	async handleDropCommand(): Promise<void> {
 		if (!this.ctx.sessionManager.getSessionFile()) {
 			this.ctx.showError("Nothing to drop (in-memory session)");
@@ -956,11 +844,10 @@ export class CommandController {
 
 		const sessionFile = this.ctx.session.sessionFile;
 		const shortPath = sessionFile ? sessionFile.split("/").pop() : "new session";
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(
+		this.ctx.present([
+			new Spacer(1),
 			new Text(`${theme.fg("accent", `${theme.status.success} Session forked to ${shortPath}`)}`, 1, 1),
-		);
-		this.ctx.ui.requestRender();
+		]);
 	}
 
 	async handleMoveCommand(targetPath: string): Promise<void> {
@@ -994,11 +881,10 @@ export class CommandController {
 			await this.ctx.sessionManager.moveTo(resolvedPath);
 			await this.ctx.applyCwdChange(resolvedPath);
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
+			this.ctx.present([
+				new Spacer(1),
 				new Text(`${theme.fg("accent", `${theme.status.success} Session moved to ${resolvedPath}`)}`, 1, 1),
-			);
-			this.ctx.ui.requestRender();
+			]);
 		} catch (err) {
 			this.ctx.showError(`Move failed: ${err instanceof Error ? err.message : String(err)}`);
 		}
@@ -1029,7 +915,7 @@ export class CommandController {
 			this.ctx.pendingMessagesContainer.addChild(this.ctx.bashComponent);
 			this.ctx.pendingBashComponents.push(this.ctx.bashComponent);
 		} else {
-			this.ctx.chatContainer.addChild(this.ctx.bashComponent);
+			this.ctx.present(this.ctx.bashComponent);
 		}
 		this.ctx.ui.requestRender();
 
@@ -1070,7 +956,7 @@ export class CommandController {
 			this.ctx.pendingMessagesContainer.addChild(this.ctx.pythonComponent);
 			this.ctx.pendingPythonComponents.push(this.ctx.pythonComponent);
 		} else {
-			this.ctx.chatContainer.addChild(this.ctx.pythonComponent);
+			this.ctx.present(this.ctx.pythonComponent);
 		}
 		this.ctx.ui.requestRender();
 
@@ -1259,10 +1145,10 @@ export class CommandController {
 			this.ctx.updateEditorBorderColor();
 			await this.ctx.reloadTodos();
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
+			this.ctx.present([
+				new Spacer(1),
 				new Text(`${theme.fg("accent", `${theme.status.success} New session started with handoff context`)}`, 1, 1),
-			);
+			]);
 			if (result.savedPath) {
 				this.ctx.showStatus(`Handoff document saved to: ${result.savedPath}`);
 			}
@@ -1388,6 +1274,8 @@ function formatAccountLabel(limit: UsageLimit, report: UsageReport, index: numbe
 	if (email) return email;
 	const accountId = (report.metadata?.accountId as string | undefined) ?? limit.scope.accountId;
 	if (accountId) return accountId;
+	const projectId = (report.metadata?.projectId as string | undefined) ?? limit.scope.projectId;
+	if (projectId) return projectId;
 	return `account ${index + 1}`;
 }
 
@@ -1396,6 +1284,8 @@ function formatUnlimitedReportLabel(report: UsageReport, index: number): string
 	if (email) return email;
 	const accountId = report.metadata?.accountId as string | undefined;
 	if (accountId) return accountId;
+	const projectId = report.metadata?.projectId as string | undefined;
+	if (projectId) return projectId;
 	return `account ${index + 1}`;
 }
 
@@ -1481,6 +1371,13 @@ function formatAggregateAmount(limits: UsageLimit[]): string {
 		return `${formatNumber(remainingPct)}% free`;
 	}
 
+	// Count unique accounts from limit scopes — not limits.length.
+	const uniqueAccountIds = new Set(
+		limits.map(limit => limit.scope.accountId).filter((id): id is string => typeof id === "string" && id.length > 0),
+	);
+	if (uniqueAccountIds.size > 0) return `${uniqueAccountIds.size} ${uniqueAccountIds.size === 1 ? "acct" : "accts"}`;
+	// No account IDs available — keep the pre-existing fallback so providers
+	// that don't populate scope.accountId still show a summary.
 	return `${limits.length} accts`;
 }
 
diff --git a/packages/coding-agent/src/modes/controllers/event-controller.ts b/packages/coding-agent/src/modes/controllers/event-controller.ts
index 2bc4734a1..d1f1e8e28 100644
--- a/packages/coding-agent/src/modes/controllers/event-controller.ts
+++ b/packages/coding-agent/src/modes/controllers/event-controller.ts
@@ -1,7 +1,7 @@
 import { INTENT_FIELD } from "@oh-my-pi/pi-agent-core";
 import { calculatePromptTokens } from "@oh-my-pi/pi-agent-core/compaction/compaction";
 import type { AssistantMessage, ImageContent } from "@oh-my-pi/pi-ai";
-import { type Component, Loader, TERMINAL, Text } from "@oh-my-pi/pi-tui";
+import { type Component, Loader, TERMINAL } from "@oh-my-pi/pi-tui";
 import { settings } from "../../config/settings";
 import { getFileSnapshotStore } from "../../edit/file-snapshot-store";
 import { AssistantMessageComponent } from "../../modes/components/assistant-message";
@@ -17,9 +17,10 @@ import { getSymbolTheme, theme } from "../../modes/theme/theme";
 import type { InteractiveModeContext, TodoPhase } from "../../modes/types";
 import type { PlanApprovalDetails } from "../../plan-mode/approved-plan";
 import type { AgentSessionEvent } from "../../session/agent-session";
-import { isSilentAbort, readPendingDisplayTag } from "../../session/messages";
+import { isSilentAbort, readPendingDisplayTag, resolveAbortLabel } from "../../session/messages";
 import type { ResolveToolDetails } from "../../tools/resolve";
 import { interruptHint } from "../shared";
+import { StreamingRevealController } from "./streaming-reveal";
 
 type AgentSessionEventKind = AgentSessionEvent["type"];
 
@@ -44,19 +45,37 @@ type AgentSessionEventHandlers = {
 
 export class EventController {
 	#lastReadGroup: ReadToolGroupComponent | undefined = undefined;
-	#lastThinkingCount = 0;
+	// Count of visible assistant content blocks (rendered non-empty text/thinking)
+	// already seen in the current streaming message. A newly appearing one breaks
+	// the read run: the rendered reasoning/answer is a visual separator, so reads
+	// after it start a fresh group. Empty/absent thinking — common when a model
+	// emits one read per completion — does not break it, so a run of consecutive
+	// reads collapses into one group even across completion boundaries.
+	#lastVisibleBlockCount = 0;
 	#renderedCustomMessages = new Set<string>();
 	#lastIntent: string | undefined = undefined;
 	#backgroundToolCallIds = new Set<string>();
 	#assistantMessageStreaming = false;
+	#agentTurnActive = false;
 	#readToolCallArgs = new Map<string, Record<string, unknown>>();
 	#readToolCallAssistantComponents = new Map<string, AssistantMessageComponent>();
 	#lastAssistantComponent: AssistantMessageComponent | undefined = undefined;
+	// Assistant component whose turn-ending error is currently mirrored in the
+	// pinned banner. Its inline `Error: …` line is suppressed while pinned and
+	// restored when the banner clears at the next `agent_start` (see
+	// #handleMessageEnd / #handleAgentStart).
+	#pinnedErrorComponent: AssistantMessageComponent | undefined = undefined;
 	#idleCompactionTimer?: NodeJS.Timeout;
 	#ircExpiryTimers = new Map<string, NodeJS.Timeout>();
+	#streamingReveal: StreamingRevealController;
 	#handlers: AgentSessionEventHandlers;
 
 	constructor(private ctx: InteractiveModeContext) {
+		this.#streamingReveal = new StreamingRevealController({
+			getSmoothStreaming: () => this.ctx.settings.get("display.smoothStreaming"),
+			getHideThinkingBlock: () => this.ctx.hideThinkingBlock,
+			requestRender: () => this.ctx.ui.requestRender(),
+		});
 		this.#handlers = {
 			agent_start: e => this.#handleAgentStart(e),
 			agent_end: e => this.#handleAgentEnd(e),
@@ -89,6 +108,7 @@ export class EventController {
 	}
 
 	dispose(): void {
+		this.#streamingReveal.stop();
 		this.#cancelIdleCompaction();
 		for (const timer of this.#ircExpiryTimers.values()) {
 			clearTimeout(timer);
@@ -97,12 +117,12 @@ export class EventController {
 	}
 
 	#resetReadGroup(): void {
+		this.#lastReadGroup?.finalize();
 		this.#lastReadGroup = undefined;
 	}
 
 	#getReadGroup(): ReadToolGroupComponent {
 		if (!this.#lastReadGroup) {
-			this.ctx.chatContainer.addChild(new Text("", 0, 0));
 			const group = new ReadToolGroupComponent({
 				showContentPreview: this.ctx.settings.get("read.toolResultPreview"),
 			});
@@ -172,21 +192,21 @@ export class EventController {
 
 		const run = this.#handlers[event.type] as (e: AgentSessionEvent) => Promise<void>;
 		await run(event);
-		// While assistant text or a foreground tool is streaming, rows above the
-		// viewport can re-layout after they have already entered native scrollback
-		// (Markdown fences, wrapping, previews). Let the TUI rebuild history on
-		// those offscreen edits instead of deferring, which otherwise leaves stale
-		// tail rows duplicated above the live viewport.
-		// Background-running tools are excluded so late async updates outside the
-		// active foreground stream keep the no-yank deferral; agent_start resets
-		// the mode at every turn boundary.
+		// While an assistant turn is active, visible status chrome and foreground
+		// transcript blocks can re-render after rows have entered native scrollback
+		// (idle Working loader, Markdown fences, wrapping, tool previews). Let the
+		// TUI use its foreground live-region path instead of idle deferral, which
+		// can otherwise leave the loader/status frame frozen until the next input.
+		// Background-running tools after the turn ends are excluded so late async
+		// updates keep the no-yank deferral; agent_start/agent_end bracket the
+		// foreground turn.
 		if (STREAM_RENDER_MODE_EVENTS[event.type]) {
 			this.#refreshToolRenderMode();
 		}
 	}
 
 	#refreshToolRenderMode(): void {
-		let foregroundToolActive = this.#assistantMessageStreaming;
+		let foregroundToolActive = this.#agentTurnActive || this.#assistantMessageStreaming;
 		if (!foregroundToolActive) {
 			for (const toolCallId of this.ctx.pendingTools.keys()) {
 				if (!this.#backgroundToolCallIds.has(toolCallId)) {
@@ -199,11 +219,17 @@ export class EventController {
 	}
 
 	async #handleAgentStart(_event: Extract<AgentSessionEvent, { type: "agent_start" }>): Promise<void> {
+		this.#agentTurnActive = true;
 		this.#lastIntent = undefined;
 		this.#readToolCallArgs.clear();
 		this.#readToolCallAssistantComponents.clear();
+		this.#resetReadGroup();
 		this.#assistantMessageStreaming = false;
 		this.#lastAssistantComponent = undefined;
+		// Restore the previous turn's inline error in the transcript before dropping
+		// the banner, so the error stays in history once the banner is gone.
+		this.#pinnedErrorComponent?.setErrorPinned(false);
+		this.#pinnedErrorComponent = undefined;
 		this.ctx.clearPinnedError();
 		if (this.ctx.retryEscapeHandler) {
 			this.ctx.editor.onEscape = this.ctx.retryEscapeHandler;
@@ -215,6 +241,7 @@ export class EventController {
 			this.ctx.statusContainer.clear();
 		}
 		this.#cancelIdleCompaction();
+		this.#refreshToolRenderMode();
 		this.ctx.ensureLoadingAnimation();
 		this.ctx.ui.requestRender();
 	}
@@ -287,9 +314,8 @@ export class EventController {
 			this.ctx.addMessageToChat(event.message);
 			this.ctx.ui.requestRender();
 		} else if (event.message.role === "assistant") {
-			this.#lastThinkingCount = 0;
 			this.#assistantMessageStreaming = true;
-			this.#resetReadGroup();
+			this.#lastVisibleBlockCount = 0;
 			this.ctx.streamingComponent = new AssistantMessageComponent(
 				undefined,
 				this.ctx.hideThinkingBlock,
@@ -299,7 +325,7 @@ export class EventController {
 			);
 			this.ctx.streamingMessage = event.message;
 			this.ctx.chatContainer.addChild(this.ctx.streamingComponent);
-			this.ctx.streamingComponent.updateContent(this.ctx.streamingMessage);
+			this.#streamingReveal.begin(this.ctx.streamingComponent, this.ctx.streamingMessage);
 			this.ctx.ui.requestRender();
 		}
 	}
@@ -343,16 +369,17 @@ export class EventController {
 	async #handleMessageUpdate(event: Extract<AgentSessionEvent, { type: "message_update" }>): Promise<void> {
 		if (this.ctx.streamingComponent && event.message.role === "assistant") {
 			this.ctx.streamingMessage = event.message;
-			this.ctx.streamingComponent.updateContent(this.ctx.streamingMessage);
+			this.#streamingReveal.setTarget(this.ctx.streamingMessage);
 
-			const thinkingCount = this.ctx.streamingMessage.content.filter(
-				content => content.type === "thinking" && content.thinking.trim(),
+			const visibleBlockCount = this.ctx.streamingMessage.content.filter(
+				content =>
+					(content.type === "text" && content.text.trim().length > 0) ||
+					(content.type === "thinking" && content.thinking.trim().length > 0),
 			).length;
-			if (thinkingCount > this.#lastThinkingCount) {
+			if (visibleBlockCount > this.#lastVisibleBlockCount) {
 				this.#resetReadGroup();
-				this.#lastThinkingCount = thinkingCount;
+				this.#lastVisibleBlockCount = visibleBlockCount;
 			}
-
 			for (const content of this.ctx.streamingMessage.content) {
 				if (content.type !== "toolCall") continue;
 				if (content.name === "read") {
@@ -385,7 +412,6 @@ export class EventController {
 						: content.arguments;
 				if (!this.ctx.pendingTools.has(content.id)) {
 					this.#resetReadGroup();
-					this.ctx.chatContainer.addChild(new Text("", 0, 0));
 					const tool = this.ctx.session.getToolByName(content.name);
 					const component = new ToolExecutionComponent(
 						content.name,
@@ -444,21 +470,20 @@ export class EventController {
 		}
 		if (this.ctx.streamingComponent && event.message.role === "assistant") {
 			this.ctx.streamingMessage = event.message;
+			this.#streamingReveal.stop();
 			let errorMessage: string | undefined;
 			const aborted = this.ctx.streamingMessage.stopReason === "aborted";
 			const silentlyAborted = aborted && isSilentAbort(this.ctx.streamingMessage.errorMessage);
 			const ttsrSilenced = aborted && this.ctx.session.isTtsrAbortPending;
 			if (aborted && !silentlyAborted && !ttsrSilenced) {
-				// Real user-cancel / network / provider abort: surface the standard
-				// operator-facing label. AgentSession.#handleAgentEvent already stamped
-				// SILENT_ABORT_MARKER for the plan-compact transition before this
-				// controller ran, so reaching this branch implies the abort was NOT a
-				// silent internal transition.
-				const retryAttempt = this.ctx.session.retryAttempt;
-				errorMessage =
-					retryAttempt > 0
-						? `Aborted after ${retryAttempt} retry attempt${retryAttempt > 1 ? "s" : ""}`
-						: "Operation aborted";
+				// Resolve the operator-facing label: a user-interrupt (Esc) abort
+				// carries USER_INTERRUPT_LABEL on errorMessage (threaded through the
+				// AbortController), which is preserved verbatim; any other abort with
+				// no threaded reason falls back to the retry-aware generic label.
+				// AgentSession.#handleAgentEvent already stamped SILENT_ABORT_MARKER for
+				// the plan-compact transition before this controller ran, so reaching
+				// this branch implies the abort was NOT a silent internal transition.
+				errorMessage = resolveAbortLabel(this.ctx.streamingMessage.errorMessage, this.ctx.session.retryAttempt);
 				this.ctx.streamingMessage.errorMessage = errorMessage;
 			}
 			if (silentlyAborted || ttsrSilenced) {
@@ -493,12 +518,15 @@ export class EventController {
 			this.ctx.streamingMessage = undefined;
 			// Pin a turn-ending provider error (e.g. Anthropic content-filter block)
 			// above the editor so it survives transcript scroll. Cleared at the next
-			// turn's agent_start.
+			// turn's agent_start. Suppress the transcript's inline `Error: …` line for
+			// the same message while pinned so the error isn't rendered twice.
 			if (
 				event.message.stopReason === "error" &&
 				event.message.errorMessage &&
 				!isSilentAbort(event.message.errorMessage)
 			) {
+				this.#lastAssistantComponent?.setErrorPinned(true);
+				this.#pinnedErrorComponent = this.#lastAssistantComponent;
 				this.ctx.showPinnedError(event.message.errorMessage);
 			}
 			this.ctx.statusLine.invalidate();
@@ -646,7 +674,9 @@ export class EventController {
 		}
 	}
 	async #handleAgentEnd(_event: Extract<AgentSessionEvent, { type: "agent_end" }>): Promise<void> {
+		this.#agentTurnActive = false;
 		this.#assistantMessageStreaming = false;
+		this.#streamingReveal.stop();
 		if (this.ctx.loadingAnimation) {
 			this.ctx.loadingAnimation.stop();
 			this.ctx.loadingAnimation = undefined;
@@ -673,6 +703,7 @@ export class EventController {
 		);
 		this.#readToolCallArgs.clear();
 		this.#readToolCallAssistantComponents.clear();
+		this.#resetReadGroup();
 		this.#lastAssistantComponent = undefined;
 		this.ctx.ui.requestRender();
 		this.#scheduleIdleCompaction();
@@ -816,14 +847,12 @@ export class EventController {
 	async #handleTtsrTriggered(event: Extract<AgentSessionEvent, { type: "ttsr_triggered" }>): Promise<void> {
 		const component = new TtsrNotificationComponent(event.rules);
 		component.setExpanded(this.ctx.toolOutputExpanded);
-		this.ctx.chatContainer.addChild(component);
-		this.ctx.ui.requestRender();
+		this.ctx.present(component);
 	}
 
 	async #handleTodoReminder(event: Extract<AgentSessionEvent, { type: "todo_reminder" }>): Promise<void> {
 		const component = new TodoReminderComponent(event.todos, event.attempt, event.maxAttempts);
-		this.ctx.chatContainer.addChild(component);
-		this.ctx.ui.requestRender();
+		this.ctx.present(component);
 	}
 
 	async #handleTodoAutoClear(_event: Extract<AgentSessionEvent, { type: "todo_auto_clear" }>): Promise<void> {
@@ -876,7 +905,6 @@ export class EventController {
 	}
 
 	sendCompletionNotification(): void {
-		if (this.ctx.isBackgrounded === false) return;
 		const notify = settings.get("completion.notify");
 		if (notify === "off") return;
 
@@ -895,15 +923,4 @@ export class EventController {
 			actions: "focus",
 		});
 	}
-
-	async handleBackgroundEvent(event: AgentSessionEvent): Promise<void> {
-		if (event.type !== "agent_end") {
-			return;
-		}
-		if (this.ctx.session.queuedMessageCount > 0 || this.ctx.session.isStreaming) {
-			return;
-		}
-		this.sendCompletionNotification();
-		await this.ctx.shutdown();
-	}
 }
diff --git a/packages/coding-agent/src/modes/controllers/extension-ui-controller.ts b/packages/coding-agent/src/modes/controllers/extension-ui-controller.ts
index e2b994fce..31d6e7cc0 100644
--- a/packages/coding-agent/src/modes/controllers/extension-ui-controller.ts
+++ b/packages/coding-agent/src/modes/controllers/extension-ui-controller.ts
@@ -1,6 +1,5 @@
 import type { Component, OverlayHandle, TUI } from "@oh-my-pi/pi-tui";
 import { Container, Spacer, Text } from "@oh-my-pi/pi-tui";
-import { logger } from "@oh-my-pi/pi-utils";
 import { KeybindingsManager } from "../../config/keybindings";
 import type {
 	CompactOptions,
@@ -176,10 +175,10 @@ export class ExtensionUiController {
 				this.ctx.streamingMessage = undefined;
 				this.ctx.pendingTools.clear();
 
-				this.ctx.chatContainer.addChild(new Spacer(1));
-				this.ctx.chatContainer.addChild(
+				this.ctx.present([
+					new Spacer(1),
 					new Text(`${theme.fg("accent", `${theme.status.success} New session started`)}`, 1, 1),
-				);
+				]);
 				await this.ctx.reloadTodos();
 				this.ctx.ui.requestRender(true, { clearScrollback: true });
 
@@ -326,10 +325,6 @@ export class ExtensionUiController {
 					.then(() => this.#applyCustomMessageDisplay(wasStreaming, message.display))
 					.catch((err: unknown) => {
 						const errorText = `Extension sendMessage failed: ${err instanceof Error ? err.message : String(err)}`;
-						if (this.ctx.isBackgrounded) {
-							logger.error(errorText);
-							return;
-						}
 						this.ctx.showError(errorText);
 					});
 			},
@@ -374,9 +369,6 @@ export class ExtensionUiController {
 			getContextUsage: () => this.ctx.session.getContextUsage(),
 			waitForIdle: () => this.ctx.session.agent.waitForIdle(),
 			reload: async () => {
-				if (this.ctx.isBackgrounded) {
-					return;
-				}
 				await this.ctx.session.reload();
 				this.ctx.chatContainer.clear();
 				this.ctx.renderInitialMessages(undefined, { clearTerminalHistory: true });
@@ -384,9 +376,6 @@ export class ExtensionUiController {
 				this.ctx.showStatus("Reloaded session");
 			},
 			newSession: async options => {
-				if (this.ctx.isBackgrounded) {
-					return { cancelled: true };
-				}
 				// Stop any loading animation
 				if (this.ctx.loadingAnimation) {
 					this.ctx.loadingAnimation.stop();
@@ -415,19 +404,16 @@ export class ExtensionUiController {
 				this.ctx.streamingMessage = undefined;
 				this.ctx.pendingTools.clear();
 
-				this.ctx.chatContainer.addChild(new Spacer(1));
-				this.ctx.chatContainer.addChild(
+				this.ctx.present([
+					new Spacer(1),
 					new Text(`${theme.fg("accent", `${theme.status.success} New session started`)}`, 1, 1),
-				);
+				]);
 				await this.ctx.reloadTodos();
 				this.ctx.ui.requestRender(true, { clearScrollback: true });
 
 				return { cancelled: false };
 			},
 			branch: async entryId => {
-				if (this.ctx.isBackgrounded) {
-					return { cancelled: true };
-				}
 				const result = await this.ctx.session.branch(entryId);
 				if (result.cancelled) {
 					return { cancelled: true };
@@ -443,9 +429,6 @@ export class ExtensionUiController {
 				return { cancelled: false };
 			},
 			navigateTree: async (targetId, options) => {
-				if (this.ctx.isBackgrounded) {
-					return { cancelled: true };
-				}
 				const result = await this.ctx.session.navigateTree(targetId, { summarize: options?.summarize });
 				if (result.cancelled) {
 					return { cancelled: true };
@@ -464,9 +447,6 @@ export class ExtensionUiController {
 			},
 			compact: async instructionsOrOptions => this.#handleInteractiveCompact(instructionsOrOptions),
 			switchSession: async sessionPath => {
-				if (this.ctx.isBackgrounded) {
-					return { cancelled: true };
-				}
 				this.clearHookWidgets();
 				const result = await this.ctx.session.switchSession(sessionPath);
 				if (!result) {
@@ -482,36 +462,6 @@ export class ExtensionUiController {
 		extensionRunner.initialize(actions, contextActions, commandActions, uiContext);
 	}
 
-	createBackgroundUiContext(): ExtensionUIContext {
-		return {
-			select: async (_title: string, _options: ExtensionUISelectItem[], _dialogOptions) => undefined,
-			confirm: async (_title: string, _message: string, _dialogOptions) => false,
-			input: async (_title: string, _placeholder?: string, _dialogOptions?: unknown) => undefined,
-			notify: () => {},
-			onTerminalInput: () => () => {},
-			setStatus: () => {},
-			setWorkingMessage: () => {},
-			setWidget: () => {},
-			setTitle: () => {},
-			custom: async () => undefined as never,
-			setEditorText: () => {},
-			pasteToEditor: () => {},
-			getEditorText: () => "",
-			editor: async () => undefined,
-			get theme() {
-				return theme;
-			},
-			getAllThemes: () => Promise.resolve([]),
-			getTheme: () => Promise.resolve(undefined),
-			setTheme: () => Promise.resolve({ success: false, error: "Background mode" }),
-			setFooter: () => {},
-			setHeader: () => {},
-			setEditorComponent: () => {},
-			getToolsExpanded: () => false,
-			setToolsExpanded: () => {},
-		};
-	}
-
 	/**
 	 * Emit session event to all extension tools.
 	 */
@@ -531,7 +481,7 @@ export class ExtensionUiController {
 						ui: uiContext,
 						getContextUsage: () => this.ctx.session.getContextUsage(),
 						compact: instructionsOrOptions => this.#compactSession(instructionsOrOptions),
-						hasUI: !this.ctx.isBackgrounded,
+						hasUI: true,
 						cwd: this.ctx.sessionManager.getCwd(),
 						sessionManager: this.ctx.session.sessionManager,
 						modelRegistry: this.ctx.session.modelRegistry,
@@ -557,22 +507,14 @@ export class ExtensionUiController {
 	 * Show a tool error in the chat.
 	 */
 	showToolError(toolName: string, error: string): void {
-		if (this.ctx.isBackgrounded) {
-			logger.error(`Tool "${toolName}" error: ${error}`);
-			return;
-		}
 		const errorText = new Text(theme.fg("error", `Tool "${toolName}" error: ${error}`), 1, 0);
-		this.ctx.chatContainer.addChild(errorText);
-		this.ctx.ui.requestRender();
+		this.ctx.present(errorText);
 	}
 
 	/**
 	 * Set hook status text in the footer.
 	 */
 	setHookStatus(key: string, text: string | undefined): void {
-		if (this.ctx.isBackgrounded) {
-			return;
-		}
 		this.ctx.statusLine.setHookStatus(key, text);
 		this.ctx.ui.requestRender();
 	}
@@ -860,14 +802,9 @@ export class ExtensionUiController {
 
 	showExtensionError(extensionPath: string, error: string): void {
 		const errorText = new Text(theme.fg("error", `Extension "${extensionPath}" error: ${error}`), 1, 0);
-		this.ctx.chatContainer.addChild(errorText);
-		this.ctx.ui.requestRender();
+		this.ctx.present(errorText);
 	}
 	async #handleInteractiveCompact(instructionsOrOptions: string | CompactOptions | undefined): Promise<void> {
-		if (this.ctx.isBackgrounded) {
-			await this.#compactSession(instructionsOrOptions);
-			return;
-		}
 		await this.ctx.executeCompaction(instructionsOrOptions, false);
 	}
 
@@ -892,7 +829,7 @@ export class ExtensionUiController {
 	#applyCustomMessageDisplay(wasStreaming: boolean, shouldDisplay: boolean | undefined): void {
 		// For non-streaming cases with display=true, update UI
 		// (streaming cases update via message_end event)
-		if (!this.ctx.isBackgrounded && !wasStreaming && shouldDisplay) {
+		if (!wasStreaming && shouldDisplay) {
 			this.ctx.rebuildChatFromMessages();
 		}
 	}
diff --git a/packages/coding-agent/src/modes/controllers/input-controller.ts b/packages/coding-agent/src/modes/controllers/input-controller.ts
index 7a2e77573..da83ea41b 100644
--- a/packages/coding-agent/src/modes/controllers/input-controller.ts
+++ b/packages/coding-agent/src/modes/controllers/input-controller.ts
@@ -1,4 +1,5 @@
 import * as fs from "node:fs/promises";
+import type { ImageContent } from "@oh-my-pi/pi-ai";
 import type { AutocompleteProvider, SlashCommand } from "@oh-my-pi/pi-tui";
 import { $env, logger, sanitizeText } from "@oh-my-pi/pi-utils";
 import { getRoleInfo } from "../../config/model-registry";
@@ -9,16 +10,16 @@ import { expandEmoticons } from "../../modes/emoji-autocomplete";
 import { materializeImageReferenceLinks } from "../../modes/image-references";
 import { createPromptActionAutocompleteProvider } from "../../modes/prompt-action-autocomplete";
 import type { InteractiveModeContext } from "../../modes/types";
-import type { AgentSessionEvent } from "../../session/agent-session";
-import { SKILL_PROMPT_MESSAGE_TYPE, type SkillPromptDetails } from "../../session/messages";
+import { SKILL_PROMPT_MESSAGE_TYPE, type SkillPromptDetails, USER_INTERRUPT_LABEL } from "../../session/messages";
 import { executeBuiltinSlashCommand } from "../../slash-commands/builtin-registry";
 import { isTinyTitleLocalModelKey } from "../../tiny/models";
 import { isLowSignalTitleInput } from "../../tiny/text";
 import { tinyTitleClient } from "../../tiny/title-client";
 import type { TinyTitleProgressEvent } from "../../tiny/title-protocol";
 import { copyToClipboard, readImageFromClipboard, readTextFromClipboard } from "../../utils/clipboard";
+import { EnhancedPasteController } from "../../utils/enhanced-paste";
 import { getEditorCommand, openInEditor } from "../../utils/external-editor";
-import { ensureSupportedImageInput } from "../../utils/image-loading";
+import { ensureSupportedImageInput, ImageInputTooLargeError, loadImageInput } from "../../utils/image-loading";
 import { resizeImage } from "../../utils/image-resize";
 import { generateSessionTitle, setSessionTerminalTitle } from "../../utils/title-generator";
 
@@ -40,8 +41,10 @@ const TINY_TITLE_PROGRESS_REVEAL_DELAY_MS = 1_000;
 export class InputController {
 	constructor(private ctx: InteractiveModeContext) {}
 
+	#enhancedPaste?: EnhancedPasteController;
+
 	#showTinyTitleDownloadProgress(modelKey: string): void {
-		if (!isTinyTitleLocalModelKey(modelKey) || this.ctx.isBackgrounded) return;
+		if (!isTinyTitleLocalModelKey(modelKey)) return;
 		const component = new TinyTitleDownloadProgressComponent(modelKey);
 		let added = false;
 		let disposed = false;
@@ -91,7 +94,7 @@ export class InputController {
 			if (this.ctx.loopModeEnabled) {
 				this.ctx.pauseLoop();
 				if (this.ctx.session.isStreaming) {
-					void this.ctx.session.abort();
+					void this.ctx.session.abort({ reason: USER_INTERRUPT_LABEL });
 				} else {
 					this.ctx.cancelPendingSubmission();
 				}
@@ -121,7 +124,7 @@ export class InputController {
 				this.ctx.isPythonMode = false;
 				this.ctx.updateEditorBorderColor();
 			} else if (this.ctx.session.isStreaming) {
-				void this.ctx.session.abort();
+				void this.ctx.session.abort({ reason: USER_INTERRUPT_LABEL });
 			} else if (!this.ctx.editor.getText().trim()) {
 				// Double-interrupt with empty editor triggers /tree, /branch, or nothing based on setting
 				const action = settings.get("doubleEscapeAction");
@@ -144,6 +147,8 @@ export class InputController {
 		this.ctx.editor.setActionKeys("app.clear", this.ctx.keybindings.getKeys("app.clear"));
 		this.ctx.editor.onClear = () => this.handleCtrlC();
 		this.ctx.editor.setActionKeys("app.exit", this.ctx.keybindings.getKeys("app.exit"));
+		this.ctx.editor.setActionKeys("app.display.reset", this.ctx.keybindings.getKeys("app.display.reset"));
+		this.ctx.editor.onDisplayReset = () => this.ctx.ui.resetDisplay();
 		this.ctx.editor.onExit = () => this.handleCtrlD();
 		this.ctx.editor.setActionKeys("app.suspend", this.ctx.keybindings.getKeys("app.suspend"));
 		this.ctx.editor.onSuspend = () => this.handleCtrlZ();
@@ -174,6 +179,7 @@ export class InputController {
 			this.ctx.keybindings.getKeys("app.clipboard.pasteImage"),
 		);
 		this.ctx.editor.onPasteImage = () => this.handleImagePaste();
+		this.ctx.editor.onPasteImagePath = path => void this.handleImagePathPaste(path);
 		this.ctx.editor.setActionKeys(
 			"app.clipboard.pasteTextRaw",
 			this.ctx.keybindings.getKeys("app.clipboard.pasteTextRaw"),
@@ -188,11 +194,9 @@ export class InputController {
 		this.ctx.editor.onExpandTools = () => this.toggleToolOutputExpansion();
 		this.ctx.editor.setActionKeys("app.message.dequeue", this.ctx.keybindings.getKeys("app.message.dequeue"));
 		this.ctx.editor.onDequeue = () => this.handleDequeue();
-
 		this.ctx.editor.clearCustomKeyHandlers();
 		// Wire up extension shortcuts
 		this.registerExtensionShortcuts();
-
 		const planModeKeys = this.ctx.keybindings.getKeys("app.plan.toggle");
 		for (const key of planModeKeys) {
 			this.ctx.editor.setCustomKeyHandler(key, () => void this.ctx.handlePlanModeCommand());
@@ -223,6 +227,8 @@ export class InputController {
 			this.ctx.editor.setCustomKeyHandler(key, () => this.ctx.showSessionObserver());
 		}
 
+		this.#setupEnhancedPaste();
+
 		this.ctx.editor.onChange = (text: string) => {
 			const wasBashMode = this.ctx.isBashMode;
 			const wasPythonMode = this.ctx.isPythonMode;
@@ -235,16 +241,45 @@ export class InputController {
 		};
 	}
 
+	#setupEnhancedPaste(): void {
+		if (this.#enhancedPaste) return;
+
+		this.#enhancedPaste = new EnhancedPasteController({
+			write: data => this.ctx.ui.terminal.write(data),
+			pasteText: text => {
+				this.ctx.editor.pasteText(text);
+				this.ctx.ui.requestRender(false, { allowUnknownViewportMutation: true });
+			},
+			pasteImage: async image => {
+				await this.#normalizeAndInsertPastedImage(image, `Unsupported pasted image format: ${image.mimeType}`);
+			},
+			showStatus: message => this.ctx.showStatus(message),
+		});
+		this.ctx.ui.addInputListener(data => (this.#enhancedPaste?.handleInput(data) ? { consume: true } : undefined));
+		this.ctx.ui.addStartListener(() => this.#enhancedPaste?.enable());
+	}
+
 	setupEditorSubmitHandler(): void {
 		this.ctx.editor.onSubmit = async (text: string) => {
 			text = text.trim();
 			if ((!isSettingsInitialized() || settings.get("emojiAutocomplete")) && text) text = expandEmoticons(text);
 
-			// Empty submit while streaming with queued messages: flush queues immediately
-			if (!text && this.ctx.session.isStreaming && this.ctx.session.queuedMessageCount > 0) {
-				// Abort current stream and let queued messages be processed
-				await this.ctx.session.abort();
-				return;
+			// Empty submit while streaming with queued steering: interrupt now and
+			// immediately resume so the visible `Steer:` entry is sent without
+			// waiting for the current tool/model boundary.
+			if (!text && this.ctx.session.isStreaming) {
+				const queuedMessages = this.ctx.session.getQueuedMessages();
+				if (queuedMessages.steering.length > 0) {
+					await this.ctx.session.interruptAndFlushQueuedMessages({ reason: USER_INTERRUPT_LABEL });
+					this.ctx.updatePendingMessagesDisplay();
+					this.ctx.ui.requestRender();
+					return;
+				}
+				if (this.ctx.session.queuedMessageCount > 0) {
+					// Preserve the existing empty-submit flush for non-steer queues.
+					await this.ctx.session.abort({ reason: USER_INTERRUPT_LABEL });
+					return;
+				}
 			}
 
 			if (!text) return;
@@ -291,7 +326,6 @@ export class InputController {
 			// Handle built-in slash commands
 			const slashResult = await executeBuiltinSlashCommand(text, {
 				ctx: this.ctx,
-				handleBackgroundCommand: () => this.handleBackgroundCommand(),
 			});
 			if (slashResult === true) {
 				return;
@@ -600,7 +634,7 @@ export class InputController {
 		if (allQueued.length === 0) {
 			this.ctx.updatePendingMessagesDisplay();
 			if (options?.abort) {
-				this.ctx.session.abort();
+				this.ctx.session.abort({ reason: USER_INTERRUPT_LABEL });
 			}
 			return 0;
 		}
@@ -610,128 +644,99 @@ export class InputController {
 		this.ctx.editor.setText(combinedText);
 		this.ctx.updatePendingMessagesDisplay();
 		if (options?.abort) {
-			this.ctx.session.abort();
+			this.ctx.session.abort({ reason: USER_INTERRUPT_LABEL });
 		}
 		return allQueued.length;
 	}
 
-	handleBackgroundCommand(): void {
-		if (this.ctx.isBackgrounded) {
-			this.ctx.showStatus("Background mode already enabled");
-			return;
-		}
-		if (!this.ctx.session.isStreaming && this.ctx.session.queuedMessageCount === 0) {
-			this.ctx.showWarning("Agent is idle; nothing to background");
-			return;
-		}
-		if (this.ctx.hasActiveBtw()) {
-			this.ctx.handleBtwEscape();
-		}
-		if (this.ctx.hasActiveOmfg()) {
-			this.ctx.handleOmfgEscape();
-		}
-
-		this.ctx.isBackgrounded = true;
-		const backgroundUiContext = this.ctx.createBackgroundUiContext();
-
-		// Background mode disables interactive UI so tools like ask fail fast.
-		this.ctx.setToolUIContext(backgroundUiContext, false);
-		this.ctx.initializeHookRunner(backgroundUiContext, false);
-
-		if (this.ctx.loadingAnimation) {
-			this.ctx.loadingAnimation.stop();
-			this.ctx.loadingAnimation = undefined;
-		}
-		if (this.ctx.autoCompactionLoader) {
-			this.ctx.autoCompactionLoader.stop();
-			this.ctx.autoCompactionLoader = undefined;
-		}
-		if (this.ctx.retryLoader) {
-			this.ctx.retryLoader.stop();
-			this.ctx.retryLoader = undefined;
-		}
-		this.ctx.statusContainer.clear();
-		this.ctx.statusLine.dispose();
-
-		if (this.ctx.unsubscribe) {
-			this.ctx.unsubscribe();
-		}
-		this.ctx.unsubscribe = this.ctx.session.subscribe(async (event: AgentSessionEvent) => {
-			await this.ctx.handleBackgroundEvent(event);
+	async #insertPendingImage(imageData: ImageContent): Promise<void> {
+		const imageLink = (
+			await materializeImageReferenceLinks(
+				[
+					{
+						type: "image",
+						data: imageData.data,
+						mimeType: imageData.mimeType,
+					},
+				],
+				this.ctx.sessionManager.putBlob.bind(this.ctx.sessionManager),
+			)
+		)?.[0];
+		this.ctx.pendingImages.push({
+			type: "image",
+			data: imageData.data,
+			mimeType: imageData.mimeType,
 		});
+		this.ctx.pendingImageLinks.push(imageLink);
+		this.ctx.editor.imageLinks = this.ctx.pendingImageLinks;
+		const imageNum = this.ctx.pendingImages.length;
+		this.ctx.editor.insertText(`[Image #${imageNum}] `);
+		this.ctx.ui.requestRender(false, { allowUnknownViewportMutation: true });
+	}
 
-		// Backgrounding keeps the current process to preserve in-flight agent state.
-		if (this.ctx.isInitialized) {
-			this.ctx.ui.stop();
-			this.ctx.isInitialized = false;
+	async #normalizeAndInsertPastedImage(image: ImageContent, unsupportedMessage: string): Promise<boolean> {
+		let imageData = await ensureSupportedImageInput(image);
+		if (!imageData) {
+			this.ctx.showStatus(unsupportedMessage);
+			return false;
 		}
-
-		process.stdout.write("Background mode enabled. Run `bg` to continue in background.\n");
-
-		if (process.platform === "win32" || !process.stdout.isTTY) {
-			process.stdout.write("Backgrounding requires POSIX job control; continuing in foreground.\n");
-			return;
+		if (settings.get("images.autoResize")) {
+			try {
+				const resized = await resizeImage({
+					type: "image",
+					data: imageData.data,
+					mimeType: imageData.mimeType,
+				});
+				imageData = { type: "image", data: resized.data, mimeType: resized.mimeType };
+			} catch {
+				// Keep the normalized image when resize fails.
+			}
 		}
+		await this.#insertPendingImage(imageData);
+		return true;
+	}
 
-		process.kill(0, "SIGTSTP");
+	async handleImagePathPaste(path: string): Promise<void> {
+		try {
+			const image = await loadImageInput({
+				path,
+				cwd: this.ctx.sessionManager.getCwd(),
+				autoResize: false,
+			});
+			if (!image) {
+				this.ctx.editor.pasteText(path);
+				this.ctx.ui.requestRender(false, { allowUnknownViewportMutation: true });
+				this.ctx.showStatus("Pasted path is not a supported image");
+				return;
+			}
+			await this.#normalizeAndInsertPastedImage(
+				{ type: "image", data: image.data, mimeType: image.mimeType },
+				`Unsupported pasted image format: ${image.mimeType}`,
+			);
+		} catch (error) {
+			this.ctx.editor.pasteText(path);
+			this.ctx.ui.requestRender(false, { allowUnknownViewportMutation: true });
+			this.ctx.showStatus(
+				error instanceof ImageInputTooLargeError ? error.message : "Failed to read pasted image path",
+			);
+		}
 	}
 
 	async handleImagePaste(): Promise<boolean> {
 		try {
 			const image = await readImageFromClipboard();
-			if (image) {
-				const base64Data = image.data.toBase64();
-				let imageData = await ensureSupportedImageInput({
-					type: "image",
-					data: base64Data,
-					mimeType: image.mimeType,
-				});
-				if (!imageData) {
-					this.ctx.showStatus(`Unsupported clipboard image format: ${image.mimeType}`);
-					return false;
-				}
-				if (settings.get("images.autoResize")) {
-					try {
-						const resized = await resizeImage({
-							type: "image",
-							data: imageData.data,
-							mimeType: imageData.mimeType,
-						});
-						imageData = { type: "image", data: resized.data, mimeType: resized.mimeType };
-					} catch {
-						// Keep the normalized image when resize fails.
-					}
-				}
-
-				const imageLink = (
-					await materializeImageReferenceLinks(
-						[
-							{
-								type: "image",
-								data: imageData.data,
-								mimeType: imageData.mimeType,
-							},
-						],
-						this.ctx.sessionManager.putBlob.bind(this.ctx.sessionManager),
-					)
-				)?.[0];
-				this.ctx.pendingImages.push({
-					type: "image",
-					data: imageData.data,
-					mimeType: imageData.mimeType,
-				});
-				this.ctx.pendingImageLinks.push(imageLink);
-				this.ctx.editor.imageLinks = this.ctx.pendingImageLinks;
-				// Insert placeholder at cursor like Claude does
-				const imageNum = this.ctx.pendingImages.length;
-				const placeholder = `[Image #${imageNum}]`;
-				this.ctx.editor.insertText(`${placeholder} `);
-				this.ctx.ui.requestRender();
-				return true;
+			if (!image) {
+				this.ctx.showStatus("No image in clipboard (use terminal paste for text)");
+				return false;
 			}
-			// No image in clipboard - show hint
-			this.ctx.showStatus("No image in clipboard (use terminal paste for text)");
-			return false;
+			return await this.#normalizeAndInsertPastedImage(
+				{
+					type: "image",
+					data: image.data.toBase64(),
+					mimeType: image.mimeType,
+				},
+				`Unsupported clipboard image format: ${image.mimeType}`,
+			);
 		} catch {
 			this.ctx.showStatus("Failed to read clipboard");
 			return false;
@@ -846,7 +851,15 @@ export class InputController {
 				child.setExpanded(expanded);
 			}
 		}
-		this.ctx.ui.requestRender(false, { allowUnknownViewportMutation: true });
+		// Toggling expansion mutates every block, but on ED3-risk terminals the
+		// transcript freezes a snapshot of each block once it scrolls past the live
+		// region (committed native scrollback is immutable there). A plain repaint
+		// replays those stale snapshots, so the toggle appears to do nothing above
+		// the live block. resetDisplay() invalidates the snapshots and forces a
+		// full clear + replay — the keyboard-accessible resize-reset equivalent —
+		// which is the only path that re-emits the whole transcript at its new
+		// heights.
+		this.ctx.ui.resetDisplay();
 	}
 
 	toggleThinkingBlockVisibility(): void {
diff --git a/packages/coding-agent/src/modes/controllers/mcp-command-controller.ts b/packages/coding-agent/src/modes/controllers/mcp-command-controller.ts
index 756328b4b..c63e57162 100644
--- a/packages/coding-agent/src/modes/controllers/mcp-command-controller.ts
+++ b/packages/coding-agent/src/modes/controllers/mcp-command-controller.ts
@@ -37,7 +37,9 @@ import type { MCPAuthConfig, MCPServerConfig, MCPServerConnection } from "../../
 import type { OAuthCredential } from "../../session/auth-storage";
 import { shortenPath } from "../../tools/render-utils";
 import { openPath } from "../../utils/open";
+import { ChatBlock } from "../components/chat-block";
 import { MCPAddWizard } from "../components/mcp-add-wizard";
+import { TranscriptBlock } from "../components/transcript-container";
 import { parseCommandArgs } from "../shared";
 import { theme } from "../theme/theme";
 import type { InteractiveModeContext } from "../types";
@@ -49,6 +51,42 @@ function withTimeout<T>(promise: Promise<T>, timeoutMs: number, message: string)
 	return Promise.race([promise, timeoutPromise]).finally(() => clearTimeout(timer));
 }
 
+/**
+ * Animated "Connecting to …" transcript block. Owns its spinner interval: it
+ * starts on mount and is cleared on {@link ChatBlock.finish}/dispose, so callers
+ * never juggle `setInterval`/`clearInterval` or `requestRender` by hand.
+ */
+class McpConnectingBlock extends ChatBlock {
+	readonly #text: Text;
+
+	constructor(private readonly serverName: string) {
+		super();
+		this.addChild(new Spacer(1));
+		const frame = theme.spinnerFrames[0] ?? "|";
+		this.#text = new Text(theme.fg("muted", `${frame} Connecting to "${serverName}"...`), 1, 0);
+		this.addChild(this.#text);
+	}
+
+	protected override onMount(): void {
+		const frames = theme.spinnerFrames;
+		let frame = 0;
+		const interval = setInterval(() => {
+			frame++;
+			this.#text.setText(
+				theme.fg("muted", `${frames[frame % frames.length] ?? "|"} Connecting to "${this.serverName}"...`),
+			);
+			this.requestRender();
+		}, 80);
+		this.onCleanup(() => clearInterval(interval));
+	}
+
+	/** Replace the spinner line with a terminal status; pair with {@link finish}. */
+	setStatus(text: string): void {
+		this.#text.setText(text);
+		this.requestRender();
+	}
+}
+
 /**
  * Outcome of {@link MCPCommandController}'s OAuth handler.
  *
@@ -547,63 +585,45 @@ export class MCPCommandController {
 				},
 				{
 					onAuth: (info: { url: string; instructions?: string }) => {
-						// Show auth URL prominently in chat
-						this.ctx.chatContainer.addChild(new Spacer(1));
-						this.ctx.chatContainer.addChild(
-							new Text(theme.fg("accent", "━━━ OAuth Authorization Required ━━━"), 1, 0),
-						);
-						this.ctx.chatContainer.addChild(new Spacer(1));
-						this.ctx.chatContainer.addChild(
-							new Text(theme.fg("muted", "Preparing browser authorization..."), 1, 0),
-						);
-						this.ctx.chatContainer.addChild(new Spacer(1));
-						this.ctx.chatContainer.addChild(
+						// Show auth URL prominently in chat as one block
+						const block = new TranscriptBlock();
+						this.ctx.present(block);
+						block.addChild(new Text(theme.fg("accent", "━━━ OAuth Authorization Required ━━━"), 1, 0));
+						block.addChild(new Spacer(1));
+						block.addChild(new Text(theme.fg("muted", "Preparing browser authorization..."), 1, 0));
+						block.addChild(new Spacer(1));
+						block.addChild(
 							new Text(
 								theme.fg("muted", "Waiting for authorization... (Press Ctrl+C to cancel, 5 minute timeout)"),
 								1,
 								0,
 							),
 						);
-						this.ctx.chatContainer.addChild(new Spacer(1));
-						this.ctx.chatContainer.addChild(
-							new Text(theme.fg("accent", "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"), 1, 0),
-						);
-						this.ctx.ui.requestRender();
+						block.addChild(new Spacer(1));
+						block.addChild(new Text(theme.fg("accent", "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"), 1, 0));
 						// Try to open browser automatically
 						try {
 							openPath(info.url);
 
 							// Show confirmation that browser should open
-							this.ctx.chatContainer.addChild(new Spacer(1));
-							this.ctx.chatContainer.addChild(
-								new Text(theme.fg("success", "→ Opening browser automatically..."), 1, 0),
-							);
-							this.ctx.chatContainer.addChild(new Spacer(1));
-							this.ctx.chatContainer.addChild(
-								new Text(theme.fg("muted", "Alternative if browser did not open:"), 1, 0),
-							);
-							this.ctx.chatContainer.addChild(
-								new Text(theme.fg("success", "Copy this exact URL in your browser:"), 1, 0),
-							);
-							this.ctx.chatContainer.addChild(new Text(theme.fg("accent", info.url), 1, 0));
+							block.addChild(new Spacer(1));
+							block.addChild(new Text(theme.fg("success", "→ Opening browser automatically..."), 1, 0));
+							block.addChild(new Spacer(1));
+							block.addChild(new Text(theme.fg("muted", "Alternative if browser did not open:"), 1, 0));
+							block.addChild(new Text(theme.fg("success", "Copy this exact URL in your browser:"), 1, 0));
+							block.addChild(new Text(theme.fg("accent", info.url), 1, 0));
 							this.ctx.ui.requestRender();
 						} catch (_error) {
 							// Show error if browser doesn't open
-							this.ctx.chatContainer.addChild(new Spacer(1));
-							this.ctx.chatContainer.addChild(
-								new Text(theme.fg("warning", "→ Could not open browser automatically"), 1, 0),
-							);
-							this.ctx.chatContainer.addChild(
-								new Text(theme.fg("success", "Copy this exact URL in your browser:"), 1, 0),
-							);
-							this.ctx.chatContainer.addChild(new Text(theme.fg("accent", info.url), 1, 0));
+							block.addChild(new Spacer(1));
+							block.addChild(new Text(theme.fg("warning", "→ Could not open browser automatically"), 1, 0));
+							block.addChild(new Text(theme.fg("success", "Copy this exact URL in your browser:"), 1, 0));
+							block.addChild(new Text(theme.fg("accent", info.url), 1, 0));
 							this.ctx.ui.requestRender();
 						}
 					},
 					onProgress: (message: string) => {
-						this.ctx.chatContainer.addChild(new Spacer(1));
-						this.ctx.chatContainer.addChild(new Text(theme.fg("muted", message), 1, 0));
-						this.ctx.ui.requestRender();
+						this.ctx.present([new Spacer(1), new Text(theme.fg("muted", message), 1, 0)]);
 					},
 				},
 			);
@@ -611,9 +631,10 @@ export class MCPCommandController {
 			// Execute OAuth flow with 5 minute timeout
 			const credentials = await withTimeout(flow.login(), 5 * 60 * 1000, "OAuth flow timed out after 5 minutes");
 
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(new Text(theme.fg("success", "✓ Authorization completed in browser."), 1, 0));
-			this.ctx.ui.requestRender();
+			this.ctx.present([
+				new Spacer(1),
+				new Text(theme.fg("success", "✓ Authorization completed in browser."), 1, 0),
+			]);
 
 			// Generate a unique credential ID
 			const credentialId = `mcp_oauth_${Date.now()}_${Math.random().toString(36).slice(2, 11)}`;
@@ -766,19 +787,8 @@ export class MCPCommandController {
 	): Promise<"connected" | "connecting" | "disconnected"> {
 		if (!this.ctx.mcpManager) return "disconnected";
 
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		const frames = theme.spinnerFrames;
-		const initialFrame = frames[0] ?? "|";
-		const statusText = new Text(theme.fg("muted", `${initialFrame} Connecting to "${name}"...`), 1, 0);
-		this.ctx.chatContainer.addChild(statusText);
-		this.ctx.ui.requestRender();
-
-		let frame = 0;
-		const interval = setInterval(() => {
-			statusText.setText(theme.fg("muted", `${frames[frame % frames.length]} Connecting to "${name}"...`));
-			frame++;
-			this.ctx.ui.requestRender();
-		}, 80);
+		const block = new McpConnectingBlock(name);
+		this.ctx.present(block);
 
 		try {
 			try {
@@ -792,20 +802,19 @@ export class MCPCommandController {
 				await this.ctx.session.refreshMCPTools(this.ctx.mcpManager.getTools());
 			}
 			if (state === "connected") {
-				statusText.setText(theme.fg("success", `✓ Connected to "${name}"`));
+				block.setStatus(theme.fg("success", `✓ Connected to "${name}"`));
 			} else if (state === "connecting") {
-				statusText.setText(theme.fg("muted", `◌ "${name}" is still connecting...`));
+				block.setStatus(theme.fg("muted", `◌ "${name}" is still connecting...`));
 			} else {
-				statusText.setText(
+				block.setStatus(
 					options?.suppressDisconnectedWarning
 						? theme.fg("muted", `◌ Connection check complete for "${name}"`)
 						: theme.fg("warning", `⚠ Could not connect to "${name}" yet`),
 				);
 			}
-			this.ctx.ui.requestRender();
 			return state;
 		} finally {
-			clearInterval(interval);
+			block.finish();
 		}
 	}
 
diff --git a/packages/coding-agent/src/modes/controllers/selector-controller.ts b/packages/coding-agent/src/modes/controllers/selector-controller.ts
index 7ba39df43..2d2728db2 100644
--- a/packages/coding-agent/src/modes/controllers/selector-controller.ts
+++ b/packages/coding-agent/src/modes/controllers/selector-controller.ts
@@ -7,7 +7,6 @@ import { getAgentDbPath, getProjectDir, normalizePathForComparison } from "@oh-m
 import { getRoleInfo } from "../../config/model-registry";
 import { formatModelSelectorValue } from "../../config/model-resolver";
 import { settings } from "../../config/settings";
-import { DebugSelectorComponent } from "../../debug";
 import { disableProvider, enableProvider } from "../../discovery";
 import { clearPluginRootsAndCaches, resolveActiveProjectRegistryPath } from "../../discovery/helpers";
 import {
@@ -37,9 +36,11 @@ import {
 	setPreferredSearchProvider,
 } from "../../tools";
 import { shortenPath } from "../../tools/render-utils";
+import { copyToClipboard } from "../../utils/clipboard";
 import { setSessionTerminalTitle } from "../../utils/title-generator";
 import { AgentDashboard } from "../components/agent-dashboard";
 import { AssistantMessageComponent } from "../components/assistant-message";
+import { CopySelectorComponent } from "../components/copy-selector";
 import { ExtensionDashboard } from "../components/extensions";
 import { HistorySearchComponent } from "../components/history-search";
 import { ModelSelectorComponent } from "../components/model-selector";
@@ -49,9 +50,12 @@ import { SessionObserverOverlayComponent } from "../components/session-observer-
 import { SessionSelectorComponent } from "../components/session-selector";
 import { SettingsSelectorComponent } from "../components/settings-selector";
 import { ToolExecutionComponent } from "../components/tool-execution";
+import { TranscriptBlock } from "../components/transcript-container";
 import { TreeSelectorComponent } from "../components/tree-selector";
 import { UserMessageSelectorComponent } from "../components/user-message-selector";
 import type { SessionObserverRegistry } from "../session-observer-registry";
+import { computeContextBreakdown } from "../utils/context-usage";
+import { buildCopyTargets } from "../utils/copy-targets";
 
 const CALLBACK_SERVER_PROVIDERS = new Set<OAuthProvider>([
 	"anthropic",
@@ -407,6 +411,7 @@ export class SelectorController {
 	}
 
 	showModelSelector(options?: { temporaryOnly?: boolean }): void {
+		const currentContextTokens = computeContextBreakdown(this.ctx.session).usedTokens;
 		this.showSelector(done => {
 			const selector = new ModelSelectorComponent(
 				this.ctx.ui,
@@ -470,7 +475,7 @@ export class SelectorController {
 					done();
 					this.ctx.ui.requestRender();
 				},
-				options,
+				{ ...options, currentContextTokens },
 			);
 			return { component: selector, focus: selector };
 		});
@@ -598,6 +603,38 @@ export class SelectorController {
 		});
 	}
 
+	showCopySelector(): void {
+		const targets = buildCopyTargets(this.ctx.session);
+		if (targets.length === 0) {
+			this.ctx.showStatus("Nothing to copy yet.");
+			return;
+		}
+
+		let overlayHandle: OverlayHandle | undefined;
+		const done = () => {
+			overlayHandle?.hide();
+			this.ctx.ui.requestRender();
+		};
+		const selector = new CopySelectorComponent(targets, {
+			onPick: target => {
+				done();
+				if (target.content === undefined) return;
+				void copyToClipboard(target.content);
+				this.ctx.showStatus(target.copyMessage ?? "Copied to clipboard");
+			},
+			onCancel: done,
+		});
+
+		overlayHandle = this.ctx.ui.showOverlay(selector, {
+			anchor: "bottom-center",
+			width: "100%",
+			maxHeight: "100%",
+			margin: 0,
+		});
+		this.ctx.ui.setFocus(selector);
+		this.ctx.ui.requestRender();
+	}
+
 	showTreeSelector(): void {
 		const tree = this.ctx.sessionManager.getTree();
 		const realLeafId = this.ctx.sessionManager.getLeafId();
@@ -895,28 +932,28 @@ export class SelectorController {
 		try {
 			await this.ctx.session.modelRegistry.authStorage.login(providerId as OAuthProvider, {
 				onAuth: (info: { url: string; instructions?: string }) => {
-					this.ctx.chatContainer.addChild(new Spacer(1));
-					this.ctx.chatContainer.addChild(new Text(theme.fg("dim", info.url), 1, 0));
+					const block = new TranscriptBlock();
+					block.addChild(new Text(theme.fg("dim", info.url), 1, 0));
 					const hyperlink = `\x1b]8;;${info.url}\x07Click here to login\x1b]8;;\x07`;
-					this.ctx.chatContainer.addChild(new Text(theme.fg("accent", hyperlink), 1, 0));
+					block.addChild(new Text(theme.fg("accent", hyperlink), 1, 0));
 					if (info.instructions) {
-						this.ctx.chatContainer.addChild(new Spacer(1));
-						this.ctx.chatContainer.addChild(new Text(theme.fg("warning", info.instructions), 1, 0));
+						block.addChild(new Spacer(1));
+						block.addChild(new Text(theme.fg("warning", info.instructions), 1, 0));
 					}
 					if (useManualInput) {
-						this.ctx.chatContainer.addChild(new Spacer(1));
-						this.ctx.chatContainer.addChild(new Text(theme.fg("dim", MANUAL_LOGIN_TIP), 1, 0));
+						block.addChild(new Spacer(1));
+						block.addChild(new Text(theme.fg("dim", MANUAL_LOGIN_TIP), 1, 0));
 					}
-					this.ctx.ui.requestRender();
+					this.ctx.present(block);
 					this.ctx.openInBrowser(info.url);
 				},
 				onPrompt: async (prompt: { message: string; placeholder?: string }) => {
-					this.ctx.chatContainer.addChild(new Spacer(1));
-					this.ctx.chatContainer.addChild(new Text(theme.fg("warning", prompt.message), 1, 0));
+					const promptBlock = new TranscriptBlock();
+					promptBlock.addChild(new Text(theme.fg("warning", prompt.message), 1, 0));
 					if (prompt.placeholder) {
-						this.ctx.chatContainer.addChild(new Text(theme.fg("dim", prompt.placeholder), 1, 0));
+						promptBlock.addChild(new Text(theme.fg("dim", prompt.placeholder), 1, 0));
 					}
-					this.ctx.ui.requestRender();
+					this.ctx.present(promptBlock);
 					const { promise, resolve } = Promise.withResolvers<string>();
 					const codeInput = new Input();
 					codeInput.onSubmit = () => {
@@ -933,18 +970,17 @@ export class SelectorController {
 					return promise;
 				},
 				onProgress: (message: string) => {
-					this.ctx.chatContainer.addChild(new Text(theme.fg("dim", message), 1, 0));
-					this.ctx.ui.requestRender();
+					this.ctx.present(new Text(theme.fg("dim", message), 1, 0));
 				},
 				onManualCodeInput: useManualInput ? () => manualInput.waitForInput(providerId) : undefined,
 			});
 			await this.ctx.session.modelRegistry.refresh();
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
+			const block = new TranscriptBlock();
+			block.addChild(
 				new Text(theme.fg("success", `${theme.status.success} Successfully logged in to ${providerId}`), 1, 0),
 			);
-			this.ctx.chatContainer.addChild(new Text(theme.fg("dim", `Credentials saved to ${getAgentDbPath()}`), 1, 0));
-			this.ctx.ui.requestRender();
+			block.addChild(new Text(theme.fg("dim", `Credentials saved to ${getAgentDbPath()}`), 1, 0));
+			this.ctx.present(block);
 		} catch (error: unknown) {
 			this.ctx.showError(`Login failed: ${error instanceof Error ? error.message : String(error)}`);
 		} finally {
@@ -966,20 +1002,18 @@ export class SelectorController {
 
 			await authStorage.logout(providerId);
 			await this.ctx.session.modelRegistry.refresh();
-			this.ctx.chatContainer.addChild(new Spacer(1));
-			this.ctx.chatContainer.addChild(
+			const block = new TranscriptBlock();
+			block.addChild(
 				new Text(theme.fg("success", `${theme.status.success} Successfully logged out of ${providerId}`), 1, 0),
 			);
-			this.ctx.chatContainer.addChild(
-				new Text(theme.fg("dim", `Credentials removed from ${getAgentDbPath()}`), 1, 0),
-			);
+			block.addChild(new Text(theme.fg("dim", `Credentials removed from ${getAgentDbPath()}`), 1, 0));
 			const remainingSource = authStorage.describeCredentialSource(providerId, this.ctx.session.sessionId);
 			if (remainingSource) {
-				this.ctx.chatContainer.addChild(
+				block.addChild(
 					new Text(theme.fg("warning", `${providerId} is still authenticated via ${remainingSource}`), 1, 0),
 				);
 			}
-			this.ctx.ui.requestRender();
+			this.ctx.present(block);
 		} catch (error: unknown) {
 			this.ctx.showError(`Logout failed: ${error instanceof Error ? error.message : String(error)}`);
 		}
@@ -1043,7 +1077,8 @@ export class SelectorController {
 		});
 	}
 
-	showDebugSelector(): void {
+	async showDebugSelector(): Promise<void> {
+		const { DebugSelectorComponent } = await import("../../debug");
 		this.showSelector(done => {
 			const selector = new DebugSelectorComponent(this.ctx, done);
 			return { component: selector, focus: selector };
diff --git a/packages/coding-agent/src/modes/controllers/streaming-reveal.ts b/packages/coding-agent/src/modes/controllers/streaming-reveal.ts
new file mode 100644
index 000000000..1e4edccbf
--- /dev/null
+++ b/packages/coding-agent/src/modes/controllers/streaming-reveal.ts
@@ -0,0 +1,212 @@
+import type { AssistantMessage } from "@oh-my-pi/pi-ai";
+import { getSegmenter } from "@oh-my-pi/pi-tui";
+import type { AssistantMessageComponent } from "../components/assistant-message";
+
+export const STREAMING_REVEAL_FRAME_MS = 1000 / 30;
+export const MIN_STEP = 3;
+export const CATCHUP_FRAMES = 8;
+
+type AssistantContentBlock = AssistantMessage["content"][number];
+type StreamingRevealComponent = Pick<AssistantMessageComponent, "updateContent">;
+
+type StreamingRevealControllerOptions = {
+	getSmoothStreaming(): boolean;
+	getHideThinkingBlock(): boolean;
+	requestRender(): void;
+};
+
+function countGraphemes(text: string): number {
+	let count = 0;
+	for (const _segment of getSegmenter().segment(text)) {
+		count += 1;
+	}
+	return count;
+}
+
+function sliceGraphemes(text: string, units: number): string {
+	if (units <= 0 || text.length === 0) return "";
+	let count = 0;
+	for (const { index, segment } of getSegmenter().segment(text)) {
+		count += 1;
+		if (count >= units) {
+			const end = index + segment.length;
+			return end >= text.length ? text : text.slice(0, end);
+		}
+	}
+	return text;
+}
+
+export function visibleUnits(message: AssistantMessage, hideThinking: boolean): number {
+	let total = 0;
+	for (const block of message.content) {
+		if (block.type === "text") {
+			total += countGraphemes(block.text);
+		} else if (block.type === "thinking" && !hideThinking) {
+			total += countGraphemes(block.thinking);
+		}
+	}
+	return total;
+}
+
+function revealTextBlock(
+	block: Extract<AssistantContentBlock, { type: "text" }>,
+	remaining: number,
+): AssistantContentBlock {
+	if (remaining <= 0) return block.text.length === 0 ? block : { ...block, text: "" };
+	const units = countGraphemes(block.text);
+	if (remaining >= units) return block;
+	return { ...block, text: sliceGraphemes(block.text, remaining) };
+}
+
+function revealThinkingBlock(
+	block: Extract<AssistantContentBlock, { type: "thinking" }>,
+	remaining: number,
+): AssistantContentBlock {
+	if (remaining <= 0) return block.thinking.length === 0 ? block : { ...block, thinking: "" };
+	const units = countGraphemes(block.thinking);
+	if (remaining >= units) return block;
+	return { ...block, thinking: sliceGraphemes(block.thinking, remaining) };
+}
+
+export function buildDisplayMessage(
+	target: AssistantMessage,
+	revealed: number,
+	hideThinking: boolean,
+): AssistantMessage {
+	let remaining = Math.max(0, Math.floor(revealed));
+	const content: AssistantContentBlock[] = [];
+	for (const block of target.content) {
+		if (block.type === "text") {
+			content.push(revealTextBlock(block, remaining));
+			remaining = Math.max(0, remaining - countGraphemes(block.text));
+		} else if (block.type === "thinking" && !hideThinking) {
+			content.push(revealThinkingBlock(block, remaining));
+			remaining = Math.max(0, remaining - countGraphemes(block.thinking));
+		} else {
+			content.push(block);
+		}
+	}
+	return { ...target, content };
+}
+
+export function nextStep(backlog: number): number {
+	return Math.max(MIN_STEP, Math.ceil(Math.max(0, backlog) / CATCHUP_FRAMES));
+}
+
+export class StreamingRevealController {
+	readonly #getSmoothStreaming: () => boolean;
+	readonly #getHideThinkingBlock: () => boolean;
+	readonly #requestRender: () => void;
+	#target: AssistantMessage | undefined;
+	#component: StreamingRevealComponent | undefined;
+	#timer: NodeJS.Timeout | undefined;
+	#revealed = 0;
+	#hideThinkingBlock = false;
+	#smoothStreaming = true;
+
+	constructor(options: StreamingRevealControllerOptions) {
+		this.#getSmoothStreaming = options.getSmoothStreaming;
+		this.#getHideThinkingBlock = options.getHideThinkingBlock;
+		this.#requestRender = options.requestRender;
+	}
+
+	begin(component: StreamingRevealComponent, message: AssistantMessage): void {
+		this.stop();
+		this.#component = component;
+		this.#target = message;
+		this.#revealed = 0;
+		this.#hideThinkingBlock = this.#getHideThinkingBlock();
+		this.#smoothStreaming = this.#getSmoothStreaming();
+		if (!this.#smoothStreaming) {
+			component.updateContent(message);
+			return;
+		}
+		const total = visibleUnits(message, this.#hideThinkingBlock);
+		if (message.content.some(block => block.type === "toolCall")) {
+			// A tool call is a transcript-order boundary: finish any leading
+			// assistant text before EventController renders the separate tool card.
+			this.#revealed = total;
+			component.updateContent(buildDisplayMessage(message, this.#revealed, this.#hideThinkingBlock));
+			return;
+		}
+		this.#renderCurrent();
+		this.#syncTimer(total);
+	}
+
+	setTarget(message: AssistantMessage): void {
+		this.#target = message;
+		if (!this.#component) return;
+		if (!this.#smoothStreaming) {
+			this.#component.updateContent(message);
+			return;
+		}
+		const total = visibleUnits(message, this.#hideThinkingBlock);
+		if (message.content.some(block => block.type === "toolCall")) {
+			// A tool call is a transcript-order boundary: finish any leading
+			// assistant text before EventController renders the separate tool card.
+			this.#revealed = total;
+			this.#stopTimer();
+			this.#component.updateContent(buildDisplayMessage(message, this.#revealed, this.#hideThinkingBlock));
+			return;
+		}
+		if (this.#revealed > total) {
+			this.#revealed = total;
+		}
+		this.#renderCurrent();
+		this.#syncTimer(total);
+	}
+
+	stop(): void {
+		this.#stopTimer();
+		this.#target = undefined;
+		this.#component = undefined;
+		this.#revealed = 0;
+	}
+
+	#renderCurrent(): void {
+		if (!this.#target || !this.#component) return;
+		this.#component.updateContent(buildDisplayMessage(this.#target, this.#revealed, this.#hideThinkingBlock));
+	}
+
+	#syncTimer(total = this.#target ? visibleUnits(this.#target, this.#hideThinkingBlock) : 0): void {
+		if (!this.#target || !this.#component || this.#revealed >= total) {
+			this.#stopTimer();
+			return;
+		}
+		this.#startTimer();
+	}
+
+	#startTimer(): void {
+		if (this.#timer) return;
+		this.#timer = setInterval(() => {
+			this.#tick();
+		}, STREAMING_REVEAL_FRAME_MS);
+		this.#timer.unref?.();
+	}
+
+	#stopTimer(): void {
+		if (!this.#timer) return;
+		clearInterval(this.#timer);
+		this.#timer = undefined;
+	}
+
+	#tick(): void {
+		const target = this.#target;
+		const component = this.#component;
+		if (!target || !component) {
+			this.stop();
+			return;
+		}
+		const total = visibleUnits(target, this.#hideThinkingBlock);
+		if (this.#revealed >= total) {
+			this.#stopTimer();
+			return;
+		}
+		this.#revealed = Math.min(total, this.#revealed + nextStep(total - this.#revealed));
+		component.updateContent(buildDisplayMessage(target, this.#revealed, this.#hideThinkingBlock));
+		this.#requestRender();
+		if (this.#revealed >= total) {
+			this.#stopTimer();
+		}
+	}
+}
diff --git a/packages/coding-agent/src/modes/controllers/tan-command-controller.ts b/packages/coding-agent/src/modes/controllers/tan-command-controller.ts
new file mode 100644
index 000000000..b9d71b71d
--- /dev/null
+++ b/packages/coding-agent/src/modes/controllers/tan-command-controller.ts
@@ -0,0 +1,173 @@
+import * as fs from "node:fs/promises";
+import type { AssistantMessage } from "@oh-my-pi/pi-ai";
+import { prompt, Snowflake } from "@oh-my-pi/pi-utils";
+import backgroundTanDispatchPrompt from "../../prompts/system/background-tan-dispatch.md" with { type: "text" };
+import { AgentRegistry, MAIN_AGENT_ID } from "../../registry/agent-registry";
+import * as sdk from "../../sdk";
+import type { AgentSession } from "../../session/agent-session";
+import { SessionManager } from "../../session/session-manager";
+import { createMCPProxyTools, createSubagentSettings } from "../../task/executor";
+import type { InteractiveModeContext } from "../types";
+
+const TAN_LABEL_PREVIEW_LENGTH = 80;
+
+function previewWork(work: string): string {
+	const singleLine = work.trim().replace(/\s+/g, " ");
+	if (singleLine.length <= TAN_LABEL_PREVIEW_LENGTH) return singleLine;
+	return `${singleLine.slice(0, TAN_LABEL_PREVIEW_LENGTH - 1)}…`;
+}
+
+function extractAssistantText(message: AssistantMessage | undefined): string {
+	if (!message) return "";
+	return message.content
+		.filter(content => content.type === "text")
+		.map(content => content.text)
+		.join("")
+		.trim();
+}
+
+async function removeCloneSession(cloneFile: string): Promise<void> {
+	await Promise.allSettled([
+		fs.rm(cloneFile, { force: true }),
+		fs.rm(cloneFile.slice(0, -6), { recursive: true, force: true }),
+	]);
+}
+
+export class TanCommandController {
+	constructor(private readonly ctx: InteractiveModeContext) {}
+
+	async start(work: string): Promise<void> {
+		const trimmedWork = work.trim();
+		if (!trimmedWork) {
+			this.ctx.showStatus("Usage: /tan <work>");
+			return;
+		}
+
+		const session = this.ctx.session;
+		if (session.isStreaming) {
+			this.ctx.showWarning("Wait for the current response to finish or abort it before using /tan.");
+			return;
+		}
+
+		const model = session.model;
+		if (!model) {
+			this.ctx.showError("No active model available for /tan.");
+			return;
+		}
+
+		const manager = session.asyncJobManager;
+		if (!manager) {
+			this.ctx.showError("Background jobs are disabled; enable async jobs to use /tan.");
+			return;
+		}
+
+		const parentFile = this.ctx.sessionManager.getSessionFile();
+		if (!parentFile) {
+			this.ctx.showError("/tan requires a persisted session.");
+			return;
+		}
+
+		const parentSessionId = session.sessionId;
+		const thinkingLevel = session.configuredThinkingLevel();
+		const systemPrompt = [...session.systemPrompt];
+		const toolNames = session.getActiveToolNames();
+		const modelRegistry = session.modelRegistry;
+		const ownerId = session.getAgentId() ?? MAIN_AGENT_ID;
+		const mcpManager = this.ctx.mcpManager;
+		const cwd = this.ctx.sessionManager.getCwd();
+		// Nest the clone inside the parent's artifact directory (like a subagent
+		// session) rather than as a top-level sibling, so it shares the parent's
+		// artifacts in place — no copy needed.
+		const sessionDir = parentFile.slice(0, -6);
+		const settings = createSubagentSettings(this.ctx.settings);
+		const customTools = mcpManager ? createMCPProxyTools(mcpManager) : undefined;
+		const enableLsp = this.ctx.settings.get("task.enableLsp") !== false;
+		const agentRegistry = AgentRegistry.global();
+		const cloneId = `Tan-${Snowflake.next()}`;
+		const label = `/tan ${previewWork(trimmedWork)}`;
+
+		await this.ctx.sessionManager.ensureOnDisk();
+		await this.ctx.sessionManager.flush();
+
+		let cloneFile = "";
+		let jobId = "";
+		try {
+			const cloneManager = await SessionManager.forkFrom(parentFile, cwd, sessionDir, undefined, {
+				suppressBreadcrumb: true,
+			});
+			cloneFile = cloneManager.getSessionFile() ?? "";
+			if (!cloneFile) throw new Error("Forked session did not create a session file.");
+
+			jobId = manager.register(
+				"task",
+				label,
+				async ({ signal }) => {
+					if (signal.aborted) throw new Error("Aborted before execution");
+
+					let clone: AgentSession | undefined;
+					try {
+						const created = await sdk.createAgentSession({
+							cwd,
+							sessionManager: cloneManager,
+							model,
+							thinkingLevel,
+							systemPrompt,
+							toolNames,
+							providerSessionId: `${parentSessionId}:tan:${Snowflake.next()}`,
+							providerPromptCacheKey: parentSessionId,
+							modelRegistry,
+							authStorage: modelRegistry.authStorage,
+							settings,
+							hasUI: false,
+							enableMCP: false,
+							customTools,
+							enableLsp,
+							agentId: cloneId,
+							agentDisplayName: "tan",
+							parentTaskPrefix: cloneId,
+							agentRegistry,
+							disableExtensionDiscovery: true,
+						});
+						clone = created.session;
+						const abortClone = () => {
+							void clone?.abort();
+						};
+						signal.addEventListener("abort", abortClone, { once: true });
+						try {
+							if (signal.aborted) {
+								abortClone();
+								throw new Error("Aborted before execution");
+							}
+							await clone.prompt(trimmedWork, { attribution: "user" });
+							await clone.waitForIdle();
+							return extractAssistantText(clone.getLastAssistantMessage()) || "(no output)";
+						} finally {
+							signal.removeEventListener("abort", abortClone);
+						}
+					} finally {
+						await clone?.dispose();
+					}
+				},
+				{ ownerId },
+			);
+		} catch (error) {
+			if (cloneFile) await removeCloneSession(cloneFile);
+			this.ctx.showError(error instanceof Error ? error.message : String(error));
+			return;
+		}
+
+		const content = prompt.render(backgroundTanDispatchPrompt, { jobId, work: trimmedWork });
+		await session.sendCustomMessage(
+			{
+				customType: "background-tan-dispatch",
+				content,
+				display: true,
+				attribution: "user",
+				details: { jobId, work: trimmedWork, sessionFile: cloneFile },
+			},
+			{ triggerTurn: false },
+		);
+		this.ctx.rebuildChatFromMessages();
+		this.ctx.showStatus(`Dispatched background tan ${jobId}`);
+	}
+}
diff --git a/packages/coding-agent/src/modes/index.ts b/packages/coding-agent/src/modes/index.ts
index 9b2726d68..ac9f12896 100644
--- a/packages/coding-agent/src/modes/index.ts
+++ b/packages/coding-agent/src/modes/index.ts
@@ -2,11 +2,13 @@ import { emergencyTerminalRestore } from "@oh-my-pi/pi-tui";
 import { postmortem } from "@oh-my-pi/pi-utils";
 
 /**
- * Run modes for the coding agent.
+ * Interactive mode and embeddable RPC client exports for the coding agent.
+ *
+ * Branch-specific runners live in their concrete modules so importing this
+ * barrel does not pull print, RPC server, or ACP server mode into the normal
+ * TUI graph.
  */
-export { runAcpMode } from "./acp";
 export { InteractiveMode, type InteractiveModeOptions } from "./interactive-mode";
-export { type PrintModeOptions, runPrintMode } from "./print-mode";
 export {
 	defineRpcClientTool,
 	type ModelInfo,
@@ -17,7 +19,6 @@ export {
 	type RpcClientToolResult,
 	type RpcEventListener,
 } from "./rpc/rpc-client";
-export { runRpcMode } from "./rpc/rpc-mode";
 export type {
 	RpcCommand,
 	RpcHostToolCallRequest,
diff --git a/packages/coding-agent/src/modes/interactive-mode.ts b/packages/coding-agent/src/modes/interactive-mode.ts
index c1659f1ac..0b0e42641 100644
--- a/packages/coding-agent/src/modes/interactive-mode.ts
+++ b/packages/coding-agent/src/modes/interactive-mode.ts
@@ -20,7 +20,7 @@ import {
 	modelsAreEqual,
 	type UsageReport,
 } from "@oh-my-pi/pi-ai";
-import type { Component, EditorTheme, SlashCommand } from "@oh-my-pi/pi-tui";
+import type { Component, EditorTheme, LoaderMessageColorFn, OverlayHandle, SlashCommand } from "@oh-my-pi/pi-tui";
 import {
 	Container,
 	clearRenderCache,
@@ -65,12 +65,7 @@ import { BUILTIN_SLASH_COMMANDS, loadSlashCommands } from "../extensibility/slas
 import type { Goal, GoalModeState } from "../goals/state";
 import { resolveLocalUrlToPath } from "../internal-urls";
 import { LSP_STARTUP_EVENT_CHANNEL, type LspStartupEvent } from "../lsp/startup-events";
-import {
-	humanizePlanTitle,
-	type PlanApprovalDetails,
-	renameApprovedPlanFile,
-	resolvePlanTitle,
-} from "../plan-mode/approved-plan";
+import { humanizePlanTitle, type PlanApprovalDetails, resolveApprovedPlan } from "../plan-mode/approved-plan";
 import planModeApprovedPrompt from "../prompts/system/plan-mode-approved.md" with { type: "text" };
 import planModeCompactInstructionsPrompt from "../prompts/system/plan-mode-compact-instructions.md" with {
 	type: "text",
@@ -94,6 +89,7 @@ import { getSessionAccentAnsi, getSessionAccentHex } from "../utils/session-colo
 import { popTerminalTitle, pushTerminalTitle, setSessionTerminalTitle } from "../utils/title-generator";
 import type { AssistantMessageComponent } from "./components/assistant-message";
 import type { BashExecutionComponent } from "./components/bash-execution";
+import { ChatBlock, type ChatBlockHost } from "./components/chat-block";
 import { CustomEditor } from "./components/custom-editor";
 import { DynamicBorder } from "./components/dynamic-border";
 import { ErrorBannerComponent } from "./components/error-banner";
@@ -101,6 +97,7 @@ import type { EvalExecutionComponent } from "./components/eval-execution";
 import type { HookEditorComponent } from "./components/hook-editor";
 import type { HookInputComponent } from "./components/hook-input";
 import type { HookSelectorComponent, HookSelectorSlider } from "./components/hook-selector";
+import { PlanReviewOverlay } from "./components/plan-review-overlay";
 import { StatusLineComponent } from "./components/status-line";
 import type { ToolExecutionHandle } from "./components/tool-execution";
 import { TranscriptContainer } from "./components/transcript-container";
@@ -114,6 +111,7 @@ import { MCPCommandController } from "./controllers/mcp-command-controller";
 import { OmfgController } from "./controllers/omfg-controller";
 import { SelectorController } from "./controllers/selector-controller";
 import { SSHCommandController } from "./controllers/ssh-command-controller";
+import { TanCommandController } from "./controllers/tan-command-controller";
 import { TodoCommandController } from "./controllers/todo-command-controller";
 import {
 	consumeLoopLimitIteration,
@@ -273,7 +271,6 @@ export class InteractiveMode implements InteractiveModeContext {
 	statusLine: StatusLineComponent;
 
 	isInitialized = false;
-	isBackgrounded = false;
 	isBashMode = false;
 	toolOutputExpanded = false;
 	todoExpanded = false;
@@ -341,12 +338,14 @@ export class InteractiveMode implements InteractiveModeContext {
 	#planModePreviousModelState: { model: Model; thinkingLevel?: ThinkingLevel } | undefined;
 	#pendingModelSwitch: { model: Model; thinkingLevel?: ThinkingLevel } | undefined;
 	#planModeHasEntered = false;
-	#planReviewContainer: Container | undefined;
+	#planReviewOverlay: PlanReviewOverlay | undefined;
+	#planReviewOverlayHandle: OverlayHandle | undefined;
 	readonly lspServers: LspStartupServerInfo[] | undefined = undefined;
 	mcpManager?: import("../mcp").MCPManager;
 	readonly #toolUiContextSetter: (uiContext: ExtensionUIContext, hasUI: boolean) => void;
 
 	readonly #btwController: BtwController;
+	readonly #tanCommandController: TanCommandController;
 	readonly #omfgController: OmfgController;
 	readonly #commandController: CommandController;
 	readonly #todoCommandController: TodoCommandController;
@@ -365,6 +364,7 @@ export class InteractiveMode implements InteractiveModeContext {
 	#eventBus?: EventBus;
 	#eventBusUnsubscribers: Array<() => void> = [];
 	#welcomeComponent?: WelcomeComponent;
+	readonly #chatHost: ChatBlockHost = { requestRender: () => this.ui.requestRender() };
 
 	constructor(
 		session: AgentSession,
@@ -470,6 +470,7 @@ export class InteractiveMode implements InteractiveModeContext {
 
 		this.#uiHelpers = new UiHelpers(this);
 		this.#btwController = new BtwController(this);
+		this.#tanCommandController = new TanCommandController(this);
 		this.#omfgController = new OmfgController(this);
 		this.#extensionUiController = new ExtensionUiController(this);
 		this.#eventController = new EventController(this);
@@ -599,8 +600,9 @@ export class InteractiveMode implements InteractiveModeContext {
 		// Load initial todos
 		await this.#loadTodoList();
 
-		// Start the UI
-		this.ui.start();
+		// Start the UI. Cold `omp` launch opts into clearing on the first paint so
+		// the initial welcome frame does not append over the previous run's scrollback.
+		this.ui.start({ clearScrollback: options.clearInitialTerminalHistory === true });
 		pushTerminalTitle();
 		setSessionTerminalTitle(this.sessionManager.getSessionName(), this.sessionManager.getCwd());
 		this.updateEditorBorderColor();
@@ -1522,22 +1524,15 @@ export class InteractiveMode implements InteractiveModeContext {
 				if (!state?.enabled) {
 					throw new ToolError("Plan mode is not active.");
 				}
-				const planFilePath = state.planFilePath;
-				const planContent = await this.#readPlanFile(planFilePath);
-				if (planContent === null) {
-					throw new ToolError(
-						`Plan file not found at ${planFilePath}. Write the finalized plan to ${planFilePath} before requesting approval.`,
-					);
-				}
-				const normalized = resolvePlanTitle({
+				const { planFilePath, title } = await resolveApprovedPlan({
 					suppliedTitle: extra?.title,
-					planContent,
-					planFilePath,
+					statePlanFilePath: state.planFilePath,
+					readPlan: url => this.#readPlanFile(url),
+					listPlanFiles: () => this.#listLocalPlanFiles(),
 				});
 				const details: PlanApprovalDetails = {
 					planFilePath,
-					finalPlanFilePath: `local://${normalized.fileName}`,
-					title: normalized.title,
+					title,
 					planExists: true,
 				};
 				return {
@@ -1677,22 +1672,87 @@ export class InteractiveMode implements InteractiveModeContext {
 		}
 	}
 
-	#renderPlanPreview(planContent: string, options?: { append?: boolean }): void {
-		const existingContainer = this.#planReviewContainer;
-		const replaceExisting = options?.append !== true && existingContainer !== undefined;
-		const planReviewContainer = replaceExisting ? existingContainer : new Container();
-		planReviewContainer.clear();
-		planReviewContainer.addChild(new Spacer(1));
-		planReviewContainer.addChild(new DynamicBorder());
-		planReviewContainer.addChild(new Text(theme.bold(theme.fg("accent", "Plan Review")), 1, 1));
-		planReviewContainer.addChild(new Spacer(1));
-		planReviewContainer.addChild(new Markdown(planContent, 1, 1, getMarkdownTheme()));
-		planReviewContainer.addChild(new DynamicBorder());
-		if (!replaceExisting) {
-			this.chatContainer.addChild(planReviewContainer);
+	/** `local://` URLs of plan files in the session-local root, newest first.
+	 *  A fallback for `resolveApprovedPlan` when the agent dropped `extra.title`,
+	 *  so the plan it wrote is still found by scanning recent `*-plan.md` files. */
+	async #listLocalPlanFiles(): Promise<string[]> {
+		const localRoot = this.#resolvePlanFilePath("local://");
+		try {
+			const entries = await fs.readdir(localRoot, { withFileTypes: true });
+			const plans = await Promise.all(
+				entries
+					.filter(entry => entry.isFile() && /plan\.md$/i.test(entry.name))
+					.map(async name => {
+						const stat = await fs.stat(path.join(localRoot, name.name)).catch(() => null);
+						return { url: `local://${name.name}`, mtime: stat?.mtimeMs ?? 0 };
+					}),
+			);
+			return plans.sort((a, b) => b.mtime - a.mtime).map(plan => plan.url);
+		} catch {
+			return [];
 		}
-		this.#planReviewContainer = planReviewContainer;
+	}
+
+	showPlanReview(
+		planContent: string,
+		title: string,
+		options: string[],
+		dialogOptions?: {
+			helpText?: string;
+			disabledIndices?: number[];
+			onExternalEditor?: () => void;
+			onPlanEdited?: (content: string) => void;
+			onFeedbackChange?: (feedback: string) => void;
+			initialIndex?: number;
+		},
+		extra?: { slider?: HookSelectorSlider },
+	): Promise<string | undefined> {
+		this.#hidePlanReview();
+		const { promise, resolve } = Promise.withResolvers<string | undefined>();
+		let settled = false;
+		const finish = (choice: string | undefined): void => {
+			if (settled) return;
+			settled = true;
+			this.#hidePlanReview();
+			this.ui.requestRender();
+			resolve(choice);
+		};
+		const overlay = new PlanReviewOverlay(
+			planContent,
+			{
+				promptTitle: title,
+				options,
+				disabledIndices: dialogOptions?.disabledIndices,
+				helpText: dialogOptions?.helpText,
+				initialIndex: dialogOptions?.initialIndex,
+				slider: extra?.slider,
+				externalEditorLabel: this.keybindings.getDisplayString("app.editor.external") || undefined,
+			},
+			{
+				onPick: choice => finish(choice),
+				onCancel: () => finish(undefined),
+				onExternalEditor: dialogOptions?.onExternalEditor,
+				onPlanEdited: dialogOptions?.onPlanEdited,
+				onFeedbackChange: dialogOptions?.onFeedbackChange,
+			},
+		);
+		this.#planReviewOverlay = overlay;
+		this.#planReviewOverlayHandle = this.ui.showOverlay(overlay, {
+			anchor: "bottom-center",
+			width: "100%",
+			maxHeight: "100%",
+			margin: 0,
+			fullscreen: true,
+		});
+		this.ui.setFocus(overlay);
 		this.ui.requestRender();
+		return promise;
+	}
+
+	#hidePlanReview(): void {
+		this.#planReviewOverlayHandle?.hide();
+		this.#planReviewOverlayHandle = undefined;
+		this.#planReviewOverlay = undefined;
 	}
 
 	#getEditorTerminalPath(): string | null {
@@ -1714,14 +1774,6 @@ export class InteractiveMode implements InteractiveModeContext {
 		}
 	}
 
-	#getPlanReviewHelpText(): string {
-		const externalEditorKey = this.keybindings.getDisplayString("app.editor.external");
-		if (!externalEditorKey) {
-			return "up/down navigate  enter select  esc cancel";
-		}
-		return `up/down navigate  enter select  ${externalEditorKey.toLowerCase()} open in editor  esc cancel`;
-	}
-
 	#getPlanApprovalContextUsage(): ContextUsage | undefined {
 		const executionModel = this.#planModePreviousModelState?.model ?? this.session.model;
 		const contextWindow = executionModel?.contextWindow;
@@ -1780,7 +1832,7 @@ export class InteractiveMode implements InteractiveModeContext {
 			});
 			if (result !== null) {
 				await Bun.write(resolvedPath, result);
-				this.#renderPlanPreview(result);
+				this.#planReviewOverlay?.setPlanContent(result);
 				this.showStatus("Plan updated in external editor.");
 			}
 		} catch (error) {
@@ -1812,19 +1864,12 @@ export class InteractiveMode implements InteractiveModeContext {
 		planContent: string,
 		options: {
 			planFilePath: string;
-			finalPlanFilePath: string;
 			title: string;
 			preserveContext?: boolean;
 			compactBeforeExecute?: boolean;
 			executionModel?: ResolvedRoleModel;
 		},
 	): Promise<void> {
-		await renameApprovedPlanFile({
-			planFilePath: options.planFilePath,
-			finalPlanFilePath: options.finalPlanFilePath,
-			getArtifactsDir: () => this.sessionManager.getArtifactsDir(),
-			getSessionId: () => this.sessionManager.getSessionId(),
-		});
 		const previousTools = this.#planModePreviousTools ?? this.session.getActiveToolNames();
 
 		// Mark the pending abort caused by the plan-mode → compaction transition as
@@ -1843,8 +1888,8 @@ export class InteractiveMode implements InteractiveModeContext {
 			if (!options.preserveContext) {
 				await this.handleClearCommand();
 				// The new session has a fresh local:// root — persist the approved plan there
-				// so `local://<title>.md` resolves correctly in the execution session.
-				const newLocalPath = resolveLocalUrlToPath(options.finalPlanFilePath, {
+				// so `local://<slug>-plan.md` resolves correctly in the execution session.
+				const newLocalPath = resolveLocalUrlToPath(options.planFilePath, {
 					getArtifactsDir: () => this.sessionManager.getArtifactsDir(),
 					getSessionId: () => this.sessionManager.getSessionId(),
 				});
@@ -1858,7 +1903,7 @@ export class InteractiveMode implements InteractiveModeContext {
 				// Cancellation skips the synthetic-prompt dispatch (operator's explicit
 				// abort is honored); failure proceeds best-effort — approval intent stands.
 				const compactionPrompt = prompt.render(planModeCompactInstructionsPrompt, {
-					planFilePath: options.finalPlanFilePath,
+					planFilePath: options.planFilePath,
 				});
 				// Pin the plan reference path BEFORE compaction so any user messages
 				// queued during the compaction await (which `handleCompactCommand`
@@ -1866,7 +1911,7 @@ export class InteractiveMode implements InteractiveModeContext {
 				// approved plan in `#buildPlanReferenceMessage`. Reassignment after
 				// the try/finally is idempotent and kept for the !compactBeforeExecute
 				// branch.
-				this.session.setPlanReferencePath(options.finalPlanFilePath);
+				this.session.setPlanReferencePath(options.planFilePath);
 				compactOutcome = await this.handleCompactCommand(compactionPrompt);
 			}
 		} finally {
@@ -1882,7 +1927,7 @@ export class InteractiveMode implements InteractiveModeContext {
 		if (previousTools.length > 0) {
 			await this.session.setActiveToolsByName(previousTools);
 		}
-		this.session.setPlanReferencePath(options.finalPlanFilePath);
+		this.session.setPlanReferencePath(options.planFilePath);
 
 		if (compactOutcome === "cancelled") {
 			// Explicit abort: honor it. `executeCompaction` already surfaced
@@ -1919,7 +1964,7 @@ export class InteractiveMode implements InteractiveModeContext {
 		this.session.markPlanReferenceSent();
 		const planModePrompt = prompt.render(planModeApprovedPrompt, {
 			planContent,
-			finalPlanFilePath: options.finalPlanFilePath,
+			planFilePath: options.planFilePath,
 			contextPreserved: options.preserveContext === true,
 		});
 		await this.session.prompt(planModePrompt, { synthetic: true });
@@ -2209,7 +2254,6 @@ export class InteractiveMode implements InteractiveModeContext {
 			return;
 		}
 
-		this.#renderPlanPreview(planContent, { append: true });
 		const contextUsage = this.#getPlanApprovalContextUsage();
 		const keepContextLabel = this.#formatKeepContextLabel(contextUsage);
 		const keepContextDisabled = this.#isKeepContextDisabled(contextUsage);
@@ -2239,23 +2283,40 @@ export class InteractiveMode implements InteractiveModeContext {
 						},
 					}
 				: undefined;
-		const helpText = slider ? `${this.#getPlanReviewHelpText()}  ◂/▸ model` : this.#getPlanReviewHelpText();
+		// The overlay now owns the dynamic, focus-aware help line; the caller only
+		// supplies the trailing cancel hint.
+		const helpText = "esc cancel";
+		// In-overlay edits (section deletes/undo) and section annotations. Deletes
+		// update `editedContent` (and mirror to disk); annotations build `feedback`
+		// that the Refine branch re-prompts the model with.
+		let editedContent: string | undefined;
+		let feedback = "";
 
-		const choice = await this.showHookSelector(
+		const choice = await this.showPlanReview(
+			planContent,
 			"Plan mode - next step",
 			["Approve and execute", "Approve and compact context", keepContextLabel, "Refine plan"],
 			{
 				helpText,
 				onExternalEditor: () => void this.#openPlanInExternalEditor(planFilePath),
+				onPlanEdited: content => {
+					editedContent = content;
+					void Bun.write(this.#resolvePlanFilePath(planFilePath), content);
+				},
+				onFeedbackChange: value => {
+					feedback = value;
+				},
 				disabledIndices: keepContextDisabled ? [PLAN_KEEP_CONTEXT_OPTION_INDEX] : undefined,
 			},
 			{ slider },
 		);
 
 		if (choice === "Approve and execute" || choice === "Approve and compact context" || choice === keepContextLabel) {
-			const finalPlanFilePath = details.finalPlanFilePath || planFilePath;
 			try {
-				const latestPlanContent = await this.#readPlanFile(planFilePath);
+				// Prefer in-overlay edits (already in memory) over a disk re-read; the
+				// `onPlanEdited` write is fire-and-forget, so reading the file here could
+				// race ahead of it.
+				const latestPlanContent = editedContent ?? (await this.#readPlanFile(planFilePath));
 				if (!latestPlanContent) {
 					this.showError(`Plan file not found at ${planFilePath}`);
 					return;
@@ -2273,7 +2334,6 @@ export class InteractiveMode implements InteractiveModeContext {
 					cycle && selectedTierIndex !== cycle.currentIndex ? cycle.models[selectedTierIndex] : undefined;
 				await this.#approvePlan(latestPlanContent, {
 					planFilePath,
-					finalPlanFilePath,
 					title: details.title,
 					preserveContext: choice !== "Approve and execute",
 					compactBeforeExecute: choice === "Approve and compact context",
@@ -2286,6 +2346,16 @@ export class InteractiveMode implements InteractiveModeContext {
 			}
 			return;
 		}
+
+		if (choice === "Refine plan") {
+			// Section annotations entered in the overlay become a refinement prompt
+			// re-submitted to the model. With no annotations, fall back to today's
+			// behavior: close the overlay and let the operator type their own.
+			if (feedback.trim() && this.onInputCallback) {
+				this.onInputCallback(this.startPendingSubmission({ text: feedback }));
+			}
+			return;
+		}
 	}
 
 	/**
@@ -2439,9 +2509,6 @@ export class InteractiveMode implements InteractiveModeContext {
 	initializeHookRunner(uiContext: ExtensionUIContext, hasUI: boolean): void {
 		this.#extensionUiController.initializeHookRunner(uiContext, hasUI);
 	}
-	createBackgroundUiContext(): ExtensionUIContext {
-		return this.#extensionUiController.createBackgroundUiContext();
-	}
 
 	setEditorComponent(
 		factory: ((tui: TUI, theme: EditorTheme, keybindings: KeybindingsManager) => CustomEditor) | undefined,
@@ -2483,12 +2550,26 @@ export class InteractiveMode implements InteractiveModeContext {
 		this.ui.requestRender();
 	}
 
-	// Event handling
-	async handleBackgroundEvent(event: AgentSessionEvent): Promise<void> {
-		await this.#eventController.handleBackgroundEvent(event);
+	// UI helpers
+	present(content: Component | readonly Component[]): void {
+		if (Array.isArray(content)) {
+			for (const item of content) this.#mountChatChild(item);
+		} else {
+			this.#mountChatChild(content as Component);
+		}
+		this.ui.requestRender();
+	}
+
+	#mountChatChild(item: Component): void {
+		this.chatContainer.addChild(item);
+		if (item instanceof ChatBlock) item.mount(this.#chatHost);
+	}
+
+	resetTranscript(): void {
+		this.chatContainer.dispose();
+		this.chatContainer.clear();
 	}
 
-	// UI helpers
 	showStatus(message: string, options?: { dim?: boolean }): void {
 		this.#uiHelpers.showStatus(message, options);
 	}
@@ -2508,7 +2589,6 @@ export class InteractiveMode implements InteractiveModeContext {
 	}
 
 	showPinnedError(message: string): void {
-		if (this.isBackgrounded) return;
 		this.errorBannerContainer.clear();
 		this.errorBannerContainer.addChild(new ErrorBannerComponent(message));
 		this.ui.requestRender();
@@ -2579,13 +2659,18 @@ export class InteractiveMode implements InteractiveModeContext {
 	ensureLoadingAnimation(): void {
 		if (!this.loadingAnimation) {
 			this.statusContainer.clear();
+			const messageColorFn = ((message: string) =>
+				renderWorkingMessage(message, this.#getWorkingMessageAccent())) as LoaderMessageColorFn & {
+				animated: true;
+			};
+			messageColorFn.animated = true;
 			this.loadingAnimation = new Loader(
 				this.ui,
 				spinner => {
 					const accent = this.#getWorkingMessageAccent();
 					return accent ? `${accent.main}${spinner}\x1b[39m` : theme.fg("accent", spinner);
 				},
-				message => renderWorkingMessage(message, this.#getWorkingMessageAccent()),
+				messageColorFn,
 				this.#defaultWorkingMessage,
 				getSymbolTheme().spinnerFrames,
 			);
@@ -2700,10 +2785,6 @@ export class InteractiveMode implements InteractiveModeContext {
 		return this.#commandController.handleShareCommand();
 	}
 
-	handleCopyCommand(sub?: string) {
-		return this.#commandController.handleCopyCommand(sub);
-	}
-
 	handleTodoCommand(args: string): Promise<void> {
 		return this.#todoCommandController.handleTodoCommand(args);
 	}
@@ -2741,7 +2822,7 @@ export class InteractiveMode implements InteractiveModeContext {
 		this.#omfgController.dispose();
 		this.#extensionUiController.clearExtensionTerminalInputListeners();
 		this.clearPinnedError();
-		this.#planReviewContainer = undefined;
+		this.#hidePlanReview();
 	}
 
 	handleClearCommand(): Promise<void> {
@@ -2749,6 +2830,10 @@ export class InteractiveMode implements InteractiveModeContext {
 		return this.#commandController.handleClearCommand();
 	}
 
+	handleFreshCommand(): Promise<void> {
+		return this.#commandController.handleFreshCommand();
+	}
+
 	handleDropCommand(): Promise<void> {
 		this.#prepareSessionSwitch();
 		return this.#commandController.handleDropCommand();
@@ -2848,8 +2933,8 @@ export class InteractiveMode implements InteractiveModeContext {
 		}
 	}
 
-	showDebugSelector(): void {
-		this.#selectorController.showDebugSelector();
+	async showDebugSelector(): Promise<void> {
+		await this.#selectorController.showDebugSelector();
 	}
 
 	showSessionObserver(): void {
@@ -2936,6 +3021,10 @@ export class InteractiveMode implements InteractiveModeContext {
 		this.#selectorController.showUserMessageSelector();
 	}
 
+	showCopySelector(): void {
+		this.#selectorController.showCopySelector();
+	}
+
 	showTreeSelector(): void {
 		this.#selectorController.showTreeSelector();
 	}
@@ -2980,10 +3069,6 @@ export class InteractiveMode implements InteractiveModeContext {
 		this.#inputController.handleDequeue();
 	}
 
-	handleBackgroundCommand(): void {
-		this.#inputController.handleBackgroundCommand();
-	}
-
 	handleImagePaste(): Promise<boolean> {
 		return this.#inputController.handleImagePaste();
 	}
@@ -2992,6 +3077,10 @@ export class InteractiveMode implements InteractiveModeContext {
 		return this.#btwController.start(question);
 	}
 
+	handleTanCommand(work: string): Promise<void> {
+		return this.#tanCommandController.start(work);
+	}
+
 	hasActiveBtw(): boolean {
 		return this.#btwController.hasActiveRequest();
 	}
diff --git a/packages/coding-agent/src/modes/setup-version.ts b/packages/coding-agent/src/modes/setup-version.ts
new file mode 100644
index 000000000..33bed5051
--- /dev/null
+++ b/packages/coding-agent/src/modes/setup-version.ts
@@ -0,0 +1,11 @@
+/**
+ * Setup version the wizard advances a fresh install to. Bump it whenever a new
+ * setup scene lands (or an existing scene raises its `minVersion`).
+ *
+ * Kept in its own dependency-free module so the cold-launch gate in `main.ts`
+ * can answer "is the stored setup version stale?" without statically importing
+ * the full wizard — every scene (sign-in/OAuth, web search, theme previews) plus
+ * the overlay component and their TUI deps. MUST equal `max(scene.minVersion)`
+ * across `ALL_SCENES`; the `setup-wizard` barrel and test suite guard it.
+ */
+export const CURRENT_SETUP_VERSION = 1;
diff --git a/packages/coding-agent/src/modes/setup-wizard/index.ts b/packages/coding-agent/src/modes/setup-wizard/index.ts
index 29da5d873..5e5eea61d 100644
--- a/packages/coding-agent/src/modes/setup-wizard/index.ts
+++ b/packages/coding-agent/src/modes/setup-wizard/index.ts
@@ -1,4 +1,5 @@
 import type { Settings } from "../../config/settings";
+import { CURRENT_SETUP_VERSION } from "../setup-version";
 import type { InteractiveModeContext } from "../types";
 import { glyphSetupScene } from "./scenes/glyph";
 import { providersSetupScene } from "./scenes/providers";
@@ -8,14 +9,14 @@ import { SetupWizardComponent } from "./wizard-overlay";
 
 export type { SetupScene, SetupSceneController, SetupSceneHost, SetupSceneResult } from "./scenes/types";
 
+export { CURRENT_SETUP_VERSION };
+
 export const ALL_SCENES = [
 	providersSetupScene,
 	glyphSetupScene,
 	themeSetupScene,
 ] as const satisfies readonly SetupScene[];
 
-export const CURRENT_SETUP_VERSION = ALL_SCENES.reduce((max, scene) => Math.max(max, scene.minVersion), 0);
-
 export interface SetupSceneSelectionOptions {
 	resuming?: boolean;
 	isTTY?: boolean;
diff --git a/packages/coding-agent/src/modes/setup-wizard/scenes/web-search.ts b/packages/coding-agent/src/modes/setup-wizard/scenes/web-search.ts
index 906803bfe..221da8b4a 100644
--- a/packages/coding-agent/src/modes/setup-wizard/scenes/web-search.ts
+++ b/packages/coding-agent/src/modes/setup-wizard/scenes/web-search.ts
@@ -19,7 +19,8 @@ type Availability = "checking" | boolean;
 /**
  * "Web search" panel: picks the provider the web_search tool should prefer and
  * reports whether the highlighted provider is ready to use given current
- * credentials (env keys or OAuth sign-ins from the Sign in tab).
+ * credentials (env keys or OAuth sign-ins from the Sign in tab) or an
+ * unauthenticated fallback.
  */
 export class WebSearchTab implements SetupTab {
 	readonly id = "web-search";
@@ -91,7 +92,7 @@ export class WebSearchTab implements SetupTab {
 			let ready = false;
 			try {
 				const provider = await getSearchProvider(id);
-				ready = await provider.isAvailable(this.host.ctx.session.modelRegistry.authStorage);
+				ready = await provider.isExplicitlyAvailable(this.host.ctx.session.modelRegistry.authStorage);
 			} catch {
 				ready = false;
 			}
diff --git a/packages/coding-agent/src/modes/setup-wizard/wizard-overlay.ts b/packages/coding-agent/src/modes/setup-wizard/wizard-overlay.ts
index 72176a870..0d39020de 100644
--- a/packages/coding-agent/src/modes/setup-wizard/wizard-overlay.ts
+++ b/packages/coding-agent/src/modes/setup-wizard/wizard-overlay.ts
@@ -82,7 +82,7 @@ export class SetupWizardComponent implements Component {
 	}
 
 	invalidate(): void {
-		this.#activeScene?.invalidate();
+		this.#activeScene?.invalidate?.();
 	}
 
 	handleInput(data: string): void {
diff --git a/packages/coding-agent/src/modes/theme/theme-schema.json b/packages/coding-agent/src/modes/theme/theme-schema.json
index e401ca0c1..3fd9972b4 100644
--- a/packages/coding-agent/src/modes/theme/theme-schema.json
+++ b/packages/coding-agent/src/modes/theme/theme-schema.json
@@ -406,7 +406,7 @@
 					}
 				},
 				"spinnerFrames": {
-					"description": "Override the spinner frames. Use a flat array to set both `status` and `activity`, or an object to override each independently. Frames are advanced ~12.5fps for status spinners and ~60fps for activity spinners.",
+					"description": "Override the spinner frames. Use a flat array to set both `status` and `activity`, or an object to override each independently. Frames are advanced ~12.5fps for status spinners and ~30fps for activity spinners.",
 					"oneOf": [
 						{
 							"type": "array",
diff --git a/packages/coding-agent/src/modes/theme/theme.ts b/packages/coding-agent/src/modes/theme/theme.ts
index c7605fb2a..ed0d46d24 100644
--- a/packages/coding-agent/src/modes/theme/theme.ts
+++ b/packages/coding-agent/src/modes/theme/theme.ts
@@ -12,6 +12,7 @@ import {
 import type { EditorTheme, MarkdownTheme, SelectListTheme, SymbolTheme } from "@oh-my-pi/pi-tui";
 import { adjustHsv, colorLuma, getCustomThemesDir, isEnoent, logger, relativeLuminance } from "@oh-my-pi/pi-utils";
 import chalk from "chalk";
+import { LRUCache } from "lru-cache/raw";
 import * as z from "zod/v4";
 // Embed theme JSON files at build time
 import darkThemeJson from "./dark.json" with { type: "json" };
@@ -94,6 +95,7 @@ export type SymbolKey =
 	| "icon.pause"
 	| "icon.loop"
 	| "icon.folder"
+	| "icon.search"
 	| "icon.scratchFolder"
 	| "icon.file"
 	| "icon.git"
@@ -263,6 +265,7 @@ const UNICODE_SYMBOLS: SymbolMap = {
 	"icon.pause": "⏸",
 	"icon.loop": "↻",
 	"icon.folder": "📁",
+	"icon.search": "🔍",
 	"icon.scratchFolder": "🗑",
 	"icon.file": "📄",
 	"icon.git": "⎇",
@@ -487,6 +490,7 @@ const NERD_SYMBOLS: SymbolMap = {
 	"icon.loop": "\uf021",
 	// pick:  | alt:  
 	"icon.folder": "\uf115",
+	"icon.search": "\uf002",
 	// pick:  | alt:
 	"icon.scratchFolder": "\uf014",
 	// pick:  | alt:  
@@ -700,6 +704,7 @@ const ASCII_SYMBOLS: SymbolMap = {
 	"icon.pause": "||",
 	"icon.loop": "loop",
 	"icon.folder": "[D]",
+	"icon.search": "[/]",
 	"icon.scratchFolder": "[T]",
 	"icon.file": "[F]",
 	"icon.git": "git:",
@@ -2429,17 +2434,54 @@ function getHighlightColors(t: Theme): NativeHighlightColors {
 	return cachedHighlightColors;
 }
 
+/**
+ * Memoized native syntax highlight. Returns the joined ANSI string, or `null`
+ * when the native tokenizer throws so callers can apply their own fallback.
+ *
+ * Keyed on `(lang, code)` and reset whenever the active `theme` instance
+ * changes — the ANSI colors are baked into the highlighted output, so a theme
+ * switch (which always reassigns `theme`) must invalidate every entry.
+ *
+ * Why this exists: animated tool blocks (eval/bash) repaint their box on every
+ * ~33ms border-shimmer frame, and markdown re-lexes on every streamed delta.
+ * Without memoization each frame can re-tokenize an unchanged code body through
+ * the Rust FFI — ~26ms for 100 lines, ~40ms for 150 — consuming or overrunning
+ * the 33ms frame budget and starving the spinner/render timers (the "TUI freeze").
+ */
+const HIGHLIGHT_CACHE_MAX = 256;
+const highlightCache = new LRUCache<string, string>({ max: HIGHLIGHT_CACHE_MAX });
+let highlightCacheTheme: Theme | undefined;
+
+function highlightCached(code: string, validLang: string | undefined): string | null {
+	if (highlightCacheTheme !== theme) {
+		highlightCache.clear();
+		highlightCacheTheme = theme;
+	}
+	const key = `${validLang ?? ""}\x00${code}`;
+	const hit = highlightCache.get(key);
+	if (hit !== undefined) {
+		return hit;
+	}
+	let highlighted: string;
+	try {
+		highlighted = nativeHighlightCode(code, validLang, getHighlightColors(theme));
+	} catch {
+		return null;
+	}
+	highlightCache.set(key, highlighted);
+	return highlighted;
+}
+
 /**
  * Highlight code with syntax coloring based on file extension or language.
  * Returns array of highlighted lines.
  */
 export function highlightCode(code: string, lang?: string): string[] {
 	const validLang = lang && nativeSupportsLanguage(lang) ? lang : undefined;
-	try {
-		return nativeHighlightCode(code, validLang, getHighlightColors(theme)).split("\n");
-	} catch {
-		return code.split("\n");
-	}
+	const highlighted = highlightCached(code, validLang);
+	// Always return a fresh array: callers (e.g. renderCodeCell) push extra lines
+	// onto the result, which would corrupt the cached string otherwise.
+	return (highlighted ?? code).split("\n");
 }
 
 export function getSymbolTheme(): SymbolTheme {
@@ -2484,11 +2526,9 @@ export function getMarkdownTheme(): MarkdownTheme {
 		resolveMermaidAscii,
 		highlightCode: (code: string, lang?: string): string[] => {
 			const validLang = lang && nativeSupportsLanguage(lang) ? lang : undefined;
-			try {
-				return nativeHighlightCode(code, validLang, getHighlightColors(theme)).split("\n");
-			} catch {
-				return code.split("\n").map(line => theme.fg("mdCodeBlock", line));
-			}
+			const highlighted = highlightCached(code, validLang);
+			if (highlighted !== null) return highlighted.split("\n");
+			return code.split("\n").map(line => theme.fg("mdCodeBlock", line));
 		},
 	};
 	cachedMarkdownTheme = markdownTheme;
diff --git a/packages/coding-agent/src/modes/types.ts b/packages/coding-agent/src/modes/types.ts
index 9116494ff..59fdce436 100644
--- a/packages/coding-agent/src/modes/types.ts
+++ b/packages/coding-agent/src/modes/types.ts
@@ -14,7 +14,7 @@ import type {
 import type { CompactOptions } from "../extensibility/extensions/types";
 import type { MCPManager } from "../mcp";
 import type { PlanApprovalDetails } from "../plan-mode/approved-plan";
-import type { AgentSession, AgentSessionEvent } from "../session/agent-session";
+import type { AgentSession } from "../session/agent-session";
 import type { HistoryStorage } from "../session/history-storage";
 import type { SessionContext, SessionManager } from "../session/session-manager";
 import type { ShakeMode } from "../session/shake-types";
@@ -63,6 +63,7 @@ export type TodoPhase = {
 
 export interface InteractiveModeInitOptions {
 	suppressWelcomeIntro?: boolean;
+	clearInitialTerminalHistory?: boolean;
 }
 
 export type InteractiveSelectorDialogOptions = ExtensionUIDialogOptions & Pick<HookSelectorOptions, "disabledIndices">;
@@ -95,7 +96,6 @@ export interface InteractiveModeContext {
 
 	// State
 	isInitialized: boolean;
-	isBackgrounded: boolean;
 	isBashMode: boolean;
 	toolOutputExpanded: boolean;
 	todoExpanded: boolean;
@@ -149,15 +149,25 @@ export interface InteractiveModeContext {
 	// Extension UI integration
 	setToolUIContext(uiContext: ExtensionUIContext, hasUI: boolean): void;
 	initializeHookRunner(uiContext: ExtensionUIContext, hasUI: boolean): void;
-	createBackgroundUiContext(): ExtensionUIContext;
 	setEditorComponent(
 		factory: ((tui: TUI, theme: EditorTheme, keybindings: KeybindingsManager) => CustomEditor) | undefined,
 	): void;
 
-	// Event handling
-	handleBackgroundEvent(event: AgentSessionEvent): Promise<void>;
-
 	// UI helpers
+	/**
+	 * Mount transcript content and repaint once. The single sink for "show this in
+	 * chat": producers build and return a `Component` (or a `ChatBlock` carrying
+	 * its own lifecycle) and hand it here instead of touching `chatContainer` /
+	 * `ui.requestRender()` directly. `ChatBlock`s are mounted (their `onMount`
+	 * runs) so their timers/subscriptions start.
+	 */
+	present(content: Component | readonly Component[]): void;
+	/**
+	 * Dispose every live block in the transcript (stopping timers/subscriptions)
+	 * and clear it. Used before a full rebuild so animated/streaming blocks do not
+	 * leak.
+	 */
+	resetTranscript(): void;
 	showStatus(message: string, options?: { dim?: boolean }): void;
 	showError(message: string): void;
 	showPinnedError(message: string): void;
@@ -222,7 +232,6 @@ export interface InteractiveModeContext {
 	// Command handling
 	handleExportCommand(text: string): Promise<void>;
 	handleShareCommand(): Promise<void>;
-	handleCopyCommand(sub?: string): void;
 	handleTodoCommand(args: string): Promise<void>;
 	handleSessionCommand(): Promise<void>;
 	handleJobsCommand(): Promise<void>;
@@ -234,6 +243,7 @@ export interface InteractiveModeContext {
 	handleDumpCommand(): void;
 	handleDebugTranscriptCommand(): Promise<void>;
 	handleClearCommand(): Promise<void>;
+	handleFreshCommand(): Promise<void>;
 	handleDropCommand(): Promise<void>;
 	handleForkCommand(): Promise<void>;
 	handleBashCommand(command: string, excludeFromContext?: boolean): Promise<void>;
@@ -263,13 +273,14 @@ export interface InteractiveModeContext {
 	showModelSelector(options?: { temporaryOnly?: boolean }): void;
 	showPluginSelector(mode?: "install" | "uninstall"): void;
 	showUserMessageSelector(): void;
+	showCopySelector(): void;
 	showTreeSelector(): void;
 	showSessionSelector(): void;
 	handleResumeSession(sessionPath: string): Promise<void>;
 	handleSessionDeleteCommand(): Promise<void>;
 	showOAuthSelector(mode: "login" | "logout", providerId?: string): Promise<void>;
 	showHookConfirm(title: string, message: string): Promise<boolean>;
-	showDebugSelector(): void;
+	showDebugSelector(): Promise<void>;
 	showSessionObserver(): void;
 	resetObserverRegistry(): void;
 
@@ -278,9 +289,9 @@ export interface InteractiveModeContext {
 	handleCtrlD(): void;
 	handleCtrlZ(): void;
 	handleDequeue(): void;
-	handleBackgroundCommand(): void;
 	handleImagePaste(): Promise<boolean>;
 	handleBtwCommand(question: string): Promise<void>;
+	handleTanCommand(work: string): Promise<void>;
 	hasActiveBtw(): boolean;
 	handleBtwEscape(): boolean;
 	handleOmfgCommand(complaint: string): Promise<void>;
diff --git a/packages/coding-agent/src/modes/utils/context-usage.ts b/packages/coding-agent/src/modes/utils/context-usage.ts
index fd93070a8..d223d9a5e 100644
--- a/packages/coding-agent/src/modes/utils/context-usage.ts
+++ b/packages/coding-agent/src/modes/utils/context-usage.ts
@@ -37,6 +37,9 @@ export interface ContextBreakdown {
 	freeTokens: number;
 }
 
+const EMPTY_STRING_PARTS: readonly string[] = [];
+const EMPTY_TOOLS: ReadonlyArray<Pick<Tool, "name" | "description" | "parameters">> = [];
+
 export function estimateSkillsTokens(skills: readonly Skill[]): number {
 	const fragments: string[] = [];
 	for (const skill of skills) {
@@ -75,15 +78,16 @@ export function estimateToolSchemaTokens(
  * messages walked incrementally as new entries append.
  */
 export function computeNonMessageTokens(session: AgentSession): number {
-	const parts = computeNonMessageBreakdown(session);
-	return parts.systemPromptTokens + parts.systemContextTokens + parts.toolsTokens + parts.skillsTokens;
+	const systemPromptParts = session.systemPrompt ?? EMPTY_STRING_PARTS;
+	const tools = session.agent?.state?.tools ?? EMPTY_TOOLS;
+	return countTokens(systemPromptParts) + estimateToolSchemaTokens(tools);
 }
 
 /**
- * Shared helper for the four non-message token totals. Single source of truth
- * for both `computeNonMessageTokens` (status-line incremental cache) and
- * `computeContextBreakdown` (/context panel). The split avoids drift between
- * the two surfaces — they MUST report the same numbers.
+ * Shared helper for the four non-message token totals used by
+ * `computeContextBreakdown` (/context panel). Keep this category split stable:
+ * the status-line fast path intentionally uses the equivalent collapsed total
+ * in `computeNonMessageTokens`.
  */
 function computeNonMessageBreakdown(session: AgentSession): {
 	skillsTokens: number;
diff --git a/packages/coding-agent/src/modes/utils/copy-targets.ts b/packages/coding-agent/src/modes/utils/copy-targets.ts
new file mode 100644
index 000000000..6ded2aba8
--- /dev/null
+++ b/packages/coding-agent/src/modes/utils/copy-targets.ts
@@ -0,0 +1,360 @@
+import type { AgentMessage } from "@oh-my-pi/pi-agent-core";
+import type { ToolCall } from "@oh-my-pi/pi-ai";
+
+/** A fenced code block extracted from assistant markdown. */
+export interface CodeBlock {
+	/** Info string after the opening fence (language id), trimmed. */
+	lang: string;
+	/** Block body with the trailing newline stripped. */
+	code: string;
+}
+
+/** A blockquote block: a maximal run of `>`-prefixed lines from markdown. */
+export interface QuoteBlock {
+	/** Block body with each line's `>` marker (and one optional space) removed. */
+	text: string;
+}
+
+/** A drillable block within an assistant message, in document order. */
+export type MessageBlock = ({ kind: "code" } & CodeBlock) | ({ kind: "quote" } & QuoteBlock);
+
+/** A runnable command found in the transcript. */
+export interface LastCommand {
+	kind: "bash" | "eval";
+	code: string;
+	/** Highlight language: "bash" for bash, "python"/"javascript" for eval. */
+	language: string;
+}
+
+/**
+ * A node in the `/copy` picker tree. Leaves carry `content` (placed on the
+ * clipboard) plus `copyMessage` (the status shown afterwards); groups carry
+ * `children` to drill into.
+ */
+export interface CopyTarget {
+	/** Stable id (e.g. "msg:1", "msg:1:code:0", "msg:1:quote:0", "msg:1:all", "cmd:1"). */
+	id: string;
+	label: string;
+	/** Dim annotation: line/block counts, language, or tool name. */
+	hint?: string;
+	/** Full text rendered in the preview pane. */
+	preview: string;
+	/** Highlight language for code/command previews (undefined = plain/markdown). */
+	language?: string;
+	/** Leaf: text copied to the clipboard. */
+	content?: string;
+	/** Leaf: status message shown after copying. */
+	copyMessage?: string;
+	/** Group: nested targets to drill into. */
+	children?: CopyTarget[];
+}
+
+/** Minimal session surface needed to assemble copy targets (eases testing). */
+export interface CopySource {
+	readonly messages: readonly AgentMessage[];
+	getLastVisibleHandoffText(): string | undefined;
+}
+
+/** Cap on how many recent assistant messages the picker lists. */
+const MAX_MESSAGES = 50;
+
+const OPEN_FENCE_RE = /^```([^\n]*)$/;
+const CLOSE_FENCE_RE = /^```/;
+const QUOTE_LINE_RE = /^>(.*)$/;
+
+/**
+ * Split assistant markdown into drillable blocks — fenced code and `>`-quoted
+ * runs — in document order. Fences mask their bodies, so a `>` line inside a
+ * code block is never mistaken for a quote. An unclosed fence is treated as
+ * ordinary text, matching the fenced-block grammar.
+ */
+export function extractBlocks(text: string): MessageBlock[] {
+	const blocks: MessageBlock[] = [];
+	const lines = text.split("\n");
+	let quote: string[] | undefined;
+	const flushQuote = () => {
+		if (quote) {
+			blocks.push({ kind: "quote", text: quote.join("\n") });
+			quote = undefined;
+		}
+	};
+
+	for (let i = 0; i < lines.length; i++) {
+		const line = lines[i]!;
+		const open = OPEN_FENCE_RE.exec(line);
+		if (open) {
+			let close = -1;
+			for (let k = i + 1; k < lines.length; k++) {
+				if (CLOSE_FENCE_RE.test(lines[k]!)) {
+					close = k;
+					break;
+				}
+			}
+			if (close !== -1) {
+				flushQuote();
+				blocks.push({ kind: "code", lang: open[1].trim(), code: lines.slice(i + 1, close).join("\n") });
+				i = close;
+				continue;
+			}
+		}
+
+		const quoted = QUOTE_LINE_RE.exec(line);
+		if (quoted) {
+			// Strip the `>` marker plus one optional following space.
+			quote ??= [];
+			quote.push(quoted[1].startsWith(" ") ? quoted[1].slice(1) : quoted[1]);
+		} else {
+			flushQuote();
+		}
+	}
+	flushQuote();
+	return blocks;
+}
+
+/** Extract fenced code blocks from assistant markdown, in document order. */
+export function extractCodeBlocks(text: string): CodeBlock[] {
+	return extractBlocks(text)
+		.filter((b): b is { kind: "code" } & CodeBlock => b.kind === "code")
+		.map(b => ({ lang: b.lang, code: b.code }));
+}
+
+/** Extract `>`-quoted blocks from assistant markdown, in document order. */
+export function extractQuoteBlocks(text: string): QuoteBlock[] {
+	return extractBlocks(text)
+		.filter((b): b is { kind: "quote" } & QuoteBlock => b.kind === "quote")
+		.map(b => ({ text: b.text }));
+}
+
+function extractEvalCode(args: unknown): { code: string; language: string } | undefined {
+	if (!args || typeof args !== "object") return undefined;
+	const cells = (args as { cells?: unknown }).cells;
+	if (!Array.isArray(cells)) return undefined;
+
+	const codeBlocks: string[] = [];
+	let language = "python";
+	let languageResolved = false;
+	for (const cell of cells) {
+		if (!cell || typeof cell !== "object") continue;
+		const code = (cell as { code?: unknown }).code;
+		if (typeof code !== "string" || code.length === 0) continue;
+		codeBlocks.push(code);
+		if (!languageResolved) {
+			language = (cell as { language?: unknown }).language === "js" ? "javascript" : "python";
+			languageResolved = true;
+		}
+	}
+
+	return codeBlocks.length > 0 ? { code: codeBlocks.join("\n\n"), language } : undefined;
+}
+
+function commandFromToolCall(tc: ToolCall): LastCommand | undefined {
+	if (tc.name === "bash" && typeof tc.arguments.command === "string") {
+		return { kind: "bash", code: tc.arguments.command, language: "bash" };
+	}
+	if (tc.name === "eval") {
+		const evalResult = extractEvalCode(tc.arguments);
+		if (evalResult) return { kind: "eval", code: evalResult.code, language: evalResult.language };
+	}
+	return undefined;
+}
+
+/** Walk the transcript backwards for the most recent bash command or eval code. */
+export function extractLastCommand(messages: readonly AgentMessage[]): LastCommand | undefined {
+	for (let i = messages.length - 1; i >= 0; i--) {
+		const msg = messages[i];
+		if (msg.role !== "assistant") continue;
+		const toolCalls = msg.content.filter((c): c is ToolCall => c.type === "toolCall");
+		for (let j = toolCalls.length - 1; j >= 0; j--) {
+			const command = commandFromToolCall(toolCalls[j]!);
+			if (command) return command;
+		}
+	}
+	return undefined;
+}
+
+/** Concatenated visible text of an assistant message, or undefined when empty. */
+function assistantText(msg: AgentMessage): string | undefined {
+	if (msg.role !== "assistant") return undefined;
+	let text = "";
+	for (const content of msg.content) {
+		if (content.type === "text") text += content.text;
+	}
+	return text.trim() || undefined;
+}
+
+function pluralLines(text: string): string {
+	const count = text.length === 0 ? 0 : text.split("\n").length;
+	return `${count} line${count === 1 ? "" : "s"}`;
+}
+
+function blockHint(block: CodeBlock): string {
+	const lines = pluralLines(block.code);
+	return block.lang ? `${block.lang} · ${lines}` : lines;
+}
+
+/** First non-empty line, whitespace-collapsed, used as a message label. */
+function firstLine(text: string): string {
+	for (const line of text.split("\n")) {
+		const trimmed = line.trim();
+		if (trimmed) return trimmed.replace(/\s+/g, " ");
+	}
+	return text.trim().replace(/\s+/g, " ");
+}
+
+/** "<n> lines · <c> code · <q> quote" — omitting block kinds that are absent. */
+function blockSummaryHint(text: string, codeCount: number, quoteCount: number): string {
+	const parts = [pluralLines(text)];
+	if (codeCount > 0) parts.push(`${codeCount} code`);
+	if (quoteCount > 0) parts.push(`${quoteCount} quote`);
+	return parts.join(" · ");
+}
+
+/** Build the target node for one assistant message: a leaf when it has no
+ * drillable blocks, otherwise a group exposing the full message plus each
+ * fenced code block and `>`-quoted block (de-prefixed) as a child target. */
+function messageTarget(text: string, rank: number): CopyTarget {
+	const id = `msg:${rank}`;
+	const label = firstLine(text);
+	const blocks = extractBlocks(text);
+	const messageCopy = rank === 1 ? "Copied last message to clipboard" : "Copied message to clipboard";
+
+	if (blocks.length === 0) {
+		return { id, label, hint: pluralLines(text), preview: text, content: text, copyMessage: messageCopy };
+	}
+
+	// The message node itself copies the full message; each block is a child
+	// copy target you can drill into, kept in document order.
+	const children: CopyTarget[] = [];
+	const codeBlocks: CodeBlock[] = [];
+	const quoteBlocks: QuoteBlock[] = [];
+	for (const block of blocks) {
+		if (block.kind === "code") {
+			const j = codeBlocks.length;
+			codeBlocks.push(block);
+			children.push({
+				id: `${id}:code:${j}`,
+				label: `Block ${j + 1}`,
+				hint: blockHint(block),
+				preview: block.code,
+				language: block.lang || undefined,
+				content: block.code,
+				copyMessage: `Copied code block ${j + 1} to clipboard`,
+			});
+		} else {
+			const j = quoteBlocks.length;
+			quoteBlocks.push(block);
+			children.push({
+				id: `${id}:quote:${j}`,
+				label: `Quote ${j + 1}`,
+				hint: pluralLines(block.text),
+				preview: block.text,
+				content: block.text,
+				copyMessage: `Copied quote block ${j + 1} to clipboard`,
+			});
+		}
+	}
+
+	if (codeBlocks.length > 1) {
+		const combined = codeBlocks.map(b => b.code).join("\n\n");
+		children.push({
+			id: `${id}:all`,
+			label: `All ${codeBlocks.length} blocks`,
+			hint: pluralLines(combined),
+			preview: combined,
+			content: combined,
+			copyMessage: `Copied ${codeBlocks.length} code blocks to clipboard`,
+		});
+	}
+	if (quoteBlocks.length > 1) {
+		const combined = quoteBlocks.map(b => b.text).join("\n\n");
+		children.push({
+			id: `${id}:all-quotes`,
+			label: `All ${quoteBlocks.length} quotes`,
+			hint: pluralLines(combined),
+			preview: combined,
+			content: combined,
+			copyMessage: `Copied ${quoteBlocks.length} quote blocks to clipboard`,
+		});
+	}
+
+	const hint = blockSummaryHint(text, codeBlocks.length, quoteBlocks.length);
+	return { id, label, hint, preview: text, content: text, copyMessage: messageCopy, children };
+}
+
+function commandTitle(command: LastCommand): string {
+	return command.kind === "bash" ? "Bash command" : "Eval code";
+}
+
+function commandTarget(command: LastCommand, rank: number): CopyTarget {
+	const title = commandTitle(command);
+	return {
+		id: `cmd:${rank}`,
+		label: firstLine(command.code) || title,
+		hint: `${command.kind} · ${pluralLines(command.code)}`,
+		preview: command.code,
+		language: command.language,
+		content: command.code,
+		copyMessage: `Copied ${command.kind === "bash" ? "bash command" : "eval code"} to clipboard`,
+	};
+}
+
+/**
+ * Assemble the unified `/copy` target tree: recent assistant messages
+ * (most recent first, each drillable into its code blocks), runnable command
+ * targets interleaved after the assistant message that issued them, and a
+ * fresh-handoff fallback when no assistant message exists yet.
+ */
+export function buildCopyTargets(source: CopySource): CopyTarget[] {
+	const targets: CopyTarget[] = [];
+	const pendingCommands: LastCommand[] = [];
+	let messageRank = 0;
+	let commandRank = 0;
+
+	const appendCommands = (commands: readonly LastCommand[]) => {
+		for (const command of commands) {
+			commandRank += 1;
+			targets.push(commandTarget(command, commandRank));
+		}
+	};
+
+	for (let i = source.messages.length - 1; i >= 0 && messageRank < MAX_MESSAGES; i--) {
+		const msg = source.messages[i];
+		if (msg.role !== "assistant") continue;
+
+		const toolCalls = msg.content.filter((c): c is ToolCall => c.type === "toolCall");
+		const commands: LastCommand[] = [];
+		for (let j = toolCalls.length - 1; j >= 0; j--) {
+			const command = commandFromToolCall(toolCalls[j]!);
+			if (command) commands.push(command);
+		}
+
+		const text = assistantText(msg);
+		if (!text) {
+			pendingCommands.push(...commands);
+			continue;
+		}
+
+		messageRank += 1;
+		targets.push(messageTarget(text, messageRank));
+		appendCommands(pendingCommands);
+		appendCommands(commands);
+		pendingCommands.length = 0;
+	}
+
+	if (messageRank === 0) {
+		const handoff = source.getLastVisibleHandoffText();
+		if (handoff) {
+			targets.unshift({
+				id: "handoff",
+				label: "Handoff context",
+				hint: pluralLines(handoff),
+				preview: handoff,
+				content: handoff,
+				copyMessage: "Copied handoff context to clipboard",
+			});
+		}
+		appendCommands(pendingCommands);
+	}
+
+	return targets;
+}
diff --git a/packages/coding-agent/src/modes/utils/hotkeys-markdown.ts b/packages/coding-agent/src/modes/utils/hotkeys-markdown.ts
index 21f8b7f1a..97a179e66 100644
--- a/packages/coding-agent/src/modes/utils/hotkeys-markdown.ts
+++ b/packages/coding-agent/src/modes/utils/hotkeys-markdown.ts
@@ -37,6 +37,7 @@ export function buildHotkeysMarkdown(bindings: HotkeysMarkdownBindings): string
 		`| \`${appKey(bindings, "app.clear")}\` | Clear editor (first) / exit (second) |`,
 		`| \`${appKey(bindings, "app.exit")}\` | Exit (when editor is empty) |`,
 		`| \`${appKey(bindings, "app.suspend")}\` | Suspend to background |`,
+		`| \`${appKey(bindings, "app.display.reset")}\` | Reset terminal display |`,
 		`| \`${appKey(bindings, "app.thinking.cycle")}\` | Cycle thinking level |`,
 		`| \`${appKey(bindings, "app.model.cycleForward")}\` | Cycle role models (slow/default/smol) |`,
 		`| \`${appKey(bindings, "app.model.cycleBackward")}\` | Cycle role models (backward) |`,
diff --git a/packages/coding-agent/src/modes/utils/ui-helpers.ts b/packages/coding-agent/src/modes/utils/ui-helpers.ts
index b3d70de4b..fdf6b2876 100644
--- a/packages/coding-agent/src/modes/utils/ui-helpers.ts
+++ b/packages/coding-agent/src/modes/utils/ui-helpers.ts
@@ -17,6 +17,7 @@ import {
 } from "../../modes/components/read-tool-group";
 import { SkillMessageComponent } from "../../modes/components/skill-message";
 import { ToolExecutionComponent } from "../../modes/components/tool-execution";
+import { TranscriptBlock } from "../../modes/components/transcript-container";
 import { UserMessageComponent } from "../../modes/components/user-message";
 import { materializeImageReferenceLinksSync } from "../../modes/image-references";
 import { theme } from "../../modes/theme/theme";
@@ -24,6 +25,7 @@ import type { CompactionQueuedMessage, InteractiveModeContext } from "../../mode
 import {
 	type CustomMessage,
 	isSilentAbort,
+	resolveAbortLabel,
 	SKILL_PROMPT_MESSAGE_TYPE,
 	type SkillPromptDetails,
 } from "../../session/messages";
@@ -73,9 +75,6 @@ export class UiHelpers {
 	 * we update the previous status line instead of appending new ones to avoid log spam.
 	 */
 	showStatus(message: string, options?: { dim?: boolean }): void {
-		if (this.ctx.isBackgrounded) {
-			return;
-		}
 		const children = this.ctx.chatContainer.children;
 		const last = children.length > 0 ? children[children.length - 1] : undefined;
 		const secondLast = children.length > 1 ? children[children.length - 2] : undefined;
@@ -90,11 +89,9 @@ export class UiHelpers {
 
 		const spacer = new Spacer(1);
 		const text = new Text(rendered, 1, 0);
-		this.ctx.chatContainer.addChild(spacer);
-		this.ctx.chatContainer.addChild(text);
+		this.ctx.present([spacer, text]);
 		this.ctx.lastStatusSpacer = spacer;
 		this.ctx.lastStatusText = text;
-		this.ctx.ui.requestRender();
 	}
 
 	addMessageToChat(
@@ -153,6 +150,7 @@ export class UiHelpers {
 											durationMs: details?.durationMs,
 										},
 									];
+						const block = new TranscriptBlock();
 						for (const job of jobs) {
 							const jobId = job.jobId ?? "unknown";
 							const typeLabel = job.type ? `[${job.type}]` : "[job]";
@@ -165,8 +163,9 @@ export class UiHelpers {
 							]
 								.filter(Boolean)
 								.join(" ");
-							this.ctx.chatContainer.addChild(new Text(line, 1, 0));
+							block.addChild(new Text(line, 1, 0));
 						}
+						this.ctx.chatContainer.addChild(block);
 						break;
 					}
 					if (message.customType === SKILL_PROMPT_MESSAGE_TYPE) {
@@ -206,19 +205,18 @@ export class UiHelpers {
 							body = details?.body ?? "";
 							arrow = `${from} ⇨ ${to}`;
 						}
-						const components: Component[] = [];
+						const block = new TranscriptBlock();
 						const header = `${theme.fg("accent", `[IRC] ${arrow}`)}`;
 						const headerComponent = new Text(header, 1, 0);
-						this.ctx.chatContainer.addChild(headerComponent);
-						components.push(headerComponent);
+						block.addChild(headerComponent);
 						if (body) {
 							for (const line of body.split("\n")) {
 								const lineComponent = new Text(theme.fg("muted", `  ${line}`), 0, 0);
-								this.ctx.chatContainer.addChild(lineComponent);
-								components.push(lineComponent);
+								block.addChild(lineComponent);
 							}
 						}
-						return components;
+						this.ctx.chatContainer.addChild(block);
+						return [block];
 					}
 					const renderer = this.ctx.session.extensionRunner?.getMessageRenderer(message.customType);
 					// Both HookMessage and CustomMessage have the same structure, cast for compatibility
@@ -229,14 +227,12 @@ export class UiHelpers {
 				break;
 			}
 			case "compactionSummary": {
-				this.ctx.chatContainer.addChild(new Spacer(1));
 				const component = new CompactionSummaryMessageComponent(message);
 				component.setExpanded(this.ctx.toolOutputExpanded);
 				this.ctx.chatContainer.addChild(component);
 				break;
 			}
 			case "branchSummary": {
-				this.ctx.chatContainer.addChild(new Spacer(1));
 				const component = new BranchSummaryMessageComponent(message);
 				component.setExpanded(this.ctx.toolOutputExpanded);
 				this.ctx.chatContainer.addChild(component);
@@ -244,6 +240,7 @@ export class UiHelpers {
 			}
 			case "fileMention": {
 				// Render compact file mention display
+				const block = new TranscriptBlock();
 				for (const file of message.files) {
 					let suffix: string;
 					if (file.skippedReason === "tooLarge") {
@@ -260,8 +257,9 @@ export class UiHelpers {
 						"accent",
 						file.path,
 					)} ${theme.fg("dim", suffix)}`;
-					this.ctx.chatContainer.addChild(new Text(text, 0, 0));
+					block.addChild(new Text(text, 0, 0));
 				}
+				if (block.children.length > 0) this.ctx.chatContainer.addChild(block);
 				break;
 			}
 			case "user":
@@ -338,18 +336,21 @@ export class UiHelpers {
 				if (assistantComponent) {
 					assistantComponent.setUsageInfo(message.usage);
 				}
-				readGroup = null;
+				const hasVisibleAssistantContent = message.content.some(
+					content =>
+						(content.type === "text" && content.text.trim().length > 0) ||
+						(content.type === "thinking" && content.thinking.trim().length > 0),
+				);
+				if (hasVisibleAssistantContent) {
+					readGroup?.finalize();
+					readGroup = null;
+				}
 				const isAbortedSilently = message.stopReason === "aborted" && isSilentAbort(message.errorMessage);
 				const hasErrorStop =
 					!isAbortedSilently && (message.stopReason === "aborted" || message.stopReason === "error");
 				const errorMessage = hasErrorStop
 					? message.stopReason === "aborted"
-						? (() => {
-								const retryAttempt = this.ctx.session.retryAttempt;
-								return retryAttempt > 0
-									? `Aborted after ${retryAttempt} retry attempt${retryAttempt > 1 ? "s" : ""}`
-									: "Operation aborted";
-							})()
+						? resolveAbortLabel(message.errorMessage, this.ctx.session.retryAttempt)
 						: message.errorMessage || "Error"
 					: null;
 
@@ -391,6 +392,7 @@ export class UiHelpers {
 						continue;
 					}
 
+					readGroup?.finalize();
 					readGroup = null;
 					const tool = this.ctx.session.getToolByName(content.name);
 					const renderArgs =
@@ -478,6 +480,10 @@ export class UiHelpers {
 			}
 		}
 
+		// The trailing read run has no following break to close it; finalize so the
+		// rebuilt group commits to native scrollback like every other historical block.
+		readGroup?.finalize();
+
 		// Render deferred messages (compaction summaries) at the bottom so they're visible
 		for (const message of deferredMessages) {
 			this.ctx.addMessageToChat(message, options);
@@ -490,8 +496,15 @@ export class UiHelpers {
 	renderInitialMessages(prebuiltContext?: SessionContext, options: RenderInitialMessagesOptions = {}): void {
 		// This path is used to rebuild the visible chat transcript (e.g. after custom/debug UI).
 		// Clear existing rendered chat first to avoid duplicating the full session in the container.
+		// On a non-preserving rebuild the existing blocks are discarded for good, so
+		// dispose them (stopping any live timers/subscriptions) before clearing. When
+		// preserving, the same instances are re-added below, so detach without dispose.
 		const preservedChatChildren = options.preserveExistingChat ? this.ctx.chatContainer.children : undefined;
-		this.ctx.chatContainer.clear();
+		if (preservedChatChildren) {
+			this.ctx.chatContainer.clear();
+		} else {
+			this.ctx.resetTranscript();
+		}
 		this.ctx.pendingMessagesContainer.clear();
 		this.ctx.pendingBashComponents = [];
 		this.ctx.pendingPythonComponents = [];
@@ -527,9 +540,6 @@ export class UiHelpers {
 	}
 
 	clearEditor(): void {
-		if (this.ctx.isBackgrounded) {
-			return;
-		}
 		this.ctx.editor.setText("");
 		this.ctx.pendingImages = [];
 		this.ctx.pendingImageLinks = [];
@@ -538,29 +548,17 @@ export class UiHelpers {
 	}
 
 	showError(errorMessage: string): void {
-		if (this.ctx.isBackgrounded) {
-			process.stderr.write(`Error: ${errorMessage}\n`);
-			return;
-		}
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Text(theme.fg("error", `Error: ${errorMessage}`), 1, 0));
-		this.ctx.ui.requestRender();
+		this.ctx.present([new Spacer(1), new Text(theme.fg("error", `Error: ${errorMessage}`), 1, 0)]);
 	}
 
 	showWarning(warningMessage: string): void {
-		if (this.ctx.isBackgrounded) {
-			process.stderr.write(`Warning: ${warningMessage}\n`);
-			return;
-		}
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new Text(theme.fg("warning", `Warning: ${warningMessage}`), 1, 0));
-		this.ctx.ui.requestRender();
+		this.ctx.present([new Spacer(1), new Text(theme.fg("warning", `Warning: ${warningMessage}`), 1, 0)]);
 	}
 
 	showNewVersionNotification(newVersion: string): void {
-		this.ctx.chatContainer.addChild(new Spacer(1));
-		this.ctx.chatContainer.addChild(new DynamicBorder(text => theme.fg("warning", text)));
-		this.ctx.chatContainer.addChild(
+		const block = new TranscriptBlock();
+		block.addChild(new DynamicBorder(text => theme.fg("warning", text)));
+		block.addChild(
 			new Text(
 				theme.bold(theme.fg("warning", "Update Available")) +
 					"\n" +
@@ -570,8 +568,8 @@ export class UiHelpers {
 				0,
 			),
 		);
-		this.ctx.chatContainer.addChild(new DynamicBorder(text => theme.fg("warning", text)));
-		this.ctx.ui.requestRender();
+		block.addChild(new DynamicBorder(text => theme.fg("warning", text)));
+		this.ctx.present(block);
 	}
 
 	updatePendingMessagesDisplay(): void {
diff --git a/packages/coding-agent/src/plan-mode/approved-plan.ts b/packages/coding-agent/src/plan-mode/approved-plan.ts
index dc259e8dc..120fd4212 100644
--- a/packages/coding-agent/src/plan-mode/approved-plan.ts
+++ b/packages/coding-agent/src/plan-mode/approved-plan.ts
@@ -1,15 +1,12 @@
-import * as fs from "node:fs/promises";
-import { isEnoent } from "@oh-my-pi/pi-utils";
-import { resolveLocalUrlToPath } from "../internal-urls";
-import { normalizeLocalScheme } from "../tools/path-utils";
 import { ToolError } from "../tools/tool-errors";
 
 /** Shape forwarded from the plan-mode resolve handler to InteractiveMode's
  *  approval popup. Populated by the standing handler that the resolve tool
- *  dispatches to when the agent submits `resolve { action: "apply" }`. */
+ *  dispatches to when the agent submits `resolve { action: "apply" }`.
+ *  `planFilePath` is the agent-chosen `local://<slug>-plan.md` artifact — it is
+ *  never renamed on approval, so links to it stay valid for the session. */
 export interface PlanApprovalDetails {
 	planFilePath: string;
-	finalPlanFilePath: string;
 	title: string;
 	planExists: boolean;
 }
@@ -110,54 +107,80 @@ export function humanizePlanTitle(title: string): string {
 	return spaced.charAt(0).toUpperCase() + spaced.slice(1);
 }
 
-interface RenameApprovedPlanFileOptions {
+/** The `local://` URL a plan slug maps to. The agent writes the plan here and
+ *  passes the slug to `resolve`; the file is never renamed, so this URL — and
+ *  any hyperlink to it — stays valid for the life of the session. */
+export function planFileUrlForSlug(slug: string): string {
+	return `local://${slug}-plan.md`;
+}
+
+/** Derive a `<slug>` from an agent-supplied `extra.title`, or `undefined` when
+ *  the title is missing/non-string/unsanitizable. A trailing `-plan` is stripped
+ *  so a supplied "auth-plan" maps to `auth-plan.md`, not `auth-plan-plan.md`. */
+function planSlugFromSupplied(suppliedTitle: unknown): string | undefined {
+	if (typeof suppliedTitle !== "string" || !suppliedTitle.trim()) return undefined;
+	try {
+		const { title } = normalizePlanTitle(suppliedTitle);
+		const slug = title.replace(/-plan$/i, "");
+		return slug || title;
+	} catch {
+		return undefined;
+	}
+}
+
+export interface ResolveApprovedPlanInput {
+	/** The agent's `extra.title` from the `resolve` call, if any. */
+	suppliedTitle?: unknown;
+	/** The plan path recorded in plan-mode state (the entry default or a prior plan). */
+	statePlanFilePath: string;
+	/** Read a plan `local://` URL, returning null when the file does not exist. */
+	readPlan: (planUrl: string) => Promise<string | null>;
+	/** Optional fallback: list candidate plan `local://` URLs (newest first) so a
+	 *  plan whose name can't be reconstructed (e.g. a dropped `extra.title`) is
+	 *  still found. */
+	listPlanFiles?: () => Promise<string[]>;
+}
+
+export interface ResolvedApprovedPlan {
 	planFilePath: string;
-	finalPlanFilePath: string;
-	getArtifactsDir: () => string | null;
-	getSessionId: () => string | null;
+	planContent: string;
+	title: string;
 }
 
-function assertLocalUrl(path: string, label: "source" | "destination"): void {
-	if (!path.startsWith("local:/") && !path.startsWith("local://")) {
-		throw new Error(`Approved plan ${label} path must use local: scheme with / or // (received ${path}).`);
-	}
-}
-
-export async function renameApprovedPlanFile(options: RenameApprovedPlanFileOptions): Promise<void> {
-	const { planFilePath, finalPlanFilePath, getArtifactsDir, getSessionId } = options;
-	assertLocalUrl(planFilePath, "source");
-	assertLocalUrl(finalPlanFilePath, "destination");
-
-	const resolveOptions = {
-		getArtifactsDir: () => getArtifactsDir(),
-		getSessionId: () => getSessionId(),
+/** Locate the plan file the agent wrote and finalize its title — without
+ *  renaming anything. Tries, in order: the slug derived from `extra.title`
+ *  (`local://<slug>-plan.md`), the plan path from plan-mode state, then a scan
+ *  of recent plan files. Throws a `ToolError` guiding the agent when none exist. */
+export async function resolveApprovedPlan(input: ResolveApprovedPlanInput): Promise<ResolvedApprovedPlan> {
+	const ordered: string[] = [];
+	const consider = (url: string | undefined): void => {
+		if (url && !ordered.includes(url)) ordered.push(url);
 	};
-	const resolvedSource = resolveLocalUrlToPath(normalizeLocalScheme(planFilePath), resolveOptions);
-	const resolvedDestination = resolveLocalUrlToPath(normalizeLocalScheme(finalPlanFilePath), resolveOptions);
 
-	if (resolvedSource === resolvedDestination) {
-		return;
+	const slug = planSlugFromSupplied(input.suppliedTitle);
+	consider(slug ? planFileUrlForSlug(slug) : undefined);
+	consider(input.statePlanFilePath);
+
+	for (const url of ordered) {
+		const content = await input.readPlan(url);
+		if (content !== null) return finalizeApprovedPlan(url, content, input.suppliedTitle);
 	}
 
-	try {
-		const destinationStat = await fs.stat(resolvedDestination);
-		if (destinationStat.isFile()) {
-			throw new Error(
-				`Plan destination already exists at ${finalPlanFilePath}. Choose a different title and submit the plan for approval again.`,
-			);
-		}
-		throw new Error(`Plan destination exists but is not a file: ${finalPlanFilePath}`);
-	} catch (error) {
-		if (!isEnoent(error)) {
-			throw error;
+	if (input.listPlanFiles) {
+		for (const url of await input.listPlanFiles()) {
+			if (ordered.includes(url)) continue;
+			const content = await input.readPlan(url);
+			if (content !== null) return finalizeApprovedPlan(url, content, input.suppliedTitle);
 		}
 	}
 
-	try {
-		await fs.rename(resolvedSource, resolvedDestination);
-	} catch (error) {
-		throw new Error(
-			`Failed to rename approved plan from ${planFilePath} to ${finalPlanFilePath}: ${error instanceof Error ? error.message : String(error)}`,
-		);
-	}
+	const target = ordered[0] ?? input.statePlanFilePath;
+	throw new ToolError(
+		`Plan file not found at ${target}. Write the finalized plan to ${target} before requesting approval.`,
+	);
+}
+
+function finalizeApprovedPlan(planFilePath: string, planContent: string, suppliedTitle: unknown): ResolvedApprovedPlan {
+	const { title } = resolvePlanTitle({ suppliedTitle, planContent, planFilePath });
+	return { planFilePath, planContent, title };
 }
diff --git a/packages/coding-agent/src/plan-mode/plan-protection.ts b/packages/coding-agent/src/plan-mode/plan-protection.ts
index ed01621b9..4c20ff1dd 100644
--- a/packages/coding-agent/src/plan-mode/plan-protection.ts
+++ b/packages/coding-agent/src/plan-mode/plan-protection.ts
@@ -16,11 +16,11 @@ function readTargetsPlan(readPath: string, planTarget: string): boolean {
  * Build a compaction protection matcher that keeps `read` results for the active
  * plan file intact through prune/shake — the plan analog of skill-read
  * protection. Matches both the canonical `local://PLAN.md` alias and the
- * session's current plan reference path (e.g. a titled `local://<title>.md`), so
- * the plan survives compaction whether the agent reads it by alias or by title.
+ * session's current plan reference path (the agent-chosen `local://<slug>-plan.md`),
+ * so the plan survives compaction whether the agent reads it by alias or by name.
  *
- * `getPlanReferencePath` is evaluated at match time so a mid-session retitle
- * (plan approval renames `PLAN.md` → `<title>.md`) is honored immediately.
+ * `getPlanReferencePath` is evaluated at match time so the plan path set on
+ * approval is honored immediately.
  */
 export function createPlanReadMatcher(getPlanReferencePath: () => string): (context: ProtectedToolContext) => boolean {
 	return (context: ProtectedToolContext) => {
diff --git a/packages/coding-agent/src/prompts/ci-green-request.md b/packages/coding-agent/src/prompts/ci-green-request.md
index 55c30c912..325212a93 100644
--- a/packages/coding-agent/src/prompts/ci-green-request.md
+++ b/packages/coding-agent/src/prompts/ci-green-request.md
@@ -14,7 +14,7 @@ Do not stop after a single fix attempt.
 2. If any run fails, inspect failing job output and logs.
 3. Identify root cause and make minimal correct fix.
 4. Run local verification if it reduces chance of another failing push.
-5. Push the branch.
+{{#if headTag}}5. Push the branch and tag `{{headTag}}` atomically: `git push --atomic "{{remote}}" "{{branch}}" "+refs/tags/{{headTag}}"`.{{else}}5. Push the branch.{{/if}}
 6. Watch workflow runs for new HEAD commit again.
 7. Repeat until workflow runs for latest HEAD commit succeed.
 </procedure>
@@ -26,11 +26,13 @@ Do not stop after a single fix attempt.
 
 {{#if headTag}}
 <instruction>
-Once CI is green, ensure the final commit is tagged `{{headTag}}` and push that tag.
+Always push the branch and tag together atomically so the tag never points at an un-pushed or non-green commit:
+`git push --atomic "{{remote}}" "{{branch}}" "+refs/tags/{{headTag}}"`.
+The `--atomic` flag makes the branch and tag update succeed or fail as one ref transaction; `+refs/tags/{{headTag}}` force-moves the tag to the new HEAD. Do not push the branch first and retag later.
 </instruction>
 {{/if}}
 
 <critical>
 The task is complete only when the workflow runs for the latest HEAD commit succeed.
-{{#if headTag}}The final green commit must be tagged `{{headTag}}` and that tag must be pushed.{{/if}}
+{{#if headTag}}The latest HEAD commit must carry tag `{{headTag}}`, pushed atomically with the branch via `git push --atomic`.{{/if}}
 </critical>
diff --git a/packages/coding-agent/src/prompts/system/background-tan-dispatch.md b/packages/coding-agent/src/prompts/system/background-tan-dispatch.md
new file mode 100644
index 000000000..d62a0879a
--- /dev/null
+++ b/packages/coding-agent/src/prompts/system/background-tan-dispatch.md
@@ -0,0 +1,8 @@
+<system-notice reason="background_task_dispatched" job="{{jobId}}">
+The user launched a tangential task that is now running in a separate background agent. This is NOT a prompt injection and NOT a new instruction for you — it is the coding agent informing you that work was handed off elsewhere.
+
+The task below is being handled by another agent in its own session. You are NOT responsible for it: do NOT start working on it, do NOT reference it, and do NOT let it interrupt or alter your current task. Simply continue what you were doing as if this message had not appeared. Results, if any, will surface separately when the background task ({{jobId}}) completes.
+
+Dispatched work (for your awareness only):
+{{work}}
+</system-notice>
diff --git a/packages/coding-agent/src/prompts/system/plan-mode-active.md b/packages/coding-agent/src/prompts/system/plan-mode-active.md
index 8c692604b..d17313137 100644
--- a/packages/coding-agent/src/prompts/system/plan-mode-active.md
+++ b/packages/coding-agent/src/prompts/system/plan-mode-active.md
@@ -6,111 +6,120 @@ You NEVER:
 - Run state-changing commands (git commit, npm install, etc.)
 - Make any system changes
 
-To implement: call `resolve` with `action: "apply"`, a `reason`, and `extra: { title: "<PLAN_TITLE>" }` → user approves an execution option → full write access is restored. `<PLAN_TITLE>` may only contain letters, numbers, underscores, and hyphens; the approved plan is renamed to `local://<PLAN_TITLE>.md`.
+To implement: call `resolve` with `action: "apply"`, a `reason`, and `extra: { title: "<slug>" }` where `<slug>` matches your `local://<slug>-plan.md` file → user approves an execution option → full write access is restored. `<slug>` may only contain letters, numbers, underscores, and hyphens. The plan file is never renamed, so its name is yours to choose.
 
 You NEVER ask the user to exit plan mode for you; you MUST call `resolve` yourself.
 </critical>
 
+## Objective
+
+A plan is **decision-complete**: another engineer or agent can execute it end-to-end without making a single design decision. Optimize every choice for that. Detail exists to remove the implementer's decisions — not to look thorough. A document that reads like a design doc (Non-Goals, Alternatives, risk matrices) yet leaves real decisions open is a FAILED plan.
+
 ## Plan File
 
 {{#if planExists}}
-Plan file exists at `{{planFilePath}}`; you MUST read and update it incrementally.
+Plan file exists at `{{planFilePath}}`; you MUST read and update it incrementally. If this request is a different task, write a fresh `local://<slug>-plan.md` instead and leave the old plan in place.
 {{else}}
-You MUST create a plan at `{{planFilePath}}`.
+Choose a short kebab-case `<slug>` that names this task (letters, numbers, hyphens) and write the plan to `local://<slug>-plan.md` — e.g. `local://auth-token-refresh-plan.md`. You MUST pass that same `<slug>` as `title` when you call `resolve`.
 {{/if}}
 
-You MUST use `{{editToolName}}` for incremental updates; use `{{writeToolName}}` only for create/full replace.
+You MUST use `{{editToolName}}` for incremental updates; use `{{writeToolName}}` only for create/full replace. You MUST update the plan as you learn — you NEVER batch all writing to the end.
 
-<caution>
-The approval selector includes:
-- **Approve and execute**: starts execution in fresh context (session cleared).
-- **Approve and compact context**: distills the plan-mode discussion into a summary, then starts execution in this session.
-- **Approve and keep context**: starts execution in this session, preserving exploration history.
+## Resolving Unknowns
 
-You MUST still make the plan file self-contained: include requirements, decisions, key findings, and remaining todos.
-</caution>
+You MUST eliminate unknowns by discovering facts, not by asking. Before asking the user anything, perform at least one targeted exploration pass.
+
+Two kinds of unknowns, treated differently:
+- **Discoverable facts** — repo/system truth: file locations, current behavior, existing patterns, types, configs. You MUST explore first (`find`, `search`, `read`, parallel explore subagents). You NEVER ask what the codebase can answer (e.g. "where is this defined?"). Ask only when several plausible candidates remain or a required identifier is genuinely absent — and then present the candidates with a recommendation.
+- **Preferences and tradeoffs** — intent, UX, scope boundaries, performance-vs-simplicity: not derivable from code. You MUST surface these early via `{{askToolName}}` with 2–4 mutually exclusive options and a recommended default. If left unanswered, proceed with the default and record it under Assumptions.
+
+Every question MUST materially change the plan, confirm a load-bearing assumption, or choose between real tradeoffs. You MUST batch questions. You NEVER ask filler questions or offer obviously-wrong options.
 
 {{#if reentry}}
 ## Re-entry
 
 <procedure>
-1. Read existing plan
-2. Evaluate request against it
+1. Read the existing plan.
+2. Evaluate the new request against it.
 3. Decide:
-   - **Different task** → Overwrite plan
-   - **Same task, continuing** → Update and clean outdated sections
-4. Call `resolve` with `action: "apply"` and `extra: { title }` when complete
+   - **Different task** → overwrite the plan.
+   - **Same task, continuing** → update and delete outdated sections.
+4. Call `resolve` with `action: "apply"` and `extra: { title }` when complete.
 </procedure>
 {{/if}}
 
 {{#if iterative}}
-## Iterative Planning
+## Workflow — Iterative
 
 <procedure>
 ### 1. Explore
-You MUST use `find`, `search`, `read` to understand the codebase.
+You MUST use `find`, `search`, `read` to ground yourself in the actual code. Hunt for existing functions, utilities, and conventions to reuse before proposing anything new.
 
 ### 2. Interview
-You MUST use `{{askToolName}}` to clarify:
-- Ambiguous requirements
-- Technical decisions and tradeoffs
-- Preferences: UI/UX, performance, edge cases
+You MUST use `{{askToolName}}` to resolve preferences and tradeoffs (see Resolving Unknowns). Batch questions; never ask what exploration answers.
 
-You MUST batch questions. You NEVER ask what you can answer by exploring.
-
-### 3. Update Incrementally
-You MUST use `{{editToolName}}` to update plan file as you learn; NEVER wait until end.
+### 3. Update incrementally
+You MUST use `{{editToolName}}` to revise the plan file as you learn.
 
 ### 4. Calibrate
-- Large unspecified task → multiple interview rounds
-- Smaller task → fewer or no questions
+- Large, unspecified task → multiple interview rounds.
+- Small, well-specified task → few or no questions.
 </procedure>
-
-<caution>
-### Plan Structure
-
-You MUST use clear markdown headers; include:
-- Recommended approach (not alternatives)
-- Paths of critical files to modify
-- Verification: how to test end-to-end
-
-The plan MUST be scannable yet detailed enough to execute.
-</caution>
-
 {{else}}
-## Planning Workflow
+## Workflow — Parallel
 
 <procedure>
-### Phase 1: Understand
-You MUST focus on the request and associated code. You SHOULD launch parallel explore agents when scope spans multiple areas.
+### Phase 1 — Understand
+You MUST focus on the request and the code behind it. You SHOULD launch parallel `explore` subagents (via `task`) when scope spans multiple areas — give each a distinct focus (existing implementations, related components, test patterns). Actively hunt for reusable functions, utilities, and conventions; avoid proposing new code when a suitable implementation already exists.
 
-### Phase 2: Design
-You MUST draft an approach based on exploration. You MUST consider trade-offs briefly, then choose.
+### Phase 2 — Design
+You MUST draft an approach from your exploration, weigh trade-offs briefly, then commit to one. For large or cross-cutting changes you MAY spawn a planning/critique subagent to pressure-test the approach before you commit.
 
-### Phase 3: Review
-You MUST read critical files. You MUST verify plan matches original request. You SHOULD use `{{askToolName}}` to clarify remaining questions.
+### Phase 3 — Review
+You MUST read the critical files you intend to touch to confirm the approach holds against the real code. You MUST verify the plan still matches the original request. You SHOULD use `{{askToolName}}` to close remaining preference questions.
 
-### Phase 4: Update Plan
-You MUST update `{{planFilePath}}` (`{{editToolName}}` for changes, `{{writeToolName}}` only if creating from scratch):
-- Recommended approach only
-- Paths of critical files to modify
-- Verification section
+### Phase 4 — Write the plan
+You MUST write the plan file (see **Plan File** above) per **The Plan** below.
 </procedure>
-
-<caution>
-You MUST ask questions throughout. You NEVER make large assumptions about user intent.
-</caution>
 {{/if}}
 
+## The Plan
+
+The plan MUST be self-contained: approval may clear or compact this conversation, so the file alone must carry everything needed to execute.
+
+<caution>
+Write 3–5 short, scannable markdown sections. The usual shape:
+- **Context** — why this change: the problem or need, what prompted it, the intended outcome. 2–4 sentences.
+- **Approach** — the recommended approach only. Group bullets by subsystem or behavior, NOT file-by-file. Name existing functions/utilities to reuse, with their paths. Describe a repeated pattern once with a few representative paths — you NEVER enumerate every file or line.
+- **Critical files** — the ≤5 files that disambiguate non-obvious changes, each with a one-line reason. Skip files whose change is already obvious from the Approach.
+- **Verification** — how to test end-to-end: exact commands, tests to run or add, manual steps.
+- **Assumptions** — only the decisions you made that the user might want to override.
+
+Prefer the minimum detail needed for safe implementation, not exhaustive coverage. Compress related changes into high-signal bullets; omit branch-by-branch logic, restated invariants, and lists of unaffected behavior. Behavior-level descriptions beat symbol-by-symbol removal lists.
+</caution>
+
 <directives>
-- You MUST use `{{askToolName}}` only for clarifying requirements or choosing approaches
+- You NEVER include sections that decide nothing: Non-Goals, Out of Scope, Alternatives Considered, Risks/Mitigations boilerplate, Future Work. Omit them entirely.
+- You NEVER invent schema, validation, precedence, or fallback policy the request did not establish, unless it is required to prevent a concrete implementation mistake.
+- You NEVER present alternatives in the final plan — choose. Record a discarded option only when it is a live tradeoff the user should confirm, and put it under Assumptions.
 </directives>
 
+<caution>
+The approval selector offers:
+- **Approve and execute** — execution starts in fresh context (session cleared).
+- **Approve and compact context** — distills this discussion into a summary, then executes in this session.
+- **Approve and keep context** — executes in this session, preserving exploration history.
+
+All three rely on the plan file being self-contained.
+</caution>
+
 <critical>
+You MUST use `{{askToolName}}` only to clarify requirements or choose between approaches.
+
 Your turn ends ONLY by:
 1. Using `{{askToolName}}` to gather information, OR
-2. Calling `resolve` with `action: "apply"`, `reason`, and `extra: { title: "<PLAN_TITLE>" }` when ready — this triggers user approval, then implementation with full tool access
+2. Calling `resolve` with `action: "apply"`, `reason`, and `extra: { title: "<slug>" }` (the slug of your `local://<slug>-plan.md`) when ready — this triggers user approval, then implementation with full tool access.
 
-You NEVER ask plan approval via text or `{{askToolName}}`; you MUST use `resolve`.
-You MUST keep going until complete.
+You NEVER ask for plan approval via text or `{{askToolName}}`; you MUST use `resolve`.
+You MUST keep going until the plan is decision-complete.
 </critical>
diff --git a/packages/coding-agent/src/prompts/system/plan-mode-approved.md b/packages/coding-agent/src/prompts/system/plan-mode-approved.md
index 970e36a09..7b8b0df5f 100644
--- a/packages/coding-agent/src/prompts/system/plan-mode-approved.md
+++ b/packages/coding-agent/src/prompts/system/plan-mode-approved.md
@@ -16,7 +16,7 @@ The plan path is for subagent handoff only. You already have the plan; NEVER rea
 
 The full plan is injected below. You MUST execute it now:
 
-<plan path="{{finalPlanFilePath}}">
+<plan path="{{planFilePath}}">
 {{planContent}}
 </plan>
 
diff --git a/packages/coding-agent/src/prompts/tools/ast-edit.md b/packages/coding-agent/src/prompts/tools/ast-edit.md
index 2b2986f0c..68aefb044 100644
--- a/packages/coding-agent/src/prompts/tools/ast-edit.md
+++ b/packages/coding-agent/src/prompts/tools/ast-edit.md
@@ -14,7 +14,7 @@ Performs structural AST-aware rewrites via native ast-grep.
 </instruction>
 
 <output>
-- Replacement summary, per-file replacement counts, and change diffs as `¶src/foo.ts#0a`, `-12:before`, `+12:after` lines in hashline mode
+- Replacement summary, per-file replacement counts, and change diffs as `[src/foo.ts#1A2B]`, `-12:before`, `+12:after` lines in hashline mode
 - Parse issues when files cannot be processed
 </output>
 
diff --git a/packages/coding-agent/src/prompts/tools/ast-grep.md b/packages/coding-agent/src/prompts/tools/ast-grep.md
index 48502520b..2e7053a29 100644
--- a/packages/coding-agent/src/prompts/tools/ast-grep.md
+++ b/packages/coding-agent/src/prompts/tools/ast-grep.md
@@ -18,7 +18,7 @@ Performs structural code search using AST matching via native ast-grep.
 
 <output>
 - Grouped matches with file path, byte range, line/column ranges, metavariable captures
-- Match lines are numbered under a file snapshot tag header in hashline mode: `¶src/foo.ts#0a`, `*42:content` for the matched line, ` 43:content` for context
+- Match lines are numbered under a file snapshot tag header in hashline mode: `[src/foo.ts#1A2B]`, `*42:content` for the matched line, ` 43:content` for context
 - Summary counts (`totalMatches`, `filesWithMatches`, `filesSearched`) and parse issues when present
 </output>
 
diff --git a/packages/coding-agent/src/prompts/tools/read.md b/packages/coding-agent/src/prompts/tools/read.md
index d8b0aab25..6479105b5 100644
--- a/packages/coding-agent/src/prompts/tools/read.md
+++ b/packages/coding-agent/src/prompts/tools/read.md
@@ -28,7 +28,7 @@ Append `:<sel>` to `path`. The bare path falls back to the default mode.
 
 - Reading a directory path returns a depth-limited dirent listing.
 {{#if IS_HL_MODE}}
-- Reading a file with an explicit selector emits a file snapshot tag header and numbered lines: `¶src/foo.ts#0a` then `41:def alpha():`. Copy the `¶PATH#TAG` header for anchored edits; ops use bare line numbers. NEVER fabricate the tag.
+- Reading a file with an explicit selector emits a file snapshot tag header and numbered lines: `[src/foo.ts#1A2B]` then `41:def alpha():`. Copy the `[PATH#TAG]` header for anchored edits; ops use bare line numbers. NEVER fabricate the tag.
 {{else}}
 {{#if IS_LINE_NUMBER_MODE}}
 - Reading a file with an explicit selector returns lines prefixed with line numbers: `41|def alpha():`.
diff --git a/packages/coding-agent/src/prompts/tools/search.md b/packages/coding-agent/src/prompts/tools/search.md
index 68401694b..245515e79 100644
--- a/packages/coding-agent/src/prompts/tools/search.md
+++ b/packages/coding-agent/src/prompts/tools/search.md
@@ -9,7 +9,7 @@ Searches files using powerful regex matching.
 
 <output>
 {{#if IS_HL_MODE}}
-- Text output emits a file snapshot tag header per matched file plus numbered lines: `¶src/login.ts#1f`, `*42:if (user.id) {` (match), ` 43:return user;` (context). Copy the header for anchored edits; ops use bare line numbers.
+- Text output emits a file snapshot tag header per matched file plus numbered lines: `[src/login.ts#1A2B]`, `*42:if (user.id) {` (match), ` 43:return user;` (context). Copy the header for anchored edits; ops use bare line numbers.
 {{else}}
 {{#if IS_LINE_NUMBER_MODE}}
 - Text output is line-number-prefixed
diff --git a/packages/coding-agent/src/sdk.ts b/packages/coding-agent/src/sdk.ts
index d5f807702..a28d834e5 100644
--- a/packages/coding-agent/src/sdk.ts
+++ b/packages/coding-agent/src/sdk.ts
@@ -10,7 +10,6 @@ import {
 } from "@oh-my-pi/pi-agent-core";
 import {
 	type CredentialDisabledEvent,
-	isUsageLimitError,
 	type Message,
 	type Model,
 	type SimpleStreamOptions,
@@ -24,7 +23,6 @@ import type { Component } from "@oh-my-pi/pi-tui";
 import {
 	$env,
 	$flag,
-	extractRetryHint,
 	getAgentDbPath,
 	getAgentDir,
 	getAuthBrokerSnapshotCachePath,
@@ -36,10 +34,10 @@ import {
 } from "@oh-my-pi/pi-utils";
 import chalk from "chalk";
 import { type AsyncJob, AsyncJobManager, isBackgroundJobSupportEnabled } from "./async";
-import { createAutoresearchExtension } from "./autoresearch";
 import { loadCapability } from "./capability";
 import { type Rule, ruleCapability, setActiveRules } from "./capability/rule";
 import { bucketRules } from "./capability/rule-buckets";
+import { createApiKeyResolver } from "./config/api-key-resolver";
 import { shouldEnableAppendOnlyContext } from "./config/append-only-context-mode";
 import { ModelRegistry } from "./config/model-registry";
 import {
@@ -57,7 +55,6 @@ import { resolveConfigValue } from "./config/resolve-config-value";
 import { initializeWithSettings } from "./discovery";
 import { disposeAllKernelSessions, disposeKernelSessionsByOwner } from "./eval/py/executor";
 import { defaultEvalSessionId } from "./eval/session-id";
-import { TtsrManager } from "./export/ttsr";
 import {
 	type CustomCommandsLoadResult,
 	type LoadedCustomCommand,
@@ -90,7 +87,7 @@ import { LocalProtocolHandler, type LocalProtocolOptions } from "./internal-urls
 import { LSP_STARTUP_EVENT_CHANNEL, type LspStartupEvent } from "./lsp/startup-events";
 import { discoverAndLoadMCPTools, MCPManager, type MCPToolsLoadResult } from "./mcp";
 import { resolveMemoryBackend } from "./memory-backend";
-import { getMnemopiSessionState, type MnemopiSessionState } from "./mnemopi/state";
+import type { MnemopiSessionState } from "./mnemopi/state";
 import asyncResultTemplate from "./prompts/tools/async-result.md" with { type: "text" };
 import { AgentRegistry, MAIN_AGENT_ID } from "./registry/agent-registry";
 import {
@@ -130,6 +127,7 @@ import {
 	resolveThinkingLevelForModel,
 	toReasoningEffort,
 } from "./thinking";
+import { countToolsForAutoDiscovery, resolveEffectiveToolDiscoveryMode } from "./tool-discovery/mode";
 import {
 	collectDiscoverableTools,
 	type DiscoverableTool,
@@ -157,6 +155,7 @@ import {
 	ResolveTool,
 	renderSearchToolBm25Description,
 	SearchTool,
+	SearchToolBm25Tool,
 	setPreferredImageProvider,
 	setPreferredSearchProvider,
 	type Tool,
@@ -280,6 +279,8 @@ export interface CreateAgentSessionOptions {
 	/** Optional provider-facing session identifier for prompt caches and sticky auth selection.
 	 * Keeps persisted session files isolated while reusing provider-side caches. */
 	providerSessionId?: string;
+	/** Optional provider-facing prompt cache key, distinct from request lineage. */
+	providerPromptCacheKey?: string;
 
 	/** Custom tools to register (in addition to built-in tools). Accepts both CustomTool and ToolDefinition. */
 	customTools?: (CustomTool | ToolDefinition)[];
@@ -1148,6 +1149,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 
 	// Discover rules and bucket them in one pass to avoid repeated scans over large rule sets.
 	const { ttsrManager, rulebookRules, alwaysApplyRules } = await logger.time("discoverTtsrRules", async () => {
+		const { TtsrManager } = await import("./export/ttsr");
 		const ttsrSettings = settings.getGroup("ttsr");
 		const ttsrManager = new TtsrManager(ttsrSettings);
 		const rulesResult =
@@ -1293,7 +1295,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 				session ? session.trackEvalExecution(execution, abortController) : execution,
 			getSessionId: () => sessionManager.getSessionId?.() ?? null,
 			getHindsightSessionState: () => session?.getHindsightSessionState(),
-			getMnemopiSessionState: () => getMnemopiSessionState(session),
+			getMnemopiSessionState: () => session?.getMnemopiSessionState(),
 			getAgentId: () => resolvedAgentId,
 			getToolByName: name => session?.getToolByName(name),
 			agentRegistry,
@@ -1470,7 +1472,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 		}
 
 		const inlineExtensions: ExtensionFactory[] = options.extensions ? [...options.extensions] : [];
-		inlineExtensions.push(createAutoresearchExtension);
+		inlineExtensions.push((await import("./autoresearch")).createAutoresearchExtension);
 		if (customTools.length > 0) {
 			inlineExtensions.push(createCustomToolsExtension(customTools));
 		}
@@ -1605,9 +1607,9 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 		// `ExtensionToolWrapper` installed below is the only place the per-tool approval gate runs.
 		// A conditional runner means the approval system silently disappears for users with no
 		// extensions, contradicting non-yolo `tools.approvalMode` settings without feedback.
-		// (Today `createAutoresearchExtension` is unconditionally pushed below, so this scenario
-		// is unreachable; the unconditional construction makes that invariant explicit instead of
-		// implicit, so a future change to make autoresearch optional cannot silently re-open the hole.)
+		// (The builtin autoresearch extension is unconditionally loaded above, so this scenario
+		// is unreachable; unconditional runner construction keeps that invariant explicit and
+		// prevents future optional extensions from silently re-opening the hole.)
 		const extensionRunner: ExtensionRunner = new ExtensionRunner(
 			extensionsResult.extensions,
 			extensionsResult.runtime,
@@ -1687,6 +1689,19 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			}
 		}
 
+		const effectiveDiscoveryMode = resolveEffectiveToolDiscoveryMode(
+			settings,
+			countToolsForAutoDiscovery(toolRegistry.keys()),
+		);
+		if (effectiveDiscoveryMode !== "off" && !toolRegistry.has("search_tool_bm25")) {
+			const searchTool: Tool = new SearchToolBm25Tool(toolSession);
+			toolRegistry.set(
+				searchTool.name,
+				new ExtensionToolWrapper(wrapToolWithMetaNotice(searchTool), extensionRunner) as Tool,
+			);
+		}
+		const mcpDiscoveryEnabled = effectiveDiscoveryMode !== "off"; // back-compat: true when any discovery active
+
 		const reloadSshTool = async (): Promise<AgentTool | null> => {
 			if (!requestedToolNameSet.has("ssh")) return null;
 			const sshTool = (await loadSshTool({
@@ -1708,7 +1723,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 
 		const repeatToolDescriptions = settings.get("repeatToolDescriptions");
 		const eagerTasks = settings.get("task.eager");
-		const intentField = settings.get("tools.intentTracing") || $flag("PI_INTENT_TRACING") ? INTENT_FIELD : undefined;
+		const intentField = $flag("PI_INTENT_TRACING", settings.get("tools.intentTracing")) ? INTENT_FIELD : undefined;
 		const rebuildSystemPrompt = async (
 			toolNames: string[],
 			tools: Map<string, AgentTool>,
@@ -1734,7 +1749,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			const promptTools = buildSystemPromptToolMetadata(tools, {
 				search_tool_bm25: { description: renderSearchToolBm25Description(discoverableToolsForDesc) },
 			});
-			const memoryBackend = resolveMemoryBackend(settings);
+			const memoryBackend = await resolveMemoryBackend(settings);
 			const memoryInstructions = await memoryBackend.buildDeveloperInstructions(agentDir, settings, session);
 
 			// Build combined append prompt: memory instructions + MCP server instructions
@@ -1806,15 +1821,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 		const requestedToolNames = explicitlyRequestedToolNames ?? toolNamesFromRegistry;
 		const normalizedRequested = requestedToolNames.filter(name => toolRegistry.has(name));
 		const requestedToolNameSet = new Set(normalizedRequested);
-		// Effective discovery mode: tools.discoveryMode takes precedence; mcp.discoveryMode is back-compat alias.
-		const toolsDiscoveryModeSetting = settings.get("tools.discoveryMode");
-		const effectiveDiscoveryMode: "off" | "mcp-only" | "all" =
-			toolsDiscoveryModeSetting !== "off"
-				? (toolsDiscoveryModeSetting as "off" | "mcp-only" | "all")
-				: settings.get("mcp.discoveryMode")
-					? "mcp-only"
-					: "off";
-		const mcpDiscoveryEnabled = effectiveDiscoveryMode !== "off"; // back-compat: true when any discovery active
+		// Effective discovery mode is resolved after the full registry exists so auto mode can count MCP/extension tools.
 		const defaultInactiveToolNames = new Set(
 			registeredTools.filter(tool => tool.definition.defaultInactive).map(tool => tool.definition.name),
 		);
@@ -1995,6 +2002,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			onPayload,
 			onResponse,
 			sessionId: providerSessionId,
+			promptCacheKey: options.providerPromptCacheKey,
 			transformContext,
 			steeringMode: settings.get("steeringMode") ?? "one-at-a-time",
 			followUpMode: settings.get("followUpMode") ?? "one-at-a-time",
@@ -2011,9 +2019,15 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			kimiApiFormat: settings.get("providers.kimiApiFormat") ?? "anthropic",
 			preferWebsockets: preferOpenAICodexWebsockets,
 			getToolContext: tc => toolContextStore.getContext(tc),
-			getApiKey: async provider => {
+			getApiKey: async (provider, ctx) => {
 				// Read agent.sessionId at call time so credential selection stays aligned
 				// with metadataResolver after /new, fork, resume, or branch switches.
+				// Retry steps (ctx carries an auth error) drive the central a/b/c
+				// policy — force-refresh the same account, then rotate to a sibling —
+				// and may legitimately yield no key when every account is exhausted.
+				if (ctx?.error !== undefined) {
+					return createApiKeyResolver(modelRegistry, provider, { sessionId: agent.sessionId })(ctx);
+				}
 				const key = await modelRegistry.getApiKeyForProvider(provider, agent.sessionId);
 				if (!key) {
 					throw new Error(`No API key found for provider "${provider}"`);
@@ -2027,40 +2041,6 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 				return streamSimple(streamModel, context, {
 					...streamOptions,
 					openrouterVariant: streamOptions?.openrouterVariant ?? openrouterVariant,
-					onAuthError: async (provider, oldKey, error) => {
-						const message = error instanceof Error ? error.message : String(error);
-						// streamSimple invokes this for both 401 auth failures AND
-						// rotatable usage-limit errors (Codex usage_limit_reached,
-						// Anthropic usage_limit_reached, etc.). The two need
-						// different storage actions: a real 401 means the credential
-						// is bad and should be marked suspect; a usage limit just
-						// means this account is parked until reset and should be
-						// temporarily blocked so a sibling can pick the request up.
-						if (isUsageLimitError(message)) {
-							const retryAfterMs = extractRetryHint(undefined, message);
-							const switched = await modelRegistry.authStorage.markUsageLimitReached(provider, agent.sessionId, {
-								retryAfterMs,
-								signal: streamOptions?.signal,
-							});
-							logger.debug("Retrying provider request after usage-limit block", {
-								provider,
-								switched,
-								retryAfterMs,
-								error: message,
-							});
-							if (!switched) return undefined;
-							return modelRegistry.getApiKeyForProvider(provider, agent.sessionId);
-						}
-						await modelRegistry.authStorage.invalidateCredentialMatching(provider, oldKey, {
-							signal: streamOptions?.signal,
-							sessionId: agent.sessionId,
-						});
-						logger.debug("Retrying provider request after credential invalidation", {
-							provider,
-							error: message,
-						});
-						return modelRegistry.getApiKeyForProvider(provider, agent.sessionId);
-					},
 				});
 			},
 			cursorExecHandlers,
@@ -2260,19 +2240,18 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			}
 		}
 
-		logger.time("startMemoryStartupTask", () =>
-			Promise.resolve(
-				resolveMemoryBackend(settings).start({
-					session,
-					settings,
-					modelRegistry,
-					agentDir,
-					taskDepth,
-					parentHindsightSessionState: options.parentHindsightSessionState,
-					parentMnemopiSessionState: options.parentMnemopiSessionState,
-				}),
-			),
-		);
+		logger.time("startMemoryStartupTask", async () => {
+			const memoryBackend = await resolveMemoryBackend(settings);
+			await memoryBackend.start({
+				session,
+				settings,
+				modelRegistry,
+				agentDir,
+				taskDepth,
+				parentHindsightSessionState: options.parentHindsightSessionState,
+				parentMnemopiSessionState: options.parentMnemopiSessionState,
+			});
+		});
 
 		// Wire MCP manager callbacks to session for reactive tool updates.
 		// Skip when reusing a parent's manager — the parent owns the callbacks.
diff --git a/packages/coding-agent/src/session/agent-session.ts b/packages/coding-agent/src/session/agent-session.ts
index 31775300d..f4558c2fd 100644
--- a/packages/coding-agent/src/session/agent-session.ts
+++ b/packages/coding-agent/src/session/agent-session.ts
@@ -91,6 +91,7 @@ import {
 	extractRetryHint,
 	getAgentDbPath,
 	getInstallId,
+	isBunTestRuntime,
 	isEnoent,
 	isUnexpectedSocketCloseMessage,
 	logger,
@@ -127,7 +128,6 @@ import {
 } from "../eval/py/executor";
 import { defaultEvalSessionId } from "../eval/session-id";
 import { type BashResult, executeBash as executeBashCommand } from "../exec/bash-executor";
-import { exportSessionToHtml } from "../export/html";
 import type { TtsrManager, TtsrMatchContext } from "../export/ttsr";
 import type { LoadedCustomCommand } from "../extensibility/custom-commands";
 import type { CustomTool, CustomToolContext } from "../extensibility/custom-tools/types";
@@ -192,6 +192,7 @@ import {
 	toReasoningEffort,
 } from "../thinking";
 import { shutdownTinyTitleClient } from "../tiny/title-client";
+import { countToolsForAutoDiscovery, resolveEffectiveToolDiscoveryMode } from "../tool-discovery/mode";
 import {
 	buildDiscoverableToolSearchIndex,
 	collectDiscoverableTools,
@@ -470,6 +471,12 @@ export interface SessionStats {
 	cost: number;
 }
 
+export interface FreshSessionResult {
+	previousSessionId: string;
+	sessionId: string;
+	closedProviderSessions: number;
+}
+
 /** Internal marker for hook messages queued through the agent loop */
 // ============================================================================
 // Constants
@@ -921,6 +928,7 @@ export class AgentSession {
 	#agentId: string | undefined;
 	#agentRegistry: AgentRegistry | undefined;
 	#providerSessionId: string | undefined;
+	#freshProviderSessionId: string | undefined;
 	#isDisposed = false;
 	// Extension system
 	#extensionRunner: ExtensionRunner | undefined = undefined;
@@ -1035,6 +1043,7 @@ export class AgentSession {
 
 	#acquirePowerAssertion(): void {
 		if (process.platform !== "darwin") return;
+		if (isBunTestRuntime()) return;
 		if (this.#powerAssertion) return;
 		const idle = this.settings.get("power.preventIdleSleep");
 		const system = this.settings.get("power.preventSystemSleep");
@@ -1273,6 +1282,14 @@ export class AgentSession {
 		return this.#modelRegistry;
 	}
 
+	get asyncJobManager(): AsyncJobManager | undefined {
+		return this.#asyncJobManager;
+	}
+
+	getAgentId(): string | undefined {
+		return this.#agentId;
+	}
+
 	/** Advance the tool-choice queue and return the next directive for the upcoming LLM call. */
 	nextToolChoice(): ToolChoice | undefined {
 		return this.#toolChoiceQueue.nextToolChoice();
@@ -1679,7 +1696,7 @@ export class AgentSession {
 							// Abort the stream immediately — do not gate on extension callbacks
 							this.#ttsrAbortPending = true;
 							this.#ensureTtsrResumePromise();
-							this.agent.abort();
+							this.agent.abort(this.#formatTtsrAbortReason(matches));
 							// Notify extensions (fire-and-forget, does not block abort)
 							this.#emitSessionEvent({ type: "ttsr_triggered", rules: matches }).catch(() => {});
 							// Schedule retry after a short delay
@@ -2160,6 +2177,12 @@ export class AgentSession {
 		}
 	}
 
+	#formatTtsrAbortReason(rules: Rule[]): string {
+		const label = rules.length === 1 ? "rule" : "rules";
+		const ruleNames = rules.map(rule => rule.name).join(", ");
+		return `TTSR matched ${label}: ${ruleNames}`;
+	}
+
 	/** Get TTSR injection payload and clear pending injections. */
 	#getTtsrInjectionContent(): { content: string; rules: Rule[] } | undefined {
 		if (this.#pendingTtsrInjections.length === 0) return undefined;
@@ -2183,13 +2206,20 @@ export class AgentSession {
 	 * project, `~`-relative when it lives under home, else the raw path.
 	 */
 	#displayRulePath(rulePath: string): string {
-		const cwdRel = relativePathWithinRoot(this.sessionManager.getCwd(), rulePath);
+		const cwdRel =
+			relativePathWithinRoot(this.sessionManager.getCwd(), rulePath) ??
+			this.#displayPathWithinRoot(this.sessionManager.getCwd(), rulePath);
 		if (cwdRel) return cwdRel;
 		const homeRel = relativePathWithinRoot(os.homedir(), rulePath);
 		if (homeRel) return `~/${homeRel}`;
 		return rulePath;
 	}
 
+	#displayPathWithinRoot(root: string, candidate: string): string | null {
+		const relative = path.relative(path.resolve(root), path.resolve(candidate));
+		return relative && !relative.startsWith("..") && !path.isAbsolute(relative) ? relative : null;
+	}
+
 	#addPendingTtsrInjections(rules: Rule[]): void {
 		const seen = new Set(this.#pendingTtsrInjections.map(rule => rule.name));
 		for (const rule of rules) {
@@ -2944,6 +2974,10 @@ export class AgentSession {
 		this.#unsubscribeAgent = this.agent.subscribe(this.#handleAgentEvent);
 	}
 
+	#activeProviderSessionId(sessionId?: string): string {
+		return this.#freshProviderSessionId ?? this.#providerSessionId ?? sessionId ?? this.sessionManager.getSessionId();
+	}
+
 	/**
 	 * Set agent.sessionId from the session manager and install a dynamic
 	 * metadata resolver so every Anthropic API request carries
@@ -2956,7 +2990,7 @@ export class AgentSession {
 	 * `#syncAgentSessionId()` on every such event.
 	 */
 	#syncAgentSessionId(sessionId?: string): void {
-		const sid = this.#providerSessionId ?? sessionId ?? this.sessionManager.getSessionId();
+		const sid = this.#activeProviderSessionId(sessionId);
 		this.agent.sessionId = sid;
 		this.agent.setMetadataResolver((provider: string) =>
 			buildSessionMetadata(sid, provider, this.#modelRegistry.authStorage),
@@ -2964,14 +2998,14 @@ export class AgentSession {
 	}
 
 	#rekeyHindsightMemoryForCurrentSessionId(): void {
-		if (resolveMemoryBackend(this.settings).id !== "hindsight") return;
+		if (this.settings.get("memory.backend") !== "hindsight") return;
 		const sid = this.agent.sessionId;
 		if (!sid) return;
 		this.getHindsightSessionState()?.setSessionId(sid);
 	}
 
 	#rekeyMnemopiMemoryForCurrentSessionId(): void {
-		if (resolveMemoryBackend(this.settings).id !== "mnemopi") return;
+		if (this.settings.get("memory.backend") !== "mnemopi") return;
 		const sid = this.agent.sessionId;
 		if (!sid) return;
 		this.getMnemopiSessionState()?.setSessionId(sid);
@@ -2979,14 +3013,14 @@ export class AgentSession {
 
 	/** New session file: reset auto-recall / retain-threshold counters for the new transcript. */
 	#resetHindsightConversationTrackingIfHindsight(): void {
-		if (resolveMemoryBackend(this.settings).id !== "hindsight") return;
+		if (this.settings.get("memory.backend") !== "hindsight") return;
 		const state = this.getHindsightSessionState();
 		if (!state || state.aliasOf) return;
 		state.resetConversationTracking();
 	}
 
 	#resetMnemopiConversationTrackingIfMnemopi(): void {
-		if (resolveMemoryBackend(this.settings).id !== "mnemopi") return;
+		if (this.settings.get("memory.backend") !== "mnemopi") return;
 		const state = this.getMnemopiSessionState();
 		if (!state || state.aliasOf) return;
 		state.resetConversationTracking();
@@ -3086,6 +3120,23 @@ export class AgentSession {
 		this.#providerSessionState.clear();
 	}
 
+	freshSession(): FreshSessionResult | undefined {
+		if (this.isStreaming) return undefined;
+		const previousSessionId = this.sessionId;
+		const closedProviderSessions = this.#providerSessionState.size;
+		this.#closeAllProviderSessions("fresh session");
+		this.#freshProviderSessionId = Bun.randomUUIDv7();
+		this.#syncAgentSessionId();
+		this.#rekeyHindsightMemoryForCurrentSessionId();
+		this.#rekeyMnemopiMemoryForCurrentSessionId();
+		this.agent.appendOnlyContext?.invalidateForModelChange();
+		return {
+			previousSessionId,
+			sessionId: this.sessionId,
+			closedProviderSessions,
+		};
+	}
+
 	// =========================================================================
 	// Read-only State Access
 	// =========================================================================
@@ -3325,12 +3376,14 @@ export class AgentSession {
 
 	// ── Generic tool discovery (covers built-in + MCP + extension) ────────────
 
-	/** Resolve effective discovery mode: tools.discoveryMode wins; mcp.discoveryMode is back-compat alias. */
+	/** Resolve effective discovery mode from the current registry size. */
 	#resolveEffectiveDiscoveryMode(): "off" | "mcp-only" | "all" {
-		const toolsMode = this.settings.get("tools.discoveryMode");
-		if (toolsMode !== "off") return toolsMode as "off" | "mcp-only" | "all";
-		if (this.settings.get("mcp.discoveryMode")) return "mcp-only";
-		return "off";
+		const mode = resolveEffectiveToolDiscoveryMode(
+			this.settings,
+			countToolsForAutoDiscovery(this.#toolRegistry.keys()),
+		);
+		if (mode !== "off") return mode;
+		return this.#mcpDiscoveryEnabled ? "mcp-only" : "off";
 	}
 
 	isToolDiscoveryEnabled(): boolean {
@@ -3665,7 +3718,7 @@ export class AgentSession {
 	}
 
 	async #buildSystemPromptForAgentStart(promptText: string): Promise<string[]> {
-		const backend = resolveMemoryBackend(this.settings);
+		const backend = await resolveMemoryBackend(this.settings);
 		if (!backend.beforeAgentStartPrompt) return this.#baseSystemPrompt;
 
 		try {
@@ -3988,7 +4041,7 @@ export class AgentSession {
 
 	/** Current session ID */
 	get sessionId(): string {
-		return this.#providerSessionId ?? this.sessionManager.getSessionId();
+		return this.#activeProviderSessionId();
 	}
 	getEvalSessionId(): string | null {
 		if (this.#parentEvalSessionId !== undefined) return this.#parentEvalSessionId;
@@ -5087,8 +5140,13 @@ export class AgentSession {
 
 	/**
 	 * Abort current operation and wait for agent to become idle.
+	 *
+	 * `reason` (e.g. `USER_INTERRUPT_LABEL`) rides the agent's `AbortController`
+	 * and surfaces verbatim on the aborted assistant message's `errorMessage`, so
+	 * the transcript can distinguish a deliberate user interrupt from an opaque
+	 * abort. Omit it for internal/lifecycle aborts.
 	 */
-	async abort(options?: { goalReason?: "interrupted" | "internal" }): Promise<void> {
+	async abort(options?: { goalReason?: "interrupted" | "internal"; reason?: string }): Promise<void> {
 		this.abortRetry();
 		this.#promptGeneration++;
 		this.#scheduledHiddenNextTurnGeneration = undefined;
@@ -5097,7 +5155,7 @@ export class AgentSession {
 		this.abortBash();
 		this.abortEval();
 		const postPromptDrain = this.#cancelPostPromptTasks();
-		this.agent.abort();
+		this.agent.abort(options?.reason);
 		await postPromptDrain;
 		await this.agent.waitForIdle();
 		await this.#goalRuntime.onTaskAborted({ reason: options?.goalReason ?? "interrupted" });
@@ -5114,6 +5172,19 @@ export class AgentSession {
 		}
 	}
 
+	/**
+	 * Abort active work, then immediately resume the agent so queued steer/follow-up
+	 * messages drain instead of waiting for another natural turn boundary.
+	 */
+	async interruptAndFlushQueuedMessages(options?: { reason?: string }): Promise<void> {
+		if (!this.agent.hasQueuedMessages()) return;
+		await this.abort({ reason: options?.reason });
+		if (!this.agent.hasQueuedMessages()) return;
+		if (this.isCompacting || this.isGeneratingHandoff) return;
+		await this.#maybeRestoreRetryFallbackPrimary();
+		await this.agent.continue();
+	}
+
 	/**
 	 * Start a new session, optionally with initial messages and parent tracking.
 	 * Clears all messages and starts a new session.
@@ -5158,6 +5229,7 @@ export class AgentSession {
 		}
 		await this.sessionManager.newSession(options);
 		this.setTodoPhases([]);
+		this.#freshProviderSessionId = undefined;
 		this.#syncAgentSessionId();
 		this.#rekeyHindsightMemoryForCurrentSessionId();
 		this.#rekeyMnemopiMemoryForCurrentSessionId();
@@ -5255,6 +5327,7 @@ export class AgentSession {
 		}
 
 		// Update agent session ID
+		this.#freshProviderSessionId = undefined;
 		this.#syncAgentSessionId();
 		this.#rekeyHindsightMemoryForCurrentSessionId();
 		this.#rekeyMnemopiMemoryForCurrentSessionId();
@@ -6091,7 +6164,7 @@ export class AgentSession {
 		messagesToSummarize: AgentMessage[];
 		turnPrefixMessages: AgentMessage[];
 	}): Promise<string | undefined> {
-		const backend = resolveMemoryBackend(this.settings);
+		const backend = await resolveMemoryBackend(this.settings);
 		if (!backend.preCompactionContext) return undefined;
 		const messages = preparation.messagesToSummarize.concat(preparation.turnPrefixMessages);
 		try {
@@ -6222,6 +6295,7 @@ export class AgentSession {
 			this.#cancelOwnAsyncJobs();
 			await this.sessionManager.newSession(previousSessionFile ? { parentSession: previousSessionFile } : undefined);
 			this.agent.reset();
+			this.#freshProviderSessionId = undefined;
 			this.#syncAgentSessionId();
 			this.#rekeyHindsightMemoryForCurrentSessionId();
 			this.#rekeyMnemopiMemoryForCurrentSessionId();
@@ -8150,8 +8224,10 @@ export class AgentSession {
 
 		const currentSelector = this.model ? formatRetryFallbackSelector(this.model, this.thinkingLevel) : undefined;
 		if (!switchedCredential && currentSelector) {
-			this.#noteRetryFallbackCooldown(currentSelector, parsedRetryAfterMs, errorMessage);
-			switchedModel = await this.#tryRetryModelFallback(currentSelector);
+			if (retrySettings.modelFallback) {
+				this.#noteRetryFallbackCooldown(currentSelector, parsedRetryAfterMs, errorMessage);
+				switchedModel = await this.#tryRetryModelFallback(currentSelector);
+			}
 			if (switchedModel) {
 				delayMs = 0;
 			} else if (parsedRetryAfterMs && parsedRetryAfterMs > delayMs) {
@@ -8952,6 +9028,7 @@ export class AgentSession {
 		const previousTools = [...this.agent.state.tools];
 		const previousBaseSystemPrompt = this.#baseSystemPrompt;
 		const previousSystemPrompt = this.agent.state.systemPrompt;
+		const previousFreshProviderSessionId = this.#freshProviderSessionId;
 		const previousFallbackSelectedMCPToolNames = previousSessionFile
 			? this.#getSessionDefaultSelectedMCPToolNames(previousSessionFile)
 			: undefined;
@@ -8963,6 +9040,9 @@ export class AgentSession {
 
 		try {
 			await this.sessionManager.setSessionFile(sessionPath);
+			if (switchingToDifferentSession) {
+				this.#freshProviderSessionId = undefined;
+			}
 			this.#syncAgentSessionId();
 			this.#rekeyHindsightMemoryForCurrentSessionId();
 			this.#rekeyMnemopiMemoryForCurrentSessionId();
@@ -9072,6 +9152,7 @@ export class AgentSession {
 			return true;
 		} catch (error) {
 			this.sessionManager.restoreState(previousSessionState);
+			this.#freshProviderSessionId = previousFreshProviderSessionId;
 			this.#syncAgentSessionId(previousSessionState.sessionId);
 			this.#rekeyHindsightMemoryForCurrentSessionId();
 			this.#rekeyMnemopiMemoryForCurrentSessionId();
@@ -9170,6 +9251,7 @@ export class AgentSession {
 			this.sessionManager.createBranchedSession(selectedEntry.parentId);
 		}
 		this.#syncTodoPhasesFromBranch();
+		this.#freshProviderSessionId = undefined;
 		this.#syncAgentSessionId();
 		this.#rekeyHindsightMemoryForCurrentSessionId();
 		this.#rekeyMnemopiMemoryForCurrentSessionId();
@@ -9598,6 +9680,7 @@ export class AgentSession {
 	 */
 	async exportToHtml(outputPath?: string): Promise<string> {
 		const themeName = getCurrentThemeName();
+		const { exportSessionToHtml } = await import("../export/html");
 		return exportSessionToHtml(this.sessionManager, this.state, { outputPath, themeName });
 	}
 
diff --git a/packages/coding-agent/src/session/messages.ts b/packages/coding-agent/src/session/messages.ts
index 65b6d5de7..c3a1a0484 100644
--- a/packages/coding-agent/src/session/messages.ts
+++ b/packages/coding-agent/src/session/messages.ts
@@ -70,6 +70,32 @@ export function isSilentAbort(errorMessage: string | undefined): boolean {
 	return errorMessage === SILENT_ABORT_MARKER;
 }
 
+/** Reason threaded through `AbortController.abort(reason)` when the user aborts
+ *  the turn with Esc (see `AgentSession.abort`). The agent surfaces it verbatim
+ *  on the aborted assistant message's `errorMessage`, so the transcript reads as
+ *  a deliberate user interrupt instead of an opaque failure. */
+export const USER_INTERRUPT_LABEL = "Interrupted by user";
+
+/** Sentinel `errorMessage` the agent stamps on any abort that carried no custom
+ *  reason (bare `abort()`). Renderers treat it as "no specific reason given". */
+const GENERIC_ABORT_SENTINEL = "Request was aborted";
+
+/** Resolve the operator-facing label for an aborted assistant turn. A custom
+ *  abort reason (e.g. `USER_INTERRUPT_LABEL`) threaded onto `errorMessage` is
+ *  shown verbatim; aborts with no threaded reason fall back to the retry-aware
+ *  generic label. Centralizes the live-stream (`EventController`), replay
+ *  (`ui-helpers`), and component (`AssistantMessageComponent`) render paths so
+ *  they stay in lockstep. */
+export function resolveAbortLabel(errorMessage: string | undefined, retryAttempt = 0): string {
+	if (errorMessage && errorMessage !== GENERIC_ABORT_SENTINEL && !isSilentAbort(errorMessage)) {
+		return errorMessage;
+	}
+	if (retryAttempt > 0) {
+		return `Aborted after ${retryAttempt} retry attempt${retryAttempt > 1 ? "s" : ""}`;
+	}
+	return "Operation aborted";
+}
+
 /** Extract the optional `__pendingDisplayTag` field from a CustomMessage's
  *  `details` blob. Safe over `unknown`; returns undefined when the field is
  *  absent or non-string. */
diff --git a/packages/coding-agent/src/session/session-manager.ts b/packages/coding-agent/src/session/session-manager.ts
index 7d30bccfe..61f78cdcc 100644
--- a/packages/coding-agent/src/session/session-manager.ts
+++ b/packages/coding-agent/src/session/session-manager.ts
@@ -1967,6 +1967,7 @@ export class SessionManager {
 	#inMemoryArtifacts: Map<string, string> | null = null;
 	#inMemoryArtifactCounter = 0;
 	readonly #blobStore: BlobStore;
+	#suppressBreadcrumb = false;
 
 	private constructor(
 		private cwd: string,
@@ -1981,6 +1982,11 @@ export class SessionManager {
 		// Note: call _initSession() or _initSessionFile() after construction
 	}
 
+	#maybeWriteBreadcrumb(cwd: string, sessionFile: string): void {
+		if (this.#suppressBreadcrumb) return;
+		writeTerminalBreadcrumb(cwd, sessionFile);
+	}
+
 	/** Puts a binary blob into the blob store and returns the blob reference */
 	async putBlob(data: Buffer, options?: BlobPutOptions): Promise<BlobPutResult> {
 		return this.#blobStore.put(data, options);
@@ -2027,7 +2033,7 @@ export class SessionManager {
 		this.#adoptedArtifactManager = null;
 		this.#buildIndex();
 		if (this.#sessionFile) {
-			writeTerminalBreadcrumb(this.cwd, this.#sessionFile);
+			this.#maybeWriteBreadcrumb(this.cwd, this.#sessionFile);
 		}
 	}
 
@@ -2047,7 +2053,7 @@ export class SessionManager {
 		this.#persistError = undefined;
 		this.#persistErrorReported = false;
 		this.#sessionFile = path.resolve(sessionFile);
-		writeTerminalBreadcrumb(this.cwd, this.#sessionFile);
+		this.#maybeWriteBreadcrumb(this.cwd, this.#sessionFile);
 		this.#fileEntries = await loadEntriesFromFile(this.#sessionFile, this.storage);
 		if (this.#fileEntries.length > 0) {
 			const header = this.#fileEntries.find(e => e.type === "session") as SessionHeader | undefined;
@@ -2064,7 +2070,7 @@ export class SessionManager {
 			if (headerCwd && headerCwd !== this.cwd) {
 				this.cwd = headerCwd;
 				this.sessionDir = path.resolve(this.#sessionFile, "..");
-				writeTerminalBreadcrumb(this.cwd, this.#sessionFile);
+				this.#maybeWriteBreadcrumb(this.cwd, this.#sessionFile);
 			}
 
 			this.#needsFullRewriteOnNextPersist = migrateToCurrentVersion(this.#fileEntries);
@@ -2245,7 +2251,7 @@ export class SessionManager {
 
 		// Update terminal breadcrumb
 		if (this.#sessionFile) {
-			writeTerminalBreadcrumb(resolvedCwd, this.#sessionFile);
+			this.#maybeWriteBreadcrumb(resolvedCwd, this.#sessionFile);
 		}
 	}
 
@@ -2280,7 +2286,7 @@ export class SessionManager {
 		if (this.persist) {
 			const fileTimestamp = timestamp.replace(/[:.]/g, "-");
 			this.#sessionFile = path.join(this.getSessionDir(), `${fileTimestamp}_${this.#sessionId}.jsonl`);
-			writeTerminalBreadcrumb(this.cwd, this.#sessionFile);
+			this.#maybeWriteBreadcrumb(this.cwd, this.#sessionFile);
 		}
 		return this.#sessionFile;
 	}
@@ -3429,9 +3435,11 @@ export class SessionManager {
 		cwd: string,
 		sessionDir?: string,
 		storage: SessionStorage = new FileSessionStorage(),
+		options?: { suppressBreadcrumb?: boolean },
 	): Promise<SessionManager> {
 		const dir = sessionDir ?? SessionManager.getDefaultSessionDir(cwd, undefined, storage);
 		const manager = new SessionManager(cwd, dir, true, storage);
+		manager.#suppressBreadcrumb = options?.suppressBreadcrumb === true;
 		const forkEntries = structuredClone(await loadEntriesFromFile(sourcePath, storage)) as FileEntry[];
 		migrateToCurrentVersion(forkEntries);
 		await resolveBlobRefsInEntries(forkEntries, manager.#blobStore);
diff --git a/packages/coding-agent/src/slash-commands/builtin-registry.ts b/packages/coding-agent/src/slash-commands/builtin-registry.ts
index b5330eb9c..ea564bde5 100644
--- a/packages/coding-agent/src/slash-commands/builtin-registry.ts
+++ b/packages/coding-agent/src/slash-commands/builtin-registry.ts
@@ -21,6 +21,7 @@ import {
 } from "../extensibility/plugins/marketplace";
 import { resolveMemoryBackend } from "../memory-backend";
 import type { InteractiveModeContext } from "../modes/types";
+import type { FreshSessionResult } from "../session/agent-session";
 import { formatShakeSummary, type ShakeMode } from "../session/shake-types";
 import { getChangelogPath, parseChangelog } from "../utils/changelog";
 import { buildContextReportText } from "./helpers/context-report";
@@ -52,6 +53,11 @@ function refreshStatusLine(ctx: InteractiveModeContext): void {
 	ctx.ui.requestRender();
 }
 
+function formatFreshSessionResult(result: FreshSessionResult): string {
+	const stateLabel = result.closedProviderSessions === 1 ? "provider state" : "provider states";
+	return `Fresh provider session started (${result.closedProviderSessions} ${stateLabel} pruned).`;
+}
+
 const shutdownHandlerTui = (_command: ParsedSlashCommand, runtime: TuiSlashCommandRuntime): SlashCommandResult => {
 	runtime.ctx.editor.setText("");
 	void runtime.ctx.shutdown();
@@ -392,17 +398,9 @@ const BUILTIN_SLASH_COMMAND_REGISTRY: ReadonlyArray<SlashCommandSpec> = [
 	},
 	{
 		name: "copy",
-		description: "Copy last agent message to clipboard",
-		subcommands: [
-			{ name: "last", description: "Copy full last agent message" },
-			{ name: "code", description: "Copy last code block" },
-			{ name: "all", description: "Copy all code blocks from last message" },
-			{ name: "cmd", description: "Copy last bash/python command" },
-		],
-		allowArgs: true,
-		handleTui: async (command, runtime) => {
-			const sub = command.args.trim().toLowerCase() || undefined;
-			await runtime.ctx.handleCopyCommand(sub);
+		description: "Pick text or code from the conversation to copy",
+		handleTui: (_command, runtime) => {
+			runtime.ctx.showCopySelector();
 			runtime.ctx.editor.setText("");
 		},
 	},
@@ -778,6 +776,25 @@ const BUILTIN_SLASH_COMMAND_REGISTRY: ReadonlyArray<SlashCommandSpec> = [
 			await runtime.ctx.handleClearCommand();
 		},
 	},
+	{
+		name: "fresh",
+		description: "Reset provider stream state without changing the local transcript",
+		handle: async (_command, runtime) => {
+			const result = runtime.session.freshSession();
+			if (!result) {
+				await runtime.output(
+					"Wait for the current response to finish or abort it before refreshing provider state.",
+				);
+				return commandConsumed();
+			}
+			await runtime.output(formatFreshSessionResult(result));
+			return commandConsumed();
+		},
+		handleTui: async (_command, runtime) => {
+			runtime.ctx.editor.setText("");
+			await runtime.ctx.handleFreshCommand();
+		},
+	},
 	{
 		name: "drop",
 		description: "Delete the current session and start a new one",
@@ -876,6 +893,17 @@ const BUILTIN_SLASH_COMMAND_REGISTRY: ReadonlyArray<SlashCommandSpec> = [
 			await runtime.ctx.handleBtwCommand(question);
 		},
 	},
+	{
+		name: "tan",
+		description: "Run a full background agent on tangential work",
+		inlineHint: "<work>",
+		allowArgs: true,
+		handleTui: async (command, runtime) => {
+			const work = command.text.slice(`/${command.name}`.length).trim();
+			runtime.ctx.editor.setText("");
+			await runtime.ctx.handleTanCommand(work);
+		},
+	},
 	{
 		name: "omfg",
 		description: "Forge a TTSR rule from a complaint to stop a recurring behavior",
@@ -898,15 +926,6 @@ const BUILTIN_SLASH_COMMAND_REGISTRY: ReadonlyArray<SlashCommandSpec> = [
 			runtime.ctx.editor.setText("");
 		},
 	},
-	{
-		name: "background",
-		aliases: ["bg"],
-		description: "Detach UI and continue running in background",
-		handleTui: (_command, runtime) => {
-			runtime.ctx.editor.setText("");
-			runtime.handleBackgroundCommand();
-		},
-	},
 	{
 		name: "debug",
 		description: "Open debug tools selector",
@@ -942,7 +961,7 @@ const BUILTIN_SLASH_COMMAND_REGISTRY: ReadonlyArray<SlashCommandSpec> = [
 		allowArgs: true,
 		handle: async (command, runtime) => {
 			const verb = (command.args.trim().split(/\s+/)[0] ?? "").toLowerCase() || "view";
-			const backend = resolveMemoryBackend(runtime.settings);
+			const backend = await resolveMemoryBackend(runtime.settings);
 			switch (verb) {
 				case "view": {
 					const payload = await backend.buildDeveloperInstructions(
diff --git a/packages/coding-agent/src/slash-commands/helpers/usage-report.ts b/packages/coding-agent/src/slash-commands/helpers/usage-report.ts
index bbef96730..8b3610f9d 100644
--- a/packages/coding-agent/src/slash-commands/helpers/usage-report.ts
+++ b/packages/coding-agent/src/slash-commands/helpers/usage-report.ts
@@ -26,6 +26,8 @@ function formatUsageReportAccount(report: UsageReport, limit: UsageLimit, index:
 	if (typeof email === "string" && email) return email;
 	const accountId = report.metadata?.accountId ?? limit.scope.accountId;
 	if (typeof accountId === "string" && accountId) return accountId;
+	const projectId = report.metadata?.projectId ?? limit.scope.projectId;
+	if (typeof projectId === "string" && projectId) return projectId;
 	return `account ${index + 1}`;
 }
 
diff --git a/packages/coding-agent/src/slash-commands/types.ts b/packages/coding-agent/src/slash-commands/types.ts
index 6725815ba..15412739a 100644
--- a/packages/coding-agent/src/slash-commands/types.ts
+++ b/packages/coding-agent/src/slash-commands/types.ts
@@ -71,15 +71,13 @@ export interface SlashCommandRuntime {
 
 /**
  * Runtime visible to TUI-only handlers (`handleTui`). Carries the interactive
- * mode context plus the background-detach hook. Intentionally narrower than
- * `SlashCommandRuntime` so existing callers can keep building it from just
- * `{ ctx, handleBackgroundCommand }`; when the TUI dispatcher needs to invoke
- * a `handle` (no `handleTui` override), it synthesizes a `SlashCommandRuntime`
- * from `ctx`.
+ * mode context. Intentionally narrower than `SlashCommandRuntime` so existing
+ * callers can keep building it from just `{ ctx }`; when the TUI dispatcher
+ * needs to invoke a `handle` (no `handleTui` override), it synthesizes a
+ * `SlashCommandRuntime` from `ctx`.
  */
 export interface TuiSlashCommandRuntime {
 	ctx: InteractiveModeContext;
-	handleBackgroundCommand: () => void;
 }
 
 /** Unified slash-command spec consumed by both TUI and ACP dispatchers. */
diff --git a/packages/coding-agent/src/task/executor.ts b/packages/coding-agent/src/task/executor.ts
index 94fb9a63b..6f47ab87a 100644
--- a/packages/coding-agent/src/task/executor.ts
+++ b/packages/coding-agent/src/task/executor.ts
@@ -166,6 +166,13 @@ export interface ExecutorOptions {
 	outputSchema?: unknown;
 	/** Parent task recursion depth (0 = top-level, 1 = first child, etc.) */
 	taskDepth?: number;
+	/**
+	 * Override the `task.maxRuntimeMs` wall-clock cap for this run. When provided
+	 * it wins over the settings value; `0` disables the per-subagent wall-clock
+	 * limit entirely. Used by the eval `agent()` bridge, whose parent cell
+	 * watchdog is already suspended for the call's duration.
+	 */
+	maxRuntimeMs?: number;
 	enableLsp?: boolean;
 	signal?: AbortSignal;
 	onProgress?: (progress: AgentProgress) => void;
@@ -481,7 +488,7 @@ function getUsageTokens(usage: unknown): number {
 /**
  * Create proxy tools that reuse the parent's MCP connections.
  */
-function createMCPProxyTools(mcpManager: MCPManager): CustomTool[] {
+export function createMCPProxyTools(mcpManager: MCPManager): CustomTool[] {
 	return mcpManager.getTools().map(tool => {
 		const mcpTool = tool as { mcpToolName?: string; mcpServerName?: string };
 		return {
@@ -531,7 +538,10 @@ function createMCPProxyTools(mcpManager: MCPManager): CustomTool[] {
 	});
 }
 
-function createSubagentSettings(baseSettings: Settings, overrides?: Partial<Record<SettingPath, unknown>>): Settings {
+export function createSubagentSettings(
+	baseSettings: Settings,
+	overrides?: Partial<Record<SettingPath, unknown>>,
+): Settings {
 	const snapshot: Partial<Record<SettingPath, unknown>> = {};
 	for (const key of Object.keys(SETTINGS_SCHEMA) as SettingPath[]) {
 		snapshot[key] = baseSettings.get(key);
@@ -625,7 +635,10 @@ export async function runSubprocess(options: ExecutorOptions): Promise<SingleRes
 		agent.readSummarize === false ? { "read.summarize.enabled": false } : undefined,
 	);
 	const maxRecursionDepth = settings.get("task.maxRecursionDepth") ?? 2;
-	const maxRuntimeMs = Math.max(0, Math.trunc(Number(settings.get("task.maxRuntimeMs") ?? 0) || 0));
+	const maxRuntimeMs = Math.max(
+		0,
+		Math.trunc(Number(options.maxRuntimeMs ?? settings.get("task.maxRuntimeMs") ?? 0) || 0),
+	);
 	const parentDepth = options.taskDepth ?? 0;
 	const childDepth = parentDepth + 1;
 	const atMaxDepth = maxRecursionDepth >= 0 && childDepth >= maxRecursionDepth;
@@ -1484,7 +1497,15 @@ export async function runSubprocess(options: ExecutorOptions): Promise<SingleRes
 				if (lastAssistant.stopReason === "aborted") {
 					aborted = abortReason === "signal" || runtimeLimitExceeded || abortReason === undefined;
 					if (aborted) {
-						abortReasonText ??= resolveAbortReasonText();
+						// A real caller signal or the wall-clock timer carries a precise
+						// reason (signal.reason / "runtime limit exceeded"). An internal
+						// turn abort (abortReason === undefined) does NOT — prefer the
+						// assistant message's own errorMessage ("Request was aborted" or a
+						// specific stream error) over the misleading "Cancelled by caller".
+						abortReasonText ??=
+							abortReason === "signal" || runtimeLimitExceeded
+								? resolveAbortReasonText()
+								: lastAssistant.errorMessage?.trim() || resolveAbortReasonText();
 					}
 					exitCode = 1;
 				} else if (lastAssistant.stopReason === "error") {
diff --git a/packages/coding-agent/src/task/index.ts b/packages/coding-agent/src/task/index.ts
index fe3f6f626..befea661d 100644
--- a/packages/coding-agent/src/task/index.ts
+++ b/packages/coding-agent/src/task/index.ts
@@ -275,6 +275,10 @@ export class TaskTool implements AgentTool<TaskToolSchemaInstance, TaskToolDetai
 	readonly strict = true;
 	readonly loadMode = "discoverable";
 	readonly renderResult = renderResult;
+	// Suppress the streaming call preview once a (partial or final) result exists
+	// so the task renders as ONE block that transitions in place — not a pending
+	// call frame stacked above the result frame. Mirrors `taskToolRenderer`.
+	readonly mergeCallAndResult = true;
 	readonly #discoveredAgents: AgentDefinition[];
 	readonly #blockedAgent: string | undefined;
 
diff --git a/packages/coding-agent/src/task/render.ts b/packages/coding-agent/src/task/render.ts
index 5babac27e..1bb94527f 100644
--- a/packages/coding-agent/src/task/render.ts
+++ b/packages/coding-agent/src/task/render.ts
@@ -6,18 +6,20 @@
  */
 import path from "node:path";
 import type { Component } from "@oh-my-pi/pi-tui";
-import { Container, Text } from "@oh-my-pi/pi-tui";
+import { Container, Markdown, Text } from "@oh-my-pi/pi-tui";
 import { formatNumber } from "@oh-my-pi/pi-utils";
 import { settings } from "../config/settings";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import { formatContextUsage } from "../modes/components/status-line/context-thresholds";
-import type { Theme } from "../modes/theme/theme";
+import { shimmerEnabled, shimmerText } from "../modes/theme/shimmer";
+import { getMarkdownTheme, type Theme } from "../modes/theme/theme";
 import {
 	formatBadge,
 	formatDuration,
 	formatMoreItems,
 	formatStatusIcon,
 	replaceTabs,
+	type ToolUIStatus,
 	truncateToWidth,
 } from "../tools/render-utils";
 import {
@@ -28,7 +30,8 @@ import {
 	type ReportFindingDetails,
 	type SubmitReviewDetails,
 } from "../tools/review";
-import { Ellipsis, Hasher, type RenderCache, renderStatusLine } from "../tui";
+import { framedBlock, renderStatusLine } from "../tui";
+import { repairDoubleEncodedJsonString } from "./repair-args";
 import { subprocessToolRegistry } from "./subprocess-tool-registry";
 import type { AgentProgress, SingleResult, TaskItem, TaskParams, TaskToolDetails } from "./types";
 
@@ -117,6 +120,26 @@ function normalizeReportFindings(value: unknown): ReportFindingDetails[] {
 	return findings;
 }
 
+/**
+ * Normalize the `yield` slot of `extractedToolData` into an array of
+ * yield-detail records. The subprocess executor always populates this slot as
+ * `unknown[]` (see `executor.ts` `extractData` handler), but the renderer
+ * MUST also tolerate a stray single object — optional chaining short-circuits
+ * on `null`/`undefined` only, so calling `.map` on a plain object would throw
+ * `TypeError: completeData?.map is not a function` and crash the TUI.
+ * A single object is wrapped as a 1-element array so the review verdict still
+ * renders; non-object primitives drop out.
+ */
+function normalizeYieldData(value: unknown): Array<{ data: unknown }> {
+	if (Array.isArray(value)) {
+		return value.filter((item): item is { data: unknown } => item !== null && typeof item === "object");
+	}
+	if (value !== null && typeof value === "object") {
+		return [value as { data: unknown }];
+	}
+	return [];
+}
+
 function formatJsonScalar(value: unknown, _theme: Theme): string {
 	if (value === null) return "null";
 	if (typeof value === "string") {
@@ -486,28 +509,20 @@ function formatOutputInline(data: unknown, theme: Theme, maxWidth = 80): string
  * preview. The args stream in token by token, so the array grows over time and
  * trailing entries may be partially parsed — every field access is defensive.
  */
-function renderTaskItemLines(
-	tasks: TaskItem[] | undefined,
-	contPrefix: string,
-	expanded: boolean,
-	theme: Theme,
-): string[] {
+function renderTaskItemLines(tasks: TaskItem[] | undefined, expanded: boolean, theme: Theme): string[] {
 	const items = tasks ?? [];
 	if (items.length === 0) return [];
 
-	const branch = theme.fg("dim", theme.tree.branch);
-	const last = theme.fg("dim", theme.tree.last);
+	const bullet = theme.fg("dim", "•");
 	const cap = expanded ? items.length : Math.min(items.length, 12);
 	const truncated = cap < items.length;
 
 	const lines: string[] = [];
 	for (let i = 0; i < cap; i++) {
 		const task = items[i] as Partial<TaskItem> | undefined;
-		const isLastLine = !truncated && i === items.length - 1;
-		const connector = isLastLine ? last : branch;
 		const rawId = task?.id?.trim();
 		const idLabel = rawId ? formatTaskId(rawId) : `#${i + 1}`;
-		let line = `${contPrefix}${connector} ${theme.fg("accent", theme.bold(idLabel))}`;
+		let line = `${bullet} ${theme.fg("accent", theme.bold(idLabel))}`;
 		const desc = task?.description?.trim();
 		if (desc) {
 			line += `: ${theme.fg("muted", truncateToWidth(replaceTabs(desc), 64))}`;
@@ -515,11 +530,42 @@ function renderTaskItemLines(
 		lines.push(line);
 	}
 	if (truncated) {
-		lines.push(`${contPrefix}${last} ${theme.fg("dim", formatMoreItems(items.length - cap, "agent"))}`);
+		lines.push(`${bullet} ${theme.fg("dim", formatMoreItems(items.length - cap, "agent"))}`);
 	}
 	return lines;
 }
 
+/**
+ * Build the shared-context section (the `# Goal / # Constraints` background
+ * passed to every subagent). Rendered in both the streaming call preview and
+ * the merged result frame so the brief stays visible for the whole task
+ * lifecycle — not just until the first progress snapshot replaces the call view.
+ */
+type TaskRenderSection = { lines: string[] };
+type ContextSectionRenderer = (width: number) => TaskRenderSection;
+
+// Default output-block layout is: left border + one-cell content inset + right
+// border. Render markdown at that inner width so the output block does not need
+// to rewrap already-rendered context lines.
+const CONTEXT_FRAME_INSET = 3;
+
+function contextMarkdownWidth(frameWidth: number): number {
+	return Math.max(1, frameWidth - CONTEXT_FRAME_INSET);
+}
+
+function createContextSectionRenderer(args: TaskParams | undefined, theme: Theme): ContextSectionRenderer | undefined {
+	// `renderResult` receives the raw tool args (unlike `renderCall`, which is
+	// fed through `repairTaskParams`), so undo any per-field double-encoding here
+	// too. The repair is idempotent on already-clean text.
+	const context = repairDoubleEncodedJsonString(args?.context ?? "").trim();
+	if (!context) return undefined;
+
+	const markdown = new Markdown(context, 0, 0, getMarkdownTheme(), {
+		color: text => theme.fg("muted", text),
+	});
+	return width => ({ lines: markdown.render(contextMarkdownWidth(width)) });
+}
+
 /**
  * Render the tool call arguments.
  */
@@ -528,43 +574,34 @@ export function renderCall(
 	options: RenderResultOptions & { renderContext?: { hasResult?: boolean } },
 	theme: Theme,
 ): Component {
-	const lines: string[] = [];
-	lines.push(renderStatusLine({ icon: "pending", title: "Task", description: args.agent }, theme));
-
-	const context = (args.context ?? "").trim();
-	const hasContext = context.length > 0;
-	const branch = theme.fg("dim", theme.tree.branch);
-	const last = theme.fg("dim", theme.tree.last);
-	const vertical = theme.fg("dim", theme.tree.vertical);
 	const showIsolated = "isolated" in args && args.isolated === true;
-	const taskCount = args.tasks?.length ?? 0;
+	const header = renderStatusLine({ icon: "pending", title: "Task", description: args.agent }, theme);
+	const contextSectionRenderer = createContextSectionRenderer(args, theme);
+	return framedBlock(theme, width => {
+		const sections: Array<{ label?: string; lines: string[]; separator?: boolean }> = [];
 
-	if (hasContext) {
-		lines.push(` ${branch} ${theme.fg("dim", "Context")}`);
-		for (const line of context.split("\n")) {
-			const content = line ? theme.fg("muted", replaceTabs(line)) : "";
-			lines.push(` ${vertical}  ${content}`);
+		if (contextSectionRenderer) sections.push(contextSectionRenderer(width));
+
+		// The per-task preview list only exists to surface dispatched agents while
+		// the call args stream in. Once a result snapshot exists, `renderResult`
+		// draws the same agents as progress/result lines, so showing the Tasks
+		// section here would just repeat the count the result frame already shows.
+		if (!options.renderContext?.hasResult) {
+			sections.push({
+				separator: true,
+				lines: renderTaskItemLines(args.tasks, options.expanded, theme),
+			});
 		}
-	}
 
-	// `Tasks` is the last child unless the isolation flag follows it.
-	const tasksIsLast = !showIsolated;
-	const tasksPrefix = tasksIsLast ? last : branch;
-	lines.push(` ${tasksPrefix} ${theme.fg("dim", "Tasks")} ${theme.fg("muted", `(${taskCount})`)}`);
-	const tasksContPrefix = tasksIsLast ? "    " : ` ${vertical}  `;
-	// The per-task preview list only exists to surface dispatched agents while
-	// the call args stream in. Once a result snapshot exists, `renderResult`
-	// draws the same agents as progress/result lines (id + description), so
-	// emitting the preview here would render every task twice.
-	if (!options.renderContext?.hasResult) {
-		lines.push(...renderTaskItemLines(args.tasks, tasksContPrefix, options.expanded, theme));
-	}
-
-	if (showIsolated) {
-		lines.push(` ${last} ${theme.fg("dim", "Isolated")}: ${theme.fg("muted", "true")}`);
-	}
-
-	return new Text(lines.join("\n"), 0, 0);
+		return {
+			header,
+			headerMeta: showIsolated ? "isolated" : undefined,
+			sections,
+			state: "pending",
+			borderColor: "borderMuted",
+			width,
+		};
+	});
 }
 
 /**
@@ -572,14 +609,13 @@ export function renderCall(
  */
 function renderAgentProgress(
 	progress: AgentProgress,
-	isLast: boolean,
+	prefix: string,
+	continuePrefix: string,
 	expanded: boolean,
 	theme: Theme,
 	spinnerFrame?: number,
 ): string[] {
 	const lines: string[] = [];
-	const prefix = isLast ? theme.fg("dim", theme.tree.last) : theme.fg("dim", theme.tree.branch);
-	const continuePrefix = isLast ? "   " : `${theme.fg("dim", theme.tree.vertical)}  `;
 
 	const icon = getStatusIcon(progress.status, theme, spinnerFrame);
 	const iconColor =
@@ -593,11 +629,24 @@ function renderAgentProgress(
 	const description = progress.description?.trim();
 	const displayId = formatTaskId(progress.id);
 	const titlePart = description ? `${theme.bold(displayId)}: ${description}` : displayId;
-	let statusLine = `${prefix} ${theme.fg(iconColor, icon)} ${theme.fg("accent", titlePart)}`;
+	const indent = prefix ? `${prefix} ` : "";
+	let statusLine: string;
+	if (progress.status === "running") {
+		const bullet = theme.fg("accent", "•");
+		const name = shimmerEnabled()
+			? shimmerText(displayId, theme)
+			: theme.fg("accent", description ? theme.bold(displayId) : displayId);
+		statusLine = `${indent}${bullet} ${name}`;
+		if (description) {
+			statusLine += theme.fg("accent", `: ${description}`);
+		}
+	} else {
+		statusLine = `${indent}${theme.fg(iconColor, icon)} ${theme.fg("accent", titlePart)}`;
+	}
 
 	// Show retry-blocked badge so the parent immediately sees that a child
 	// is sleeping on a provider 429, not silently progressing. Wins over the
-	// generic running spinner because "we're waiting on a quota window" is
+	// generic running marker because "we're waiting on a quota window" is
 	// the operationally meaningful state.
 	if (progress.retryState && progress.status === "running") {
 		statusLine += ` ${formatBadge("retrying", "warning", theme)}`;
@@ -671,12 +720,12 @@ function renderAgentProgress(
 	if (progress.extractedToolData) {
 		// For completed tasks, check for review verdict from yield tool
 		if (progress.status === "completed") {
-			const completeData = progress.extractedToolData.yield as Array<{ data: unknown }> | undefined;
+			const completeData = normalizeYieldData(progress.extractedToolData.yield);
 			const reportFindingData = normalizeReportFindings(progress.extractedToolData.report_finding);
 			const reviewData = completeData
-				?.map(c => c.data as SubmitReviewDetails)
+				.map(c => c.data as SubmitReviewDetails)
 				.filter(d => d && typeof d === "object" && "overall_correctness" in d);
-			if (reviewData && reviewData.length > 0) {
+			if (reviewData.length > 0) {
 				const summary = reviewData[reviewData.length - 1];
 				const findings = reportFindingData;
 				lines.push(...renderReviewResult(summary, findings, continuePrefix, expanded, theme));
@@ -846,10 +895,14 @@ function renderFindings(
 /**
  * Render final result for a single agent.
  */
-function renderAgentResult(result: SingleResult, isLast: boolean, expanded: boolean, theme: Theme): string[] {
+function renderAgentResult(
+	result: SingleResult,
+	prefix: string,
+	continuePrefix: string,
+	expanded: boolean,
+	theme: Theme,
+): string[] {
 	const lines: string[] = [];
-	const prefix = isLast ? theme.fg("dim", theme.tree.last) : theme.fg("dim", theme.tree.branch);
-	const continuePrefix = isLast ? "   " : `${theme.fg("dim", theme.tree.vertical)}  `;
 
 	const { warning: missingCompleteWarning, rest: outputWithoutWarning } = extractMissingYieldWarning(result.output);
 	const aborted = result.aborted ?? false;
@@ -878,7 +931,7 @@ function renderAgentResult(result: SingleResult, isLast: boolean, expanded: bool
 	const description = result.description?.trim();
 	const displayId = formatTaskId(result.id);
 	const titlePart = description ? `${theme.bold(displayId)}: ${description}` : displayId;
-	let statusLine = `${prefix} ${theme.fg(iconColor, icon)} ${theme.fg("accent", titlePart)} ${formatBadge(
+	let statusLine = `${prefix ? `${prefix} ` : ""}${theme.fg(iconColor, icon)} ${theme.fg("accent", titlePart)} ${formatBadge(
 		statusText,
 		iconColor,
 		theme,
@@ -912,16 +965,21 @@ function renderAgentResult(result: SingleResult, isLast: boolean, expanded: bool
 		);
 	}
 	// Check for review result (yield with review schema + report_finding)
-	const completeData = result.extractedToolData?.yield as Array<{ data: unknown }> | undefined;
+	// Check for review result (yield with review schema + report_finding).
+	// `normalizeYieldData` guards against a stray non-array `yield` slot —
+	// optional chaining on `.map` only short-circuits on null/undefined and
+	// would otherwise crash the renderer with `TypeError: completeData?.map
+	// is not a function` when the slot is a plain object (see issue #1987).
+	const completeData = normalizeYieldData(result.extractedToolData?.yield);
 	const reportFindingData = normalizeReportFindings(result.extractedToolData?.report_finding);
 
 	// Extract review verdict from yield tool's data field if it matches SubmitReviewDetails
 	const reviewData = completeData
-		?.map(c => c.data as SubmitReviewDetails)
+		.map(c => c.data as SubmitReviewDetails)
 		.filter(d => d && typeof d === "object" && "overall_correctness" in d);
-	const submitReviewData = reviewData && reviewData.length > 0 ? reviewData : undefined;
+	const submitReviewData = reviewData.length > 0 ? reviewData : undefined;
 
-	if (submitReviewData && submitReviewData.length > 0) {
+	if (submitReviewData) {
 		// Use combined review renderer
 		const summary = submitReviewData[submitReviewData.length - 1];
 		const findings = reportFindingData;
@@ -929,7 +987,7 @@ function renderAgentResult(result: SingleResult, isLast: boolean, expanded: bool
 		return lines;
 	}
 	if (reportFindingData.length > 0) {
-		const hasCompleteData = completeData && completeData.length > 0;
+		const hasCompleteData = completeData.length > 0;
 		const message = hasCompleteData
 			? "Review verdict missing expected fields"
 			: "Review incomplete (yield not called)";
@@ -1017,101 +1075,123 @@ export function renderResult(
 	result: { content: Array<{ type: string; text?: string }>; details?: TaskToolDetails },
 	options: RenderResultOptions,
 	theme: Theme,
+	args?: TaskParams,
 ): Component {
 	const fallbackText = result.content.find(c => c.type === "text")?.text ?? "";
 	const details = result.details;
+	const contextSectionRenderer = createContextSectionRenderer(args, theme);
 
 	if (!details) {
 		const text = result.content.find(c => c.type === "text")?.text || "";
-		return new Text(theme.fg("dim", truncateToWidth(text, 100)), 0, 0);
+		const header = renderStatusLine({ icon: "success", title: "Task" }, theme);
+		return framedBlock(theme, width => ({
+			header,
+			sections: [
+				...(contextSectionRenderer ? [contextSectionRenderer(width)] : []),
+				...(text ? [{ separator: true, lines: [theme.fg("dim", truncateToWidth(text, width))] }] : []),
+			],
+			state: "success",
+			borderColor: "borderMuted",
+			width,
+		}));
 	}
 
-	let cached: RenderCache | undefined;
+	const hasResults = Boolean(details.results && details.results.length > 0);
+	const aborted = hasResults && details.results.some(r => r.aborted);
+	const failed = hasResults && details.results.some(r => !r.aborted && r.exitCode !== 0);
+	const mergeFailed = hasResults && details.results.some(r => !r.aborted && r.exitCode === 0 && Boolean(r.error));
+	const isError = aborted || failed;
+	const agentCount = hasResults ? details.results.length : (details.progress?.length ?? 0);
+	const icon: ToolUIStatus = options.isPartial ? "running" : isError ? "error" : mergeFailed ? "warning" : "success";
+	const header = renderStatusLine(
+		{
+			icon,
+			title: "Task",
+			meta: agentCount > 0 ? [`${agentCount} ${agentCount === 1 ? "agent" : "agents"}`] : undefined,
+		},
+		theme,
+	);
 
-	return {
-		render(width) {
-			const { expanded, isPartial, spinnerFrame } = options;
-			const key = new Hasher()
-				.bool(expanded)
-				.bool(isPartial)
-				.u32(spinnerFrame ?? 0)
-				.u32(width)
-				.digest();
-			if (cached?.key === key) return cached.lines;
+	return framedBlock(theme, width => {
+		const { expanded, isPartial, spinnerFrame } = options;
+		const lines: string[] = [];
 
-			const lines: string[] = [];
+		const shouldRenderProgress =
+			Boolean(details.progress && details.progress.length > 0) && (isPartial || details.results.length === 0);
+		if (shouldRenderProgress && details.progress) {
+			details.progress.forEach(progress => {
+				lines.push(...renderAgentProgress(progress, "", "  ", expanded, theme, spinnerFrame));
+			});
+		} else if (details.results && details.results.length > 0) {
+			details.results.forEach(res => {
+				lines.push(...renderAgentResult(res, "", "  ", expanded, theme));
+			});
 
-			const shouldRenderProgress =
-				Boolean(details.progress && details.progress.length > 0) && (isPartial || details.results.length === 0);
-			if (shouldRenderProgress && details.progress) {
-				details.progress.forEach((progress, i) => {
-					const isLast = i === details.progress!.length - 1;
-					lines.push(...renderAgentProgress(progress, isLast, expanded, theme, spinnerFrame));
-				});
-			} else if (details.results && details.results.length > 0) {
-				details.results.forEach((res, i) => {
-					const isLast = i === details.results.length - 1;
-					lines.push(...renderAgentResult(res, isLast, expanded, theme));
-				});
-
-				const abortedCount = details.results.filter(r => r.aborted).length;
-				const mergeFailedCount = details.results.filter(r => !r.aborted && r.exitCode === 0 && r.error).length;
-				const successCount = details.results.filter(r => !r.aborted && r.exitCode === 0 && !r.error).length;
-				const failCount = details.results.length - successCount - mergeFailedCount - abortedCount;
-				let summary = `${theme.fg("dim", "Total:")} `;
-				if (abortedCount > 0) {
-					summary += theme.fg("error", `${abortedCount} aborted`);
-					if (successCount > 0 || mergeFailedCount > 0 || failCount > 0) summary += theme.sep.dot;
-				}
-				if (successCount > 0) {
-					summary += theme.fg("success", `${successCount} succeeded`);
-					if (mergeFailedCount > 0 || failCount > 0) summary += theme.sep.dot;
-				}
-				if (mergeFailedCount > 0) {
-					summary += theme.fg("warning", `${mergeFailedCount} merge failed`);
-					if (failCount > 0) summary += theme.sep.dot;
-				}
-				if (failCount > 0) {
-					summary += theme.fg("error", `${failCount} failed`);
-				}
-				summary += `${theme.sep.dot}${theme.fg("dim", formatDuration(details.totalDurationMs))}`;
-				lines.push(summary);
-			}
-
-			if (lines.length === 0) {
-				const text = fallbackText.trim() ? fallbackText : "No results";
-				const result = [theme.fg("dim", truncateToWidth(text, width))];
-				cached = { key, lines: result };
-				return result;
-			}
-
-			if (fallbackText.trim()) {
-				const summaryLines = fallbackText.split("\n");
-				const markerIndex = summaryLines.findIndex(
-					line =>
-						line.includes("<system-notification>") ||
-						line.startsWith("Applied patches:") ||
-						line.startsWith("No changes to apply."),
-				);
-				if (markerIndex >= 0) {
-					const extra = summaryLines.slice(markerIndex);
-					for (const line of extra) {
-						if (!line.trim()) continue;
-						lines.push(theme.fg("dim", line));
-					}
-				}
-			}
-
-			const indented = lines.map(line =>
-				line.length > 0 ? truncateToWidth(`   ${line}`, width, Ellipsis.Omit) : "",
+			const abortedCount = details.results.filter(r => r.aborted).length;
+			const mergeFailedCount = details.results.filter(r => !r.aborted && r.exitCode === 0 && r.error).length;
+			const successCount = details.results.filter(r => !r.aborted && r.exitCode === 0 && !r.error).length;
+			const failCount = details.results.length - successCount - mergeFailedCount - abortedCount;
+			const summaryParts: string[] = [];
+			if (abortedCount > 0) summaryParts.push(theme.fg("error", `${abortedCount} aborted`));
+			if (successCount > 0) summaryParts.push(theme.fg("success", `${successCount} succeeded`));
+			if (mergeFailedCount > 0) summaryParts.push(theme.fg("warning", `${mergeFailedCount} merge failed`));
+			if (failCount > 0) summaryParts.push(theme.fg("error", `${failCount} failed`));
+			summaryParts.push(theme.fg("dim", formatDuration(details.totalDurationMs)));
+			// Wrap the run summary in the theme's bracket glyphs (dim chrome, colored
+			// counts) to match the bash tool's `[Wall: … | Exit: …]` footer.
+			lines.push(
+				theme.fg("dim", theme.format.bracketLeft) +
+					summaryParts.join(theme.fg("dim", theme.sep.dot)) +
+					theme.fg("dim", theme.format.bracketRight),
 			);
-			cached = { key, lines: indented };
-			return indented;
-		},
-		invalidate() {
-			cached = undefined;
-		},
-	};
+		}
+
+		const state = isPartial ? "running" : isError ? "error" : mergeFailed ? "warning" : "success";
+		const borderColor = isError ? "error" : "borderMuted";
+
+		if (lines.length === 0) {
+			const text = fallbackText.trim() ? fallbackText : "No results";
+			return {
+				header,
+				sections: [
+					...(contextSectionRenderer ? [contextSectionRenderer(width)] : []),
+					{ separator: true, lines: [theme.fg("dim", truncateToWidth(text, width))] },
+				],
+				state,
+				borderColor,
+				width,
+			};
+		}
+
+		if (fallbackText.trim()) {
+			const summaryLines = fallbackText.split("\n");
+			const markerIndex = summaryLines.findIndex(
+				line =>
+					line.includes("<system-notification>") ||
+					line.startsWith("Applied patches:") ||
+					line.startsWith("No changes to apply."),
+			);
+			if (markerIndex >= 0) {
+				const extra = summaryLines.slice(markerIndex);
+				for (const line of extra) {
+					if (!line.trim()) continue;
+					lines.push(theme.fg("dim", line));
+				}
+			}
+		}
+
+		while (lines.length > 0 && lines[0].trim() === "") lines.shift();
+		return {
+			header,
+			sections: [
+				...(contextSectionRenderer ? [contextSectionRenderer(width)] : []),
+				...(lines.length > 0 ? [{ separator: true, lines }] : []),
+			],
+			state,
+			borderColor,
+			width,
+		};
+	});
 }
 
 function isTaskToolDetails(value: unknown): value is TaskToolDetails {
@@ -1123,13 +1203,23 @@ function isTaskToolDetails(value: unknown): value is TaskToolDetails {
 	);
 }
 
+// Nested subagent snapshots sit one or more levels below the frame border, so
+// they keep tree guides to convey depth (the parent prepends its own continue
+// prefix). Only the top-level agent list drops guides (the frame is its box).
+function nestedMarkers(isLast: boolean, theme: Theme): { prefix: string; continuePrefix: string } {
+	return {
+		prefix: isLast ? theme.fg("dim", theme.tree.last) : theme.fg("dim", theme.tree.branch),
+		continuePrefix: isLast ? "   " : `${theme.fg("dim", theme.tree.vertical)}  `,
+	};
+}
+
 function renderNestedTaskResults(detailsList: TaskToolDetails[], expanded: boolean, theme: Theme): string[] {
 	const lines: string[] = [];
 	for (const details of detailsList) {
 		if (!details.results || details.results.length === 0) continue;
 		details.results.forEach((result, index) => {
-			const isLast = index === details.results.length - 1;
-			lines.push(...renderAgentResult(result, isLast, expanded, theme));
+			const { prefix, continuePrefix } = nestedMarkers(index === details.results.length - 1, theme);
+			lines.push(...renderAgentResult(result, prefix, continuePrefix, expanded, theme));
 		});
 	}
 	return lines;
@@ -1151,16 +1241,16 @@ function renderNestedTaskTree(
 		const hasResults = Boolean(details.results && details.results.length > 0);
 		if (hasResults) {
 			details.results.forEach((result, index) => {
-				const isLast = index === details.results.length - 1;
-				lines.push(...renderAgentResult(result, isLast, expanded, theme));
+				const { prefix, continuePrefix } = nestedMarkers(index === details.results.length - 1, theme);
+				lines.push(...renderAgentResult(result, prefix, continuePrefix, expanded, theme));
 			});
 			continue;
 		}
 		const inflight = details.progress;
 		if (inflight && inflight.length > 0) {
 			inflight.forEach((prog, index) => {
-				const isLast = index === inflight.length - 1;
-				lines.push(...renderAgentProgress(prog, isLast, expanded, theme, spinnerFrame));
+				const { prefix, continuePrefix } = nestedMarkers(index === inflight.length - 1, theme);
+				lines.push(...renderAgentProgress(prog, prefix, continuePrefix, expanded, theme, spinnerFrame));
 			});
 		}
 	}
@@ -1181,4 +1271,5 @@ subprocessToolRegistry.register<TaskToolDetails>("task", {
 export const taskToolRenderer = {
 	renderCall,
 	renderResult,
+	mergeCallAndResult: true,
 };
diff --git a/packages/coding-agent/src/telemetry-export.ts b/packages/coding-agent/src/telemetry-export.ts
index fbb5b3a62..234d43cf3 100644
--- a/packages/coding-agent/src/telemetry-export.ts
+++ b/packages/coding-agent/src/telemetry-export.ts
@@ -23,11 +23,7 @@
  * `sdk-trace-base@2.7` exports cleanly on Bun.
  */
 import { logger, postmortem } from "@oh-my-pi/pi-utils";
-import { AsyncLocalStorageContextManager } from "@opentelemetry/context-async-hooks";
-import { OTLPTraceExporter } from "@opentelemetry/exporter-trace-otlp-proto";
-import { resourceFromAttributes } from "@opentelemetry/resources";
-import { BatchSpanProcessor } from "@opentelemetry/sdk-trace-base";
-import { NodeTracerProvider } from "@opentelemetry/sdk-trace-node";
+import type * as TraceNode from "@opentelemetry/sdk-trace-node";
 
 /**
  * Periodic flush interval. A long-lived `omp` process (the ACP server is
@@ -36,7 +32,8 @@ import { NodeTracerProvider } from "@opentelemetry/sdk-trace-node";
  */
 const FLUSH_INTERVAL_MS = 30_000;
 
-let provider: NodeTracerProvider | undefined;
+let provider: TraceNode.NodeTracerProvider | undefined;
+let initPromise: Promise<void> | undefined;
 
 /**
  * Whether {@link initTelemetryExport} registered a real provider. The CLI uses
@@ -53,8 +50,10 @@ export function isTelemetryExportEnabled(): boolean {
  * the OTEL kill-switches are engaged), so it is safe to call unconditionally at
  * startup.
  */
-export function initTelemetryExport(): void {
+export async function initTelemetryExport(): Promise<void> {
 	if (provider) return;
+	if (initPromise) return initPromise;
+
 	// The OTEL env contract parses booleans and enum lists case-insensitively, so
 	// OTEL_SDK_DISABLED=TRUE and OTEL_TRACES_EXPORTER=None must also disable export.
 	if (process.env.OTEL_SDK_DISABLED?.trim().toLowerCase() === "true") return;
@@ -77,6 +76,25 @@ export function initTelemetryExport(): void {
 		return;
 	}
 
+	initPromise = registerProvider();
+	return initPromise;
+}
+
+async function registerProvider(): Promise<void> {
+	const [
+		{ AsyncLocalStorageContextManager },
+		{ OTLPTraceExporter },
+		{ resourceFromAttributes },
+		{ BatchSpanProcessor },
+		{ NodeTracerProvider },
+	] = await Promise.all([
+		import("@opentelemetry/context-async-hooks"),
+		import("@opentelemetry/exporter-trace-otlp-proto"),
+		import("@opentelemetry/resources"),
+		import("@opentelemetry/sdk-trace-base"),
+		import("@opentelemetry/sdk-trace-node"),
+	]);
+
 	// The exporter reads endpoint/headers/timeout from OTEL_EXPORTER_OTLP_* itself,
 	// so there is nothing to thread through here.
 	const exporter = new OTLPTraceExporter();
diff --git a/packages/coding-agent/src/tool-discovery/mode.ts b/packages/coding-agent/src/tool-discovery/mode.ts
new file mode 100644
index 000000000..21d4a613d
--- /dev/null
+++ b/packages/coding-agent/src/tool-discovery/mode.ts
@@ -0,0 +1,24 @@
+import type { Settings } from "../config/settings";
+import type { SettingValue } from "../config/settings-schema";
+
+export const TOOL_DISCOVERY_AUTO_THRESHOLD = 40;
+export const TOOL_DISCOVERY_SEARCH_TOOL_NAME = "search_tool_bm25";
+
+export type ToolDiscoveryModeSetting = SettingValue<"tools.discoveryMode">;
+export type EffectiveToolDiscoveryMode = Exclude<ToolDiscoveryModeSetting, "auto">;
+
+export function countToolsForAutoDiscovery(toolNames: Iterable<string>): number {
+	let count = 0;
+	for (const name of toolNames) {
+		if (name !== TOOL_DISCOVERY_SEARCH_TOOL_NAME) count++;
+	}
+	return count;
+}
+
+export function resolveEffectiveToolDiscoveryMode(settings: Settings, toolCount: number): EffectiveToolDiscoveryMode {
+	const configuredMode = settings.get("tools.discoveryMode");
+	if (configuredMode === "all" || configuredMode === "mcp-only") return configuredMode;
+	if (settings.get("mcp.discoveryMode")) return "mcp-only";
+	if (configuredMode === "auto" && toolCount > TOOL_DISCOVERY_AUTO_THRESHOLD) return "mcp-only";
+	return "off";
+}
diff --git a/packages/coding-agent/src/tools/archive-reader.ts b/packages/coding-agent/src/tools/archive-reader.ts
index 11b09b0ad..cdcd8ae63 100644
--- a/packages/coding-agent/src/tools/archive-reader.ts
+++ b/packages/coding-agent/src/tools/archive-reader.ts
@@ -1,5 +1,9 @@
+import * as fs from "node:fs/promises";
+import * as os from "node:os";
+import * as path from "node:path";
 import { inflateSync, strFromU8 } from "fflate";
 
+import { formatBytes } from "./render-utils";
 import { ToolError } from "./tool-errors";
 
 export type ArchiveFormat = "zip" | "tar" | "tar.gz";
@@ -123,11 +127,21 @@ function getArchiveFormatFromPath(filePath: string): ArchiveFormat | undefined {
 	return undefined;
 }
 
+export function formatArchiveEntryLines(entries: readonly ArchiveDirectoryEntry[]): string[] {
+	return entries.map(entry => {
+		if (entry.isDirectory) return `${entry.name}/`;
+
+		const sizeSuffix = entry.size > 0 ? ` (${formatBytes(entry.size)})` : "";
+		return `${entry.name}${sizeSuffix}`;
+	});
+}
+
 const ZIP_LOCAL_FILE_HEADER_SIGNATURE = 0x04034b50;
 const ZIP_CENTRAL_DIRECTORY_HEADER_SIGNATURE = 0x02014b50;
 const ZIP64_EOCD_SIGNATURE = 0x06064b50;
 const ZIP64_EOCD_LOCATOR_SIGNATURE = 0x07064b50;
 const ZIP_EOCD_SIGNATURE = 0x06054b50;
+const ZIP_DATA_DESCRIPTOR_SIGNATURE = 0x08074b50;
 const ZIP_EOCD_MIN_LENGTH = 22;
 const ZIP_EOCD_MAX_COMMENT_LENGTH = 0xffff;
 const ZIP64_EOCD_LOCATOR_LENGTH = 20;
@@ -167,6 +181,37 @@ function readUInt32LE(bytes: Uint8Array, offset: number): number {
 	return (bytes[offset]! | (bytes[offset + 1]! << 8) | (bytes[offset + 2]! << 16) | (bytes[offset + 3]! << 24)) >>> 0;
 }
 
+function bytesMatchAscii(bytes: Uint8Array, offset: number, value: string): boolean {
+	if (bytes.byteLength < offset + value.length) return false;
+	for (let index = 0; index < value.length; index++) {
+		if (bytes[offset + index] !== value.charCodeAt(index)) return false;
+	}
+	return true;
+}
+
+export function sniffArchiveFormat(bytes: Uint8Array): ArchiveFormat | undefined {
+	if (bytes.byteLength >= 4) {
+		const signature = readUInt32LE(bytes, 0);
+		if (
+			signature === ZIP_LOCAL_FILE_HEADER_SIGNATURE ||
+			signature === ZIP_EOCD_SIGNATURE ||
+			signature === ZIP_DATA_DESCRIPTOR_SIGNATURE
+		) {
+			return "zip";
+		}
+	}
+
+	if (bytes.byteLength >= 2 && bytes[0] === 0x1f && bytes[1] === 0x8b) {
+		return "tar.gz";
+	}
+
+	if (bytesMatchAscii(bytes, 257, "ustar")) {
+		return "tar";
+	}
+
+	return undefined;
+}
+
 function readUInt64LEAsNumber(bytes: Uint8Array, offset: number): number {
 	const value = readUInt32LE(bytes, offset) + readUInt32LE(bytes, offset + 4) * ZIP_UINT32_RANGE;
 	if (!Number.isSafeInteger(value)) {
@@ -627,3 +672,22 @@ export async function openArchive(filePath: string): Promise<ArchiveReader> {
 		format === "zip" ? await readZipEntries(filePath) : await readTarEntries(await Bun.file(filePath).bytes());
 	return new ArchiveReader(format, entries);
 }
+
+export async function listArchiveRoot(
+	bytes: Uint8Array,
+	format: ArchiveFormat,
+	opts: { limit?: number } = {},
+): Promise<string> {
+	const tempDir = await fs.mkdtemp(path.join(os.tmpdir(), "omp-archive-"));
+	const tempPath = path.join(tempDir, `payload.${format}`);
+	try {
+		await Bun.write(tempPath, bytes);
+		const archive = await openArchive(tempPath);
+		const entries = archive.listDirectory("");
+		const limitedEntries = opts.limit !== undefined && opts.limit > 0 ? entries.slice(0, opts.limit) : entries;
+		const lines = formatArchiveEntryLines(limitedEntries);
+		return lines.length > 0 ? lines.join("\n") : "(empty archive directory)";
+	} finally {
+		await fs.rm(tempDir, { recursive: true, force: true });
+	}
+}
diff --git a/packages/coding-agent/src/tools/ask.ts b/packages/coding-agent/src/tools/ask.ts
index edf7371c3..5ff09d52d 100644
--- a/packages/coding-agent/src/tools/ask.ts
+++ b/packages/coding-agent/src/tools/ask.ts
@@ -16,22 +16,14 @@
  */
 
 import type { AgentTool, AgentToolContext, AgentToolResult, AgentToolUpdateCallback } from "@oh-my-pi/pi-agent-core";
-import {
-	type Component,
-	Container,
-	Markdown,
-	type MarkdownTheme,
-	renderInlineMarkdown,
-	TERMINAL,
-	Text,
-} from "@oh-my-pi/pi-tui";
+import { type Component, Markdown, type MarkdownTheme, renderInlineMarkdown, TERMINAL, Text } from "@oh-my-pi/pi-tui";
 import { prompt, untilAborted } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import type { ExtensionUISelectItem } from "../extensibility/extensions";
 import { getMarkdownTheme, type Theme, theme } from "../modes/theme/theme";
 import askDescription from "../prompts/tools/ask.md" with { type: "text" };
-import { renderStatusLine } from "../tui";
+import { framedBlock, renderStatusLine } from "../tui";
 import type { ToolSession } from ".";
 import { formatErrorMessage, formatMeta, formatTitle } from "./render-utils";
 import { ToolAbortError } from "./tool-errors";
@@ -626,23 +618,14 @@ interface AskRenderArgs {
 	}>;
 }
 
-/** Render custom input as a single block with continuation lines (not one entry per line) */
-function renderCustomInput(
-	uiTheme: Theme,
-	prefix: string,
-	customInput: string,
-	isLastEntry: boolean,
-	includeLeadingNewline = true,
-): string {
+/** Render a custom free-text answer as a status line plus indented continuation rows. */
+function renderCustomInputLines(uiTheme: Theme, customInput: string): string[] {
 	const lines = customInput.split("\n");
-	const branch = isLastEntry ? uiTheme.tree.last : uiTheme.tree.branch;
-	const firstLine = lines[0] ?? "";
-	let text = `${includeLeadingNewline ? "\n" : ""}${prefix}${uiTheme.fg("dim", branch)} ${uiTheme.styledSymbol("status.success", "success")} ${uiTheme.fg("toolOutput", firstLine)}`;
-	const continuationIndent = isLastEntry ? "   " : `${uiTheme.fg("dim", uiTheme.tree.vertical)}  `;
-	for (let i = 1; i < lines.length; i++) {
-		text += `\n${prefix}${continuationIndent}   ${uiTheme.fg("toolOutput", lines[i])}`;
-	}
-	return text;
+	const out: string[] = [
+		` ${uiTheme.styledSymbol("status.success", "success")} ${uiTheme.fg("toolOutput", lines[0] ?? "")}`,
+	];
+	for (let i = 1; i < lines.length; i++) out.push(`   ${uiTheme.fg("toolOutput", lines[i])}`);
+	return out;
 }
 
 /**
@@ -654,25 +637,38 @@ function optionMarker(uiTheme: Theme, multi: boolean | undefined, selected: bool
 	return selected ? uiTheme.radio.selected : uiTheme.radio.unselected;
 }
 
-/**
- * Render the answered option list for a question: every offered option with its
- * selection marker filled in, plus any custom free-text answer. This keeps the
- * result visually identical to the question form (`renderCall`) so the answer
- * reads in place rather than as a detached summary block.
- *
- * `linePrefix` is the indent that precedes each entry's tree branch — a single
- * leading space for top-level (single-question) entries, or the question's
- * vertical continuation for nested (multi-question) entries.
- */
-function renderAnswerOptions(
+/** Render the offered options for a question form as flat marker bullets (no tree guides). */
+function renderQuestionOptionLines(
+	uiTheme: Theme,
+	mdTheme: MarkdownTheme,
+	options: AskRenderOption[],
+	multi: boolean | undefined,
+): string[] {
+	const out: string[] = [];
+	for (const opt of options) {
+		const optLabel = renderInlineMarkdown(opt.label, mdTheme, t => uiTheme.fg("muted", t));
+		out.push(` ${uiTheme.fg("dim", optionMarker(uiTheme, multi, false))} ${optLabel}`);
+		if (opt.description?.trim()) {
+			const description = renderInlineMarkdown(opt.description.trim(), mdTheme, t => uiTheme.fg("dim", t));
+			out.push(`   ${uiTheme.fg("dim", "↳")} ${description}`);
+		}
+	}
+	return out;
+}
+
+/**
+ * Render the answered option list for a question: every offered option with its
+ * selection marker filled in, plus any custom free-text answer. Flat marker
+ * bullets — the frame is the container, so no tree guides are drawn.
+ */
+function renderAnswerOptionLines(
 	uiTheme: Theme,
 	mdTheme: MarkdownTheme,
-	linePrefix: string,
 	options: string[] | undefined,
 	selectedOptions: string[] | undefined,
 	multi: boolean | undefined,
 	customInput: string | undefined,
-): string {
+): string[] {
 	const selected = new Set(selectedOptions ?? []);
 	// Prefer the full recorded option set; fall back to the selected labels when
 	// details omit the options array.
@@ -680,26 +676,21 @@ function renderAnswerOptions(
 
 	// Nothing was chosen (and no custom answer) → a lone cancelled marker.
 	if (selected.size === 0 && customInput === undefined) {
-		return `${linePrefix}${uiTheme.fg("dim", uiTheme.tree.last)} ${uiTheme.styledSymbol("status.warning", "warning")} ${uiTheme.fg("warning", "Cancelled")}`;
+		return [` ${uiTheme.styledSymbol("status.warning", "warning")} ${uiTheme.fg("warning", "Cancelled")}`];
 	}
 
-	let text = "";
-	for (let i = 0; i < list.length; i++) {
-		const label = list[i];
+	const out: string[] = [];
+	for (const label of list) {
 		const isSelected = selected.has(label);
-		const isLastEntry = i === list.length - 1 && customInput === undefined;
-		const branch = isLastEntry ? uiTheme.tree.last : uiTheme.tree.branch;
 		const marker = optionMarker(uiTheme, multi, isSelected);
 		const markerStyled = isSelected ? uiTheme.fg("success", marker) : uiTheme.fg("dim", marker);
 		const labelStyled = renderInlineMarkdown(label, mdTheme, t =>
 			isSelected ? uiTheme.fg("toolOutput", t) : uiTheme.fg("muted", t),
 		);
-		text += `${text ? "\n" : ""}${linePrefix}${uiTheme.fg("dim", branch)} ${markerStyled} ${labelStyled}`;
+		out.push(` ${markerStyled} ${labelStyled}`);
 	}
-	if (customInput !== undefined) {
-		text += renderCustomInput(uiTheme, linePrefix, customInput, true, text.length > 0);
-	}
-	return text;
+	if (customInput !== undefined) out.push(...renderCustomInputLines(uiTheme, customInput));
+	return out;
 }
 
 export const askToolRenderer = {
@@ -708,80 +699,58 @@ export const askToolRenderer = {
 		const label = formatTitle("Ask", uiTheme);
 		const mdTheme = getMarkdownTheme();
 		const accentStyle = { color: (t: string) => uiTheme.fg("accent", t) };
+		const md = (text: string, width: number) =>
+			new Markdown(text, 1, 0, mdTheme, accentStyle).render(Math.max(1, width - 3 + 1));
 
-		// Multi-part questions
+		// Multi-part questions: one divider-labelled section per question.
 		if (args.questions && args.questions.length > 0) {
-			const container = new Container();
-			container.addChild(new Text(`${label} ${uiTheme.fg("muted", `${args.questions.length} questions`)}`, 0, 0));
-
-			for (let i = 0; i < args.questions.length; i++) {
-				const q = args.questions[i];
-				const isLastQ = i === args.questions.length - 1;
-				const qBranch = isLastQ ? uiTheme.tree.last : uiTheme.tree.branch;
-				const continuation = isLastQ ? " " : uiTheme.tree.vertical;
-
-				const meta: string[] = [];
-				if (q.multi) meta.push("multi");
-				if (q.options?.length) meta.push(`options:${q.options.length}`);
-				const metaStr = meta.length > 0 ? uiTheme.fg("dim", ` · ${meta.join(" · ")}`) : "";
-
-				container.addChild(
-					new Text(` ${uiTheme.fg("dim", qBranch)} ${uiTheme.fg("dim", `[${q.id}]`)}${metaStr}`, 0, 0),
-				);
-				container.addChild(new Markdown(q.question, 3, 0, mdTheme, accentStyle));
-
-				if (q.options?.length) {
-					let optText = "";
-					for (let j = 0; j < q.options.length; j++) {
-						const opt = q.options[j];
-						const isLastOpt = j === q.options.length - 1;
-						const optBranch = isLastOpt ? uiTheme.tree.last : uiTheme.tree.branch;
-						const optLabel = renderInlineMarkdown(opt.label, mdTheme, t => uiTheme.fg("muted", t));
-						optText += `\n ${uiTheme.fg("dim", continuation)}   ${uiTheme.fg("dim", optBranch)} ${uiTheme.fg("dim", optionMarker(uiTheme, q.multi, false))} ${optLabel}`;
-						if (opt.description?.trim()) {
-							const optContinuation = isLastOpt ? " " : uiTheme.tree.vertical;
-							const description = renderInlineMarkdown(opt.description.trim(), mdTheme, t =>
-								uiTheme.fg("dim", t),
-							);
-							optText += `\n ${uiTheme.fg("dim", continuation)}   ${uiTheme.fg("dim", optContinuation)}   ${uiTheme.fg("dim", "↳")} ${description}`;
-						}
-					}
-					container.addChild(new Text(optText, 0, 0));
-				}
-			}
-			return container;
+			const questions = args.questions;
+			const header = `${label} ${uiTheme.fg("muted", `${questions.length} questions`)}`;
+			return framedBlock(uiTheme, width => {
+				const sections = questions.map(q => {
+					const meta: string[] = [];
+					if (q.multi) meta.push("multi");
+					if (q.options?.length) meta.push(`options:${q.options.length}`);
+					const metaStr = meta.length > 0 ? uiTheme.fg("dim", ` · ${meta.join(" · ")}`) : "";
+					const lines = md(q.question, width);
+					if (q.options?.length) lines.push(...renderQuestionOptionLines(uiTheme, mdTheme, q.options, q.multi));
+					return { label: `${uiTheme.fg("dim", `[${q.id}]`)}${metaStr}`, lines };
+				});
+				return { header, sections, state: "pending", borderColor: "borderMuted", width };
+			});
 		}
 
 		// Single question
 		if (!args.question) {
-			return new Text(formatErrorMessage("No question provided", uiTheme), 0, 0);
+			const errorLine = formatErrorMessage("No question provided", uiTheme);
+			return framedBlock(uiTheme, width => ({
+				header: errorLine,
+				sections: [],
+				state: "error",
+				borderColor: "error",
+				width,
+			}));
 		}
 
-		const container = new Container();
+		const question = args.question;
 		const meta: string[] = [];
 		if (args.multi) meta.push("multi");
 		if (args.options?.length) meta.push(`options:${args.options.length}`);
-		container.addChild(new Text(`${label}${formatMeta(meta, uiTheme)}`, 0, 0));
-		container.addChild(new Markdown(args.question, 1, 0, mdTheme, accentStyle));
-
-		if (args.options?.length) {
-			let optText = "";
-			for (let i = 0; i < args.options.length; i++) {
-				const opt = args.options[i];
-				const isLast = i === args.options.length - 1;
-				const branch = isLast ? uiTheme.tree.last : uiTheme.tree.branch;
-				const optLabel = renderInlineMarkdown(opt.label, mdTheme, t => uiTheme.fg("muted", t));
-				optText += `\n ${uiTheme.fg("dim", branch)} ${uiTheme.fg("dim", optionMarker(uiTheme, args.multi, false))} ${optLabel}`;
-				if (opt.description?.trim()) {
-					const continuation = isLast ? " " : uiTheme.tree.vertical;
-					const description = renderInlineMarkdown(opt.description.trim(), mdTheme, t => uiTheme.fg("dim", t));
-					optText += `\n ${uiTheme.fg("dim", continuation)}   ${uiTheme.fg("dim", "↳")} ${description}`;
-				}
-			}
-			container.addChild(new Text(optText, 0, 0));
-		}
-
-		return container;
+		const header = `${label}${formatMeta(meta, uiTheme)}`;
+		const questionOptions = args.options;
+		const multi = args.multi;
+		return framedBlock(uiTheme, width => {
+			const bodyLines = md(question, width);
+			if (questionOptions?.length)
+				bodyLines.push(...renderQuestionOptionLines(uiTheme, mdTheme, questionOptions, multi));
+			return {
+				header,
+				sections: bodyLines.length > 0 ? [{ lines: bodyLines }] : [],
+				state: "pending",
+				borderColor: "borderMuted",
+				width,
+			};
+		});
 	},
 
 	renderResult(
@@ -792,56 +761,47 @@ export const askToolRenderer = {
 		const { details } = result;
 		const mdTheme = getMarkdownTheme();
 		const accentStyle = { color: (t: string) => uiTheme.fg("accent", t) };
+		const md = (text: string, width: number) =>
+			new Markdown(text, 1, 0, mdTheme, accentStyle).render(Math.max(1, width - 3 + 1));
 
 		if (!details) {
 			const txt = result.content[0];
 			const fallback = txt?.type === "text" && txt.text ? txt.text : "";
 			const header = renderStatusLine({ icon: "warning", title: "Ask" }, uiTheme);
-			return new Text(`${header}\n${uiTheme.fg("dim", fallback)}`, 0, 0);
+			const body = fallback ? `\n${uiTheme.fg("dim", fallback)}` : "";
+			return new Text(`${header}${body}`, 0, 0);
 		}
 
-		// Multi-part results
+		// Multi-part results: one divider-labelled section per question.
 		if (details.results && details.results.length > 0) {
-			const hasAnySelection = details.results.some(
+			const results = details.results;
+			const hasAnySelection = results.some(
 				r => r.customInput !== undefined || (r.selectedOptions && r.selectedOptions.length > 0),
 			);
 			const header = renderStatusLine(
 				{
 					icon: hasAnySelection ? "success" : "warning",
 					title: "Ask",
-					meta: [`${details.results.length} questions`],
+					meta: [`${results.length} questions`],
 				},
 				uiTheme,
 			);
-			const container = new Container();
-			container.addChild(new Text(header, 0, 0));
-
-			for (let i = 0; i < details.results.length; i++) {
-				const r = details.results[i];
-				const isLastQuestion = i === details.results.length - 1;
-				const qBranch = isLastQuestion ? uiTheme.tree.last : uiTheme.tree.branch;
-				const continuation = isLastQuestion ? " " : uiTheme.tree.vertical;
-				const linePrefix = ` ${uiTheme.fg("dim", continuation)}   `;
-
-				container.addChild(new Text(` ${uiTheme.fg("dim", qBranch)} ${uiTheme.fg("dim", `[${r.id}]`)}`, 0, 0));
-				container.addChild(new Markdown(r.question, 3, 0, mdTheme, accentStyle));
-				container.addChild(
-					new Text(
-						renderAnswerOptions(
-							uiTheme,
-							mdTheme,
-							linePrefix,
-							r.options,
-							r.selectedOptions,
-							r.multi,
-							r.customInput,
-						),
-						0,
-						0,
-					),
-				);
-			}
-			return container;
+			return framedBlock(uiTheme, width => {
+				const sections = results.map(r => {
+					const lines = md(r.question, width);
+					lines.push(
+						...renderAnswerOptionLines(uiTheme, mdTheme, r.options, r.selectedOptions, r.multi, r.customInput),
+					);
+					return { label: uiTheme.fg("dim", `[${r.id}]`), lines };
+				});
+				return {
+					header,
+					sections,
+					state: hasAnySelection ? "success" : "warning",
+					borderColor: "borderMuted",
+					width,
+				};
+			});
 		}
 
 		// Single question result
@@ -851,29 +811,24 @@ export const askToolRenderer = {
 			return new Text(fallback, 0, 0);
 		}
 
+		const question = details.question;
 		const hasSelection =
 			details.customInput !== undefined || (details.selectedOptions && details.selectedOptions.length > 0);
 		const header = renderStatusLine({ icon: hasSelection ? "success" : "warning", title: "Ask" }, uiTheme);
-		const container = new Container();
-		container.addChild(new Text(header, 0, 0));
-		container.addChild(new Markdown(details.question, 1, 0, mdTheme, accentStyle));
-
-		container.addChild(
-			new Text(
-				renderAnswerOptions(
-					uiTheme,
-					mdTheme,
-					" ",
-					details.options,
-					details.selectedOptions,
-					details.multi,
-					details.customInput,
-				),
-				0,
-				0,
-			),
-		);
-
-		return container;
+		const dOptions = details.options;
+		const dSelected = details.selectedOptions;
+		const dMulti = details.multi;
+		const dCustom = details.customInput;
+		return framedBlock(uiTheme, width => {
+			const bodyLines = md(question, width);
+			bodyLines.push(...renderAnswerOptionLines(uiTheme, mdTheme, dOptions, dSelected, dMulti, dCustom));
+			return {
+				header,
+				sections: bodyLines.length > 0 ? [{ lines: bodyLines }] : [],
+				state: hasSelection ? "success" : "warning",
+				borderColor: "borderMuted",
+				width,
+			};
+		});
 	},
 };
diff --git a/packages/coding-agent/src/tools/ast-edit.ts b/packages/coding-agent/src/tools/ast-edit.ts
index 4cc1d9e13..eac7b6e92 100644
--- a/packages/coding-agent/src/tools/ast-edit.ts
+++ b/packages/coding-agent/src/tools/ast-edit.ts
@@ -3,7 +3,7 @@ import { formatHashlineHeader } from "@oh-my-pi/hashline";
 import type { AgentTool, AgentToolContext, AgentToolResult, AgentToolUpdateCallback } from "@oh-my-pi/pi-agent-core";
 import { type AstReplaceChange, type AstReplaceFileChange, astEdit } from "@oh-my-pi/pi-natives";
 import type { Component } from "@oh-my-pi/pi-tui";
-import { Text } from "@oh-my-pi/pi-tui";
+import { replaceTabs, Text } from "@oh-my-pi/pi-tui";
 import { $envpos, prompt, untilAborted } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
 import { getFileSnapshotStore } from "../edit/file-snapshot-store";
@@ -11,26 +11,24 @@ import { normalizeToLF } from "../edit/normalize";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import type { Theme } from "../modes/theme/theme";
 import astEditDescription from "../prompts/tools/ast-edit.md" with { type: "text" };
-import { Ellipsis, fileHyperlink, renderStatusLine, renderTreeList, truncateToWidth } from "../tui";
+import { Ellipsis, fileHyperlink, framedBlock, renderStatusLine, truncateToWidth } from "../tui";
 import { resolveFileDisplayMode } from "../utils/file-display-mode";
 import type { ToolSession } from ".";
 import { truncateForPrompt } from "./approval";
 import { createFileRecorder, formatResultPath } from "./file-recorder";
-import { formatGroupedFiles } from "./grouped-file-output";
+import { classifyGroupedLines, formatGroupedFiles, groupLineIndicesByBlank } from "./grouped-file-output";
 import type { OutputMeta } from "./output-meta";
 import { isInternalUrlPath, resolveToolSearchScope } from "./path-utils";
 import {
 	appendParseErrorsBulletList,
 	capParseErrors,
-	createCachedComponent,
 	formatCodeFrameLine,
 	formatCount,
-	formatEmptyMessage,
-	formatErrorMessage,
+	formatErrorDetail,
+	formatMoreItems,
 	formatParseErrors,
 	formatParseErrorsCountLabel,
 	PREVIEW_LIMITS,
-	splitGroupsByBlankLine,
 } from "./render-utils";
 import { queueResolveHandler } from "./resolve";
 import { ToolError } from "./tool-errors";
@@ -161,6 +159,9 @@ export interface AstEditToolDetails {
 	/** Absolute base directory used during the edit. Used by the renderer to resolve
 	 * display-relative paths to absolute paths for OSC 8 hyperlinks. */
 	searchPath?: string;
+	/** Session cwd at edit time. Display header paths are cwd-relative, so the
+	 * renderer resolves them against this; `searchPath` is the scope target. */
+	cwd?: string;
 }
 
 export class AstEditTool implements AgentTool<typeof astEditSchema, AstEditToolDetails> {
@@ -274,6 +275,7 @@ export class AstEditTool implements AgentTool<typeof astEditSchema, AstEditToolD
 				...(cappedParseErrors.length > 0 ? { parseErrors: cappedParseErrors, parseErrorsTotal } : {}),
 				scopePath,
 				searchPath: resolvedSearchPath,
+				cwd: this.session.cwd,
 				files: fileList,
 				fileReplacements: [],
 			};
@@ -464,6 +466,32 @@ interface AstEditRenderArgs {
 
 const COLLAPSED_CHANGE_LIMIT = PREVIEW_LIMITS.COLLAPSED_LINES * 2;
 
+/**
+ * Flatten pre-styled change groups into frame body lines. Groups are separated
+ * by a blank line and carry no tree guides — the frame border is the container,
+ * so nested `├─ │` gutters would just be noise. Collapsed mode always shows at
+ * least the first group, then fills up to `budget` lines before summarizing the
+ * rest as `… N more changes`.
+ */
+function buildChangeBody(groups: string[][], expanded: boolean, budget: number, theme: Theme): string[] {
+	const lines: string[] = [];
+	let shown = 0;
+	for (let i = 0; i < groups.length; i++) {
+		const group = groups[i]!;
+		const separator = shown > 0 ? 1 : 0;
+		const remainingAfter = groups.length - (i + 1);
+		const reserved = !expanded && remainingAfter > 0 ? 1 : 0;
+		// Always emit the first group; budget only gates subsequent ones.
+		if (!expanded && shown > 0 && lines.length + separator + group.length + reserved > budget) break;
+		if (separator) lines.push("");
+		lines.push(...group);
+		shown++;
+	}
+	const remaining = groups.length - shown;
+	if (!expanded && remaining > 0) lines.push(theme.fg("muted", formatMoreItems(remaining, "change")));
+	return lines;
+}
+
 export const astEditToolRenderer = {
 	inline: true,
 	renderCall(args: AstEditRenderArgs, _options: RenderResultOptions, uiTheme: Theme): Component {
@@ -473,8 +501,9 @@ export const astEditToolRenderer = {
 		if (rewriteCount > 1) meta.push(`${rewriteCount} rewrites`);
 
 		const description = rewriteCount === 1 ? args.ops?.[0]?.pat : rewriteCount ? `${rewriteCount} rewrites` : "?";
-		const text = renderStatusLine({ icon: "pending", title: "AST Edit", description, meta }, uiTheme);
-		return new Text(text, 0, 0);
+		const header = renderStatusLine({ icon: "pending", title: "AST Edit", description, meta }, uiTheme);
+		// Pending call has no body yet — a lone status line is sleeker than an empty frame.
+		return new Text(header, 0, 0);
 	},
 
 	renderResult(
@@ -487,7 +516,14 @@ export const astEditToolRenderer = {
 
 		if (result.isError) {
 			const errorText = result.content?.find(c => c.type === "text")?.text || "Unknown error";
-			return new Text(formatErrorMessage(errorText, uiTheme), 0, 0);
+			const header = renderStatusLine({ icon: "error", title: "AST Edit" }, uiTheme);
+			return framedBlock(uiTheme, width => ({
+				header,
+				sections: [{ lines: formatErrorDetail(errorText, uiTheme).split("\n") }],
+				state: "error",
+				borderColor: "error",
+				width,
+			}));
 		}
 
 		const totalReplacements = details?.totalReplacements ?? 0;
@@ -502,9 +538,18 @@ export const astEditToolRenderer = {
 			if (details?.scopePath) meta.push(`in ${details.scopePath}`);
 			if (filesSearched > 0) meta.push(`searched ${filesSearched}`);
 			const header = renderStatusLine({ icon: "warning", title: "AST Edit", description, meta }, uiTheme);
-			const lines = [header, formatEmptyMessage("No replacements made", uiTheme)];
-			appendParseErrorsBulletList(lines, details?.parseErrors, uiTheme, details?.parseErrorsTotal);
-			return new Text(lines.join("\n"), 0, 0);
+			// The "0 replacements" count already rides on the status line; only parse
+			// errors are worth a body, so frame solely when there are some.
+			const bodyLines: string[] = [];
+			appendParseErrorsBulletList(bodyLines, details?.parseErrors, uiTheme, details?.parseErrorsTotal);
+			if (bodyLines.length === 0) return new Text(header, 0, 0);
+			return framedBlock(uiTheme, width => ({
+				header,
+				sections: [{ lines: bodyLines }],
+				state: "warning",
+				borderColor: "borderMuted",
+				width,
+			}));
 		}
 
 		const summaryParts = [formatCount("replacement", totalReplacements), formatCount("file", filesTouched)];
@@ -516,10 +561,33 @@ export const astEditToolRenderer = {
 		const description = rewriteCount === 1 ? args?.ops?.[0]?.pat : undefined;
 
 		const textContent = result.details?.displayContent ?? result.content?.find(c => c.type === "text")?.text ?? "";
-		const allGroups = splitGroupsByBlankLine(textContent.split("\n"));
-		const changeGroups = allGroups.filter(
-			group => !group[0]?.startsWith("Safety cap reached") && !group[0]?.startsWith("Parse issues:"),
-		);
+		const allLines = textContent.split("\n");
+		// Resolve hyperlinks over the whole output so nested directory headers
+		// reconstruct across the blank-line groups the tree list collapses by.
+		const contexts = classifyGroupedLines(allLines, details?.cwd ?? details?.searchPath, details?.searchPath);
+		const styledLines = allLines.map((line, index) => {
+			const ctx = contexts[index]!;
+			// Swap the inner code-frame gutter `│` for a space so it does not nest a
+			// second vertical bar inside the frame border.
+			const display = replaceTabs(line.replace("│", " "));
+			if (ctx.kind === "dir") {
+				const styled = uiTheme.fg("accent", display);
+				return ctx.headerPath ? fileHyperlink(ctx.headerPath, styled) : styled;
+			}
+			if (ctx.kind === "file") {
+				const styled = uiTheme.fg(ctx.depth === 1 ? "accent" : "dim", display);
+				return ctx.headerPath ? fileHyperlink(ctx.headerPath, styled) : styled;
+			}
+			if (display.startsWith("+")) return uiTheme.fg("toolDiffAdded", display);
+			if (display.startsWith("-")) return uiTheme.fg("toolDiffRemoved", display);
+			return uiTheme.fg("toolOutput", display);
+		});
+		const changeGroups = groupLineIndicesByBlank(allLines)
+			.filter(indices => {
+				const first = allLines[indices[0]!]!;
+				return !first.startsWith("Safety cap reached") && !first.startsWith("Parse issues:");
+			})
+			.map(indices => indices.map(index => styledLines[index]!));
 
 		const badge = { label: "proposed", color: "warning" as const };
 		const header = renderStatusLine(
@@ -536,60 +604,19 @@ export const astEditToolRenderer = {
 				uiTheme.fg("warning", formatParseErrorsCountLabel(details.parseErrors, details.parseErrorsTotal)),
 			);
 		}
-		return createCachedComponent(
-			() => options.expanded,
-			width => {
-				const searchBase = details?.searchPath;
-				const changeLines = renderTreeList(
-					{
-						items: changeGroups,
-						expanded: options.expanded,
-						maxCollapsed: changeGroups.length,
-						maxCollapsedLines: COLLAPSED_CHANGE_LIMIT,
-						itemType: "change",
-						renderItem: group => {
-							let contextDir = searchBase ?? "";
-							return group.map(line => {
-								if (line.startsWith("## ")) {
-									// Strip ` (3 replacements)` and `#hash` suffixes from formatGroupedFiles.
-									const fileName = line
-										.slice(3)
-										.trimEnd()
-										.replace(/\s+\([^)]*\)\s*$/, "")
-										.replace(/#[0-9a-f]+$/, "");
-									const absPath = contextDir && fileName ? path.join(contextDir, fileName) : undefined;
-									const styled = uiTheme.fg("dim", line);
-									return absPath ? fileHyperlink(absPath, styled) : styled;
-								}
-								if (line.startsWith("# ")) {
-									const raw = line
-										.slice(2)
-										.trimEnd()
-										.replace(/\s+\([^)]*\)\s*$/, "");
-									const isDirectory = raw.endsWith("/");
-									const name = isDirectory ? raw.replace(/\/$/, "") : raw.replace(/#[0-9a-f]+$/, "");
-									if (isDirectory) {
-										if (searchBase) {
-											contextDir = name === "." ? searchBase : path.join(searchBase, name);
-										}
-										return uiTheme.fg("accent", line);
-									}
-									// Root-level file with optional `#hash` and ` (3 replacements)` suffixes.
-									const absPath = searchBase && name ? path.join(searchBase, name) : undefined;
-									const styled = uiTheme.fg("accent", line);
-									return absPath ? fileHyperlink(absPath, styled) : styled;
-								}
-								if (line.startsWith("+")) return uiTheme.fg("toolDiffAdded", line);
-								if (line.startsWith("-")) return uiTheme.fg("toolDiffRemoved", line);
-								return uiTheme.fg("toolOutput", line);
-							});
-						},
-					},
-					uiTheme,
-				);
-				return [header, ...changeLines, ...extraLines].map(l => truncateToWidth(l, width, Ellipsis.Omit));
-			},
-		);
+		return framedBlock(uiTheme, width => {
+			const changeLines = buildChangeBody(changeGroups, Boolean(options.expanded), COLLAPSED_CHANGE_LIMIT, uiTheme);
+			const innerWidth = Math.max(1, width - 3);
+			const bodyLines = [...changeLines, ...extraLines].map(l => truncateToWidth(l, innerWidth, Ellipsis.Omit));
+			while (bodyLines.length > 0 && bodyLines[0].trim() === "") bodyLines.shift();
+			return {
+				header,
+				sections: bodyLines.length > 0 ? [{ lines: bodyLines }] : [],
+				state: options.isPartial ? "pending" : "success",
+				borderColor: "borderMuted",
+				width,
+			};
+		});
 	},
 	mergeCallAndResult: true,
 };
diff --git a/packages/coding-agent/src/tools/ast-grep.ts b/packages/coding-agent/src/tools/ast-grep.ts
index 3354c4af5..3935915d4 100644
--- a/packages/coding-agent/src/tools/ast-grep.ts
+++ b/packages/coding-agent/src/tools/ast-grep.ts
@@ -14,7 +14,7 @@ import { Ellipsis, fileHyperlink, renderStatusLine, renderTreeList, truncateToWi
 import { resolveFileDisplayMode } from "../utils/file-display-mode";
 import type { ToolSession } from ".";
 import { createFileRecorder, formatResultPath } from "./file-recorder";
-import { formatGroupedFiles } from "./grouped-file-output";
+import { classifyGroupedLines, formatGroupedFiles, groupLineIndicesByBlank } from "./grouped-file-output";
 import { formatMatchLine } from "./match-line-format";
 import type { OutputMeta } from "./output-meta";
 import { resolveToolSearchScope } from "./path-utils";
@@ -29,7 +29,6 @@ import {
 	formatParseErrors,
 	formatParseErrorsCountLabel,
 	PREVIEW_LIMITS,
-	splitGroupsByBlankLine,
 } from "./render-utils";
 import { ToolError } from "./tool-errors";
 import { toolResult } from "./tool-result";
@@ -118,6 +117,9 @@ export interface AstGrepToolDetails {
 	/** Absolute base directory used during search. Used by the renderer to resolve
 	 * display-relative paths to absolute paths for OSC 8 hyperlinks. */
 	searchPath?: string;
+	/** Session cwd at search time. Display header/match paths are cwd-relative, so
+	 * the renderer resolves them against this; `searchPath` is the scope target. */
+	cwd?: string;
 }
 
 export class AstGrepTool implements AgentTool<typeof astGrepSchema, AstGrepToolDetails> {
@@ -207,6 +209,7 @@ export class AstGrepTool implements AgentTool<typeof astGrepSchema, AstGrepToolD
 				...(cappedParseErrors.length > 0 ? { parseErrors: cappedParseErrors, parseErrorsTotal } : {}),
 				scopePath,
 				searchPath: resolvedSearchPath,
+				cwd: this.session.cwd,
 				files: fileList,
 				fileMatches: [],
 			};
@@ -381,15 +384,41 @@ export const astGrepToolRenderer = {
 		if (limitReached) meta.push(uiTheme.fg("warning", "limit reached"));
 		const description = args?.pat;
 		const header = renderStatusLine(
-			{ icon: limitReached ? "warning" : "success", title: "AST Grep", description, meta },
+			{
+				...(limitReached
+					? { icon: "warning" as const }
+					: { iconOverride: uiTheme.fg("accent", uiTheme.symbol("icon.search")) }),
+				title: "AST Grep",
+				description,
+				meta,
+			},
 			uiTheme,
 		);
 
 		const textContent = result.details?.displayContent ?? result.content?.find(c => c.type === "text")?.text ?? "";
-		const allGroups = splitGroupsByBlankLine(textContent.split("\n"));
-		const matchGroups = allGroups.filter(
-			group => !group[0]?.startsWith("Result limit reached") && !group[0]?.startsWith("Parse issues:"),
-		);
+		const allLines = textContent.split("\n");
+		// Resolve hyperlinks over the whole output so nested directory headers
+		// reconstruct across the blank-line groups the tree list collapses by.
+		const contexts = classifyGroupedLines(allLines, details?.cwd ?? details?.searchPath, details?.searchPath);
+		const styledLines = allLines.map((line, index) => {
+			const ctx = contexts[index]!;
+			if (ctx.kind === "dir") {
+				const styled = uiTheme.fg("accent", line);
+				return ctx.headerPath ? fileHyperlink(ctx.headerPath, styled) : styled;
+			}
+			if (ctx.kind === "file") {
+				const styled = uiTheme.fg(ctx.depth === 1 ? "accent" : "dim", line);
+				return ctx.headerPath ? fileHyperlink(ctx.headerPath, styled) : styled;
+			}
+			if (line.startsWith("  meta:")) return uiTheme.fg("dim", line);
+			return uiTheme.fg("toolOutput", line);
+		});
+		const matchGroups = groupLineIndicesByBlank(allLines)
+			.filter(indices => {
+				const first = allLines[indices[0]!]!;
+				return !first.startsWith("Result limit reached") && !first.startsWith("Parse issues:");
+			})
+			.map(indices => indices.map(index => styledLines[index]!));
 
 		const extraLines: string[] = [];
 		if (limitReached) {
@@ -404,7 +433,6 @@ export const astGrepToolRenderer = {
 		return createCachedComponent(
 			() => options.expanded,
 			width => {
-				const searchBase = details?.searchPath;
 				const matchLines = renderTreeList(
 					{
 						items: matchGroups,
@@ -412,41 +440,7 @@ export const astGrepToolRenderer = {
 						maxCollapsed: matchGroups.length,
 						maxCollapsedLines: COLLAPSED_MATCH_LIMIT,
 						itemType: "match",
-						renderItem: group => {
-							let contextDir = searchBase ?? "";
-							return group.map(line => {
-								if (line.startsWith("## ")) {
-									const fileName = line
-										.slice(3)
-										.trimEnd()
-										.replace(/\s+\([^)]*\)\s*$/, "")
-										.replace(/#[0-9a-f]+$/, "");
-									const absPath = contextDir && fileName ? path.join(contextDir, fileName) : undefined;
-									const styled = uiTheme.fg("dim", line);
-									return absPath ? fileHyperlink(absPath, styled) : styled;
-								}
-								if (line.startsWith("# ")) {
-									const raw = line
-										.slice(2)
-										.trimEnd()
-										.replace(/\s+\([^)]*\)\s*$/, "");
-									const isDirectory = raw.endsWith("/");
-									const name = isDirectory ? raw.replace(/\/$/, "") : raw.replace(/#[0-9a-f]+$/, "");
-									if (isDirectory) {
-										if (searchBase) {
-											contextDir = name === "." ? searchBase : path.join(searchBase, name);
-										}
-										return uiTheme.fg("accent", line);
-									}
-									// Root-level file (single # without trailing slash) from formatGroupedFiles.
-									const absPath = searchBase && name ? path.join(searchBase, name) : undefined;
-									const styled = uiTheme.fg("accent", line);
-									return absPath ? fileHyperlink(absPath, styled) : styled;
-								}
-								if (line.startsWith("  meta:")) return uiTheme.fg("dim", line);
-								return uiTheme.fg("toolOutput", line);
-							});
-						},
+						renderItem: group => group,
 					},
 					uiTheme,
 				);
diff --git a/packages/coding-agent/src/tools/bash.ts b/packages/coding-agent/src/tools/bash.ts
index d4e227620..0238695a8 100644
--- a/packages/coding-agent/src/tools/bash.ts
+++ b/packages/coding-agent/src/tools/bash.ts
@@ -20,7 +20,7 @@ import bashDescription from "../prompts/tools/bash.md" with { type: "text" };
 import type { ClientBridgeTerminalExitStatus, ClientBridgeTerminalOutput } from "../session/client-bridge";
 import { DEFAULT_MAX_BYTES, streamTailUpdates, TailBuffer } from "../session/streaming-output";
 import { renderStatusLine } from "../tui";
-import { CachedOutputBlock } from "../tui/output-block";
+import { CachedOutputBlock, markFramedBlockComponent } from "../tui/output-block";
 import { getSixelLineMask } from "../utils/sixel";
 import type { ToolSession } from ".";
 import { truncateForPrompt } from "./approval";
@@ -31,7 +31,7 @@ import { canUseInteractiveBashPty } from "./bash-pty-selection";
 import { expandInternalUrls, type InternalUrlExpansionOptions } from "./bash-skill-urls";
 import { formatStyledTruncationWarning, type OutputMeta, stripOutputNotice } from "./output-meta";
 import { resolveToCwd } from "./path-utils";
-import { formatToolWorkingDirectory, replaceTabs } from "./render-utils";
+import { capPreviewLines, formatToolWorkingDirectory, replaceTabs } from "./render-utils";
 import { ToolAbortError, ToolError } from "./tool-errors";
 import { toolResult } from "./tool-result";
 import { clampTimeout, TOOL_TIMEOUTS } from "./tool-timeouts";
@@ -287,6 +287,35 @@ function formatExitCodeNotice(exitCode: number): string {
 	return `Command exited with code ${exitCode}`;
 }
 
+const RAW_OUTPUT_ARTIFACT_PREFIX = "[raw output: artifact://";
+const RAW_OUTPUT_ARTIFACT_SUFFIX = "]";
+
+function stripRawOutputArtifactNotice(text: string): { text: string; artifactId?: string } {
+	const trimmed = text.trimEnd();
+	const lineStart = trimmed.lastIndexOf("\n");
+	const candidateStart = lineStart === -1 ? 0 : lineStart + 1;
+	if (
+		!trimmed.startsWith(RAW_OUTPUT_ARTIFACT_PREFIX, candidateStart) ||
+		!trimmed.endsWith(RAW_OUTPUT_ARTIFACT_SUFFIX)
+	) {
+		return { text };
+	}
+
+	const idStart = candidateStart + RAW_OUTPUT_ARTIFACT_PREFIX.length;
+	const idEnd = trimmed.length - RAW_OUTPUT_ARTIFACT_SUFFIX.length;
+	if (idStart === idEnd) return { text };
+	for (let i = idStart; i < idEnd; i++) {
+		const code = trimmed.charCodeAt(i);
+		if (code < 48 || code > 57) return { text };
+	}
+
+	const artifactId = trimmed.slice(idStart, idEnd);
+	return {
+		text: trimmed.slice(0, lineStart === -1 ? 0 : lineStart).trimEnd(),
+		artifactId,
+	};
+}
+
 /**
  * Strip the trailing occurrence of `notice` (plus a single surrounding newline
  * on each side) so the TUI can echo the value via a styled footer label
@@ -1028,6 +1057,7 @@ export interface ShellRendererConfig<TArgs> {
 	resolveCommand?: (args: TArgs | undefined) => string | undefined;
 	resolveCwd?: (args: TArgs | undefined) => string | undefined;
 	resolveEnv?: (args: TArgs | undefined) => Record<string, string> | undefined;
+	showHeader?: boolean;
 }
 
 function getPartialJson<TArgs>(args: TArgs | undefined): string | undefined {
@@ -1079,20 +1109,28 @@ export function createShellRenderer<TArgs>(config: ShellRendererConfig<TArgs>) {
 	return {
 		renderCall(args: TArgs, options: RenderResultOptions, uiTheme: Theme): Component {
 			const renderArgs = toBashRenderArgs(args, config);
-			const title = config.resolveTitle(args, options);
 			const cmdLines = formatBashCommandLines(renderArgs, uiTheme);
-			const header = renderStatusLine({ icon: "pending", title }, uiTheme);
+			const header =
+				config.showHeader === false
+					? undefined
+					: renderStatusLine({ icon: "pending", title: config.resolveTitle(args, options) }, uiTheme);
 			const outputBlock = new CachedOutputBlock();
-			return {
+			return markFramedBlockComponent({
 				render: (width: number): string[] =>
 					outputBlock.render(
-						{ header, state: "pending", sections: [{ lines: cmdLines }], width, animate: true },
+						{
+							header,
+							state: "pending",
+							sections: [{ lines: capPreviewLines(cmdLines, uiTheme, { expanded: options.expanded }) }],
+							width,
+							animate: true,
+						},
 						uiTheme,
 					),
 				invalidate: () => {
 					outputBlock.invalidate();
 				},
-			};
+			});
 		},
 
 		renderResult(
@@ -1109,12 +1147,14 @@ export function createShellRenderer<TArgs>(config: ShellRendererConfig<TArgs>) {
 			const cmdLines = args ? formatBashCommandLines(renderArgs, uiTheme) : undefined;
 			const isError = result.isError === true;
 			const icon = options.isPartial ? "pending" : isError ? "error" : "success";
-			const title = config.resolveTitle(args, options);
-			const header = renderStatusLine({ icon, title }, uiTheme);
+			const header =
+				config.showHeader === false
+					? undefined
+					: renderStatusLine({ icon, title: config.resolveTitle(args, options) }, uiTheme);
 			const details = result.details;
 			const outputBlock = new CachedOutputBlock();
 
-			return {
+			return markFramedBlockComponent({
 				render: (width: number): string[] => {
 					// REACTIVE: read mutable options at render time
 					const { renderContext } = options;
@@ -1127,7 +1167,9 @@ export function createShellRenderer<TArgs>(config: ShellRendererConfig<TArgs>) {
 					const rawOutput = renderContext?.output ?? result.content?.find(c => c.type === "text")?.text ?? "";
 					const strippedOutput = stripOutputNotice(rawOutput, details?.meta);
 					const withoutExit = stripExitCodeNotice(strippedOutput, details?.exitCode);
-					const output = stripWallTimeNotice(withoutExit, details?.wallTimeMs);
+					const withoutWall = stripWallTimeNotice(withoutExit, details?.wallTimeMs);
+					const rawOutputArtifact = stripRawOutputArtifactNotice(withoutWall);
+					const output = rawOutputArtifact.text;
 					const displayOutput = output.trimEnd();
 					const showingFullOutput = expanded && renderContext?.isFullOutput === true;
 
@@ -1146,6 +1188,9 @@ export function createShellRenderer<TArgs>(config: ShellRendererConfig<TArgs>) {
 								: `Timeout: ${timeoutSeconds}s`,
 						);
 					}
+					if (rawOutputArtifact.artifactId) {
+						statsParts.push(`Artifact: ${rawOutputArtifact.artifactId}`);
+					}
 					if (isError && typeof details?.exitCode === "number") {
 						statsParts.push(`Exit: ${details.exitCode}`);
 					}
@@ -1201,11 +1246,19 @@ export function createShellRenderer<TArgs>(config: ShellRendererConfig<TArgs>) {
 							header,
 							state: options.isPartial ? "pending" : isError ? "error" : "success",
 							sections: [
-								{ lines: cmdLines ?? [] },
+								{
+									lines: options.isPartial
+										? capPreviewLines(cmdLines ?? [], uiTheme, { expanded })
+										: (cmdLines ?? []),
+								},
 								{ label: uiTheme.fg("toolTitle", "Output"), lines: outputLines },
 							],
 							width,
-							animate: options.isPartial && shimmerEnabled(),
+							// Don't animate once the command has been backgrounded: the block
+							// gets committed to scrollback and finalizes later via the async
+							// update path, so a mid-sweep frame would freeze a stray dark
+							// border segment.
+							animate: options.isPartial && shimmerEnabled() && details?.async?.state !== "running",
 						},
 						uiTheme,
 					);
@@ -1213,7 +1266,7 @@ export function createShellRenderer<TArgs>(config: ShellRendererConfig<TArgs>) {
 				invalidate: () => {
 					outputBlock.invalidate();
 				},
-			};
+			});
 		},
 		mergeCallAndResult: true,
 		inline: true,
@@ -1225,4 +1278,5 @@ export const bashToolRenderer = createShellRenderer<BashRenderArgs>({
 	resolveCommand: args => args?.command,
 	resolveCwd: args => args?.cwd,
 	resolveEnv: args => args?.env,
+	showHeader: false,
 });
diff --git a/packages/coding-agent/src/tools/browser/render.ts b/packages/coding-agent/src/tools/browser/render.ts
index 91c362917..e2b84bb61 100644
--- a/packages/coding-agent/src/tools/browser/render.ts
+++ b/packages/coding-agent/src/tools/browser/render.ts
@@ -9,7 +9,7 @@ import type { Component } from "@oh-my-pi/pi-tui";
 import { Text } from "@oh-my-pi/pi-tui";
 import type { RenderResultOptions } from "../../extensibility/custom-tools/types";
 import type { Theme } from "../../modes/theme/theme";
-import { Hasher, renderCodeCell, renderStatusLine } from "../../tui";
+import { Hasher, isFramedBlockComponent, markFramedBlockComponent, renderCodeCell, renderStatusLine } from "../../tui";
 import type { BrowserToolDetails } from "../browser";
 import { formatStyledTruncationWarning, stripOutputNotice } from "../output-meta";
 import { replaceTabs, shortenPath } from "../render-utils";
@@ -65,13 +65,14 @@ function dropTrailingBlankLines(text: string): string {
 
 function appendLine(component: Component, line: string | undefined): Component {
 	if (!line) return component;
-	return {
+	const wrapped = {
 		render: (width: number): string[] => {
 			const base = component.render(width);
 			return [...base, line];
 		},
 		invalidate: () => component.invalidate?.(),
 	};
+	return isFramedBlockComponent(component) ? markFramedBlockComponent(wrapped) : wrapped;
 }
 
 function renderRunCell(
@@ -93,7 +94,7 @@ function renderRunCell(
 	const title = titleParts.join(" · ");
 
 	let cached: { key: bigint; width: number; lines: string[] } | undefined;
-	return {
+	return markFramedBlockComponent({
 		render: (width: number): string[] => {
 			const expanded = options.renderContext?.expanded ?? options.expanded;
 			const previewLines = options.renderContext?.previewLines ?? BROWSER_DEFAULT_PREVIEW_LINES;
@@ -131,7 +132,7 @@ function renderRunCell(
 		invalidate: () => {
 			cached = undefined;
 		},
-	};
+	});
 }
 
 function renderOpenOrCloseLine(
diff --git a/packages/coding-agent/src/tools/debug.ts b/packages/coding-agent/src/tools/debug.ts
index 3089fbef7..61a009d2c 100644
--- a/packages/coding-agent/src/tools/debug.ts
+++ b/packages/coding-agent/src/tools/debug.ts
@@ -22,6 +22,7 @@ import {
 	type DapFunctionBreakpointRecord,
 	type DapInstructionBreakpointRecord,
 	type DapModule,
+	type DapResolvedAdapter,
 	type DapScope,
 	type DapSessionSummary,
 	type DapSource,
@@ -30,13 +31,15 @@ import {
 	type DapVariable,
 	dapSessionManager,
 	getAvailableAdapters,
+	type LaunchProgramKind,
+	resolveLaunchOverrides,
 	selectAttachAdapter,
 	selectLaunchAdapter,
 } from "../dap";
 import type { Theme } from "../modes/theme/theme";
 import debugDescription from "../prompts/tools/debug.md" with { type: "text" };
 import { renderStatusLine } from "../tui";
-import { CachedOutputBlock } from "../tui/output-block";
+import { CachedOutputBlock, markFramedBlockComponent } from "../tui/output-block";
 import type { ToolSession } from ".";
 import { truncateForPrompt } from "./approval";
 import type { OutputMeta } from "./output-meta";
@@ -489,16 +492,23 @@ function getConfiguredAdapters(cwd: string): string {
 	const adapters = getAvailableAdapters(cwd).map(adapter => adapter.name);
 	return adapters.length > 0 ? adapters.join(", ") : "none";
 }
-async function validateLaunchProgram(program: string, cwd: string): Promise<void> {
-	let isDirectory: boolean;
+
+async function classifyLaunchProgram(program: string): Promise<LaunchProgramKind> {
 	try {
-		isDirectory = (await fs.stat(program)).isDirectory();
+		return (await fs.stat(program)).isDirectory() ? "directory" : "file";
 	} catch (error) {
-		if (isEnoent(error)) return;
+		if (isEnoent(error)) return "missing";
 		throw error;
 	}
-	if (!isDirectory) return;
+}
 
+function validateLaunchProgram(
+	program: string,
+	cwd: string,
+	programKind: LaunchProgramKind,
+	adapter: DapResolvedAdapter,
+): void {
+	if (programKind !== "directory" || adapter.acceptsDirectoryProgram) return;
 	const displayPath = formatPathRelativeToCwd(program, cwd, { trailingSlash: true });
 	throw new ToolError(
 		`launch program resolves to a directory: ${displayPath}. Pass an executable file path, or for Python use adapter "debugpy" with program set to the .py file.`,
@@ -581,7 +591,7 @@ export const debugToolRenderer = {
 		args?: DebugRenderArgs,
 	): Component {
 		const outputBlock = new CachedOutputBlock();
-		return {
+		return markFramedBlockComponent({
 			render(width: number): string[] {
 				const action = (args?.action ?? result.details?.action ?? "debug").replaceAll("_", " ");
 				const status = options.isPartial ? "running" : result.isError ? "error" : "success";
@@ -620,7 +630,7 @@ export const debugToolRenderer = {
 			invalidate() {
 				outputBlock.invalidate();
 			},
-		};
+		});
 	},
 	mergeCallAndResult: true,
 	inline: true,
@@ -676,8 +686,8 @@ export class DebugTool implements AgentTool<typeof debugSchema, DebugToolDetails
 				}
 				const commandCwd = params.cwd ? resolveToCwd(params.cwd, this.session.cwd) : this.session.cwd;
 				const program = resolveToCwd(params.program, commandCwd);
-				await validateLaunchProgram(program, commandCwd);
-				const adapter = selectLaunchAdapter(program, commandCwd, params.adapter);
+				const programKind = await classifyLaunchProgram(program);
+				const adapter = selectLaunchAdapter(program, commandCwd, params.adapter, programKind);
 				if (!adapter) {
 					if (params.adapter === "debugpy") {
 						throw new ToolError("adapter 'debugpy' is not available: python not found in PATH");
@@ -686,8 +696,10 @@ export class DebugTool implements AgentTool<typeof debugSchema, DebugToolDetails
 						`No debugger adapter available. Installed adapters: ${getConfiguredAdapters(commandCwd)}`,
 					);
 				}
+				validateLaunchProgram(program, commandCwd, programKind, adapter);
+				const extraLaunchArguments = resolveLaunchOverrides(adapter, program, programKind);
 				const snapshot = await dapSessionManager.launch(
-					{ adapter, program, args: params.args, cwd: commandCwd },
+					{ adapter, program, args: params.args, cwd: commandCwd, extraLaunchArguments },
 					combinedSignal,
 					timeoutSec * 1000,
 				);
diff --git a/packages/coding-agent/src/tools/eval-backends.ts b/packages/coding-agent/src/tools/eval-backends.ts
index 7f9cb6e7e..7839387d9 100644
--- a/packages/coding-agent/src/tools/eval-backends.ts
+++ b/packages/coding-agent/src/tools/eval-backends.ts
@@ -1,4 +1,4 @@
-import { $env, $flag } from "@oh-my-pi/pi-utils";
+import { $flag } from "@oh-my-pi/pi-utils";
 import type { ToolSession } from ".";
 
 export interface EvalBackendsAllowance {
@@ -6,21 +6,6 @@ export interface EvalBackendsAllowance {
 	js: boolean;
 }
 
-/**
- * Parse PI_PY / PI_JS environment variables. Each is a boolean flag; unset
- * means "not specified, defer to settings". Returns null when neither is set
- * so the caller can fall through to `readEvalBackendsAllowance` per key.
- */
-function getEvalBackendsFromEnv(): EvalBackendsAllowance | null {
-	const pyEnv = $env.PI_PY;
-	const jsEnv = $env.PI_JS;
-	if (pyEnv === undefined && jsEnv === undefined) return null;
-	return {
-		python: pyEnv === undefined ? true : $flag("PI_PY"),
-		js: jsEnv === undefined ? true : $flag("PI_JS"),
-	};
-}
-
 /** Read per-backend allowance from settings (defaults true). */
 export function readEvalBackendsAllowance(session: ToolSession): EvalBackendsAllowance {
 	return {
@@ -34,5 +19,9 @@ export function readEvalBackendsAllowance(session: ToolSession): EvalBackendsAll
  * override the per-key settings; otherwise settings (defaults true) win.
  */
 export function resolveEvalBackends(session: ToolSession): EvalBackendsAllowance {
-	return getEvalBackendsFromEnv() ?? readEvalBackendsAllowance(session);
+	const settings = readEvalBackendsAllowance(session);
+	return {
+		python: $flag("PI_PY", settings.python),
+		js: $flag("PI_JS", settings.js),
+	};
 }
diff --git a/packages/coding-agent/src/tools/eval-render.ts b/packages/coding-agent/src/tools/eval-render.ts
index 6beeb223e..d9379bba2 100644
--- a/packages/coding-agent/src/tools/eval-render.ts
+++ b/packages/coding-agent/src/tools/eval-render.ts
@@ -18,7 +18,7 @@ import { formatContextUsage } from "../modes/components/status-line/context-thre
 import { truncateToVisualLines } from "../modes/components/visual-truncate";
 import { shimmerEnabled } from "../modes/theme/shimmer";
 import { getMarkdownTheme, type Theme } from "../modes/theme/theme";
-import { borderShimmerTick, renderCodeCell } from "../tui";
+import { borderShimmerTick, markFramedBlockComponent, renderCodeCell } from "../tui";
 import {
 	JSON_TREE_MAX_DEPTH_COLLAPSED,
 	JSON_TREE_MAX_DEPTH_EXPANDED,
@@ -39,7 +39,6 @@ import {
 	truncateToWidth,
 	wrapBrackets,
 } from "./render-utils";
-
 export const EVAL_DEFAULT_PREVIEW_LINES = 10;
 
 function languageForHighlighter(language: EvalLanguage | undefined): "python" | "javascript" {
@@ -490,10 +489,10 @@ export const evalToolRenderer = {
 
 		let cached: { key: string; width: number; result: string[] } | undefined;
 
-		return {
+		return markFramedBlockComponent({
 			render: (width: number): string[] => {
 				const animate = options.isPartial && shimmerEnabled();
-				const key = `${animate ? borderShimmerTick() : 0}|${cells.map(c => `${c.language}:${c.title ?? ""}:${c.code.length}`).join("|")}`;
+				const key = `${animate ? borderShimmerTick() : 0}|${options.expanded ? 1 : 0}|${cells.map(c => `${c.language}:${c.title ?? ""}:${c.code.length}`).join("|")}`;
 				if (cached && cached.key === key && cached.width === width) {
 					return cached.result;
 				}
@@ -510,8 +509,13 @@ export const evalToolRenderer = {
 							title: cell.title,
 							status: "pending",
 							width,
-							codeMaxLines: EVAL_DEFAULT_PREVIEW_LINES,
-							expanded: true,
+							// Always render the full source: the code is fixed input, not the
+							// streaming part, so it is never compacted. While still pending
+							// (args streaming) the block is not yet committed to native
+							// scrollback — its head is only committed once a result exists and
+							// the code has finalized (see `isStreamingPreviewAppendOnly`).
+							codeMaxLines: Number.POSITIVE_INFINITY,
+							expanded: options.expanded,
 							animate,
 						},
 						uiTheme,
@@ -527,7 +531,7 @@ export const evalToolRenderer = {
 			invalidate: () => {
 				cached = undefined;
 			},
-		};
+		});
 	},
 
 	renderResult(
@@ -571,7 +575,7 @@ export const evalToolRenderer = {
 		if (cellResults && cellResults.length > 0) {
 			let cached: { key: string; width: number; result: string[] } | undefined;
 
-			return {
+			return markFramedBlockComponent({
 				render: (width: number): string[] => {
 					const expanded = options.renderContext?.expanded ?? options.expanded;
 					const previewLines = options.renderContext?.previewLines ?? EVAL_DEFAULT_PREVIEW_LINES;
@@ -613,7 +617,9 @@ export const evalToolRenderer = {
 								duration: cell.durationMs,
 								output: outputLines.length > 0 ? outputLines.join("\n") : undefined,
 								outputMaxLines: outputLines.length,
-								codeMaxLines: expanded ? Number.POSITIVE_INFINITY : EVAL_DEFAULT_PREVIEW_LINES,
+								// Code is fixed input — always shown in full, never compacted.
+								// Only `output` honors the collapsed preview cap above.
+								codeMaxLines: Number.POSITIVE_INFINITY,
 								expanded,
 								width,
 								animate,
@@ -649,7 +655,7 @@ export const evalToolRenderer = {
 				invalidate: () => {
 					cached = undefined;
 				},
-			};
+			});
 		}
 
 		const displayOutput = output;
@@ -745,6 +751,18 @@ export const evalToolRenderer = {
 			},
 		};
 	},
+
+	// Append-only once a result exists (args complete → code finalized). The code
+	// is rendered in full as a fixed top-anchored prefix, and the streamed stdout
+	// below it only appends rows at the bottom, so the scrolled-off head commits
+	// to native scrollback instead of being yanked — collapsed or expanded, since
+	// the collapsed output cap keeps its sliding tail in the bottom live region.
+	// Returns false while still pending: the code is mid-stream (args incomplete)
+	// and its header still reads "pending", so committing it would strand a stale
+	// pending preview in history.
+	isStreamingPreviewAppendOnly(_args: EvalRenderArgs, _options: RenderResultOptions, result?: unknown): boolean {
+		return result != null;
+	},
 	mergeCallAndResult: true,
 	inline: true,
 };
diff --git a/packages/coding-agent/src/tools/eval.ts b/packages/coding-agent/src/tools/eval.ts
index 4b09016bc..cb8a6b80d 100644
--- a/packages/coding-agent/src/tools/eval.ts
+++ b/packages/coding-agent/src/tools/eval.ts
@@ -4,7 +4,7 @@ import { prompt } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
 import { jsBackend, pythonBackend } from "../eval";
 import type { ExecutorBackend, ExecutorBackendResult } from "../eval/backend";
-import { EVAL_HEARTBEAT_OP } from "../eval/heartbeat";
+import { EVAL_TIMEOUT_PAUSE_OP, EVAL_TIMEOUT_RESUME_OP } from "../eval/bridge-timeout";
 import { IdleTimeout } from "../eval/idle-timeout";
 import { defaultEvalSessionId } from "../eval/session-id";
 import type { EvalCellResult, EvalDisplayOutput, EvalLanguage, EvalStatusEvent, EvalToolDetails } from "../eval/types";
@@ -130,11 +130,12 @@ function timeoutSecondsFromMs(timeoutMs: number): number {
 }
 
 async function resolveBackend(session: ToolSession, language: EvalLanguage): Promise<ResolvedBackend> {
-	const allowPy = (session.settings.get("eval.py") as boolean | undefined) ?? true;
-	const allowJs = (session.settings.get("eval.js") as boolean | undefined) ?? true;
+	const backends = resolveEvalBackends(session);
+	const allowPy = backends.python;
+	const allowJs = backends.js;
 
 	if (language === "python") {
-		if (!allowPy) throw new ToolError("Python backend is disabled (eval.py = false).");
+		if (!allowPy) throw new ToolError("Python backend is disabled (PI_PY=0 or eval.py = false).");
 		if (!(await pythonBackend.isAvailable(session))) {
 			throw new ToolError(
 				'Python backend is unavailable in this session. Pass language: "js" or install the python kernel.',
@@ -142,7 +143,7 @@ async function resolveBackend(session: ToolSession, language: EvalLanguage): Pro
 		}
 		return { backend: pythonBackend };
 	}
-	if (!allowJs) throw new ToolError("JavaScript backend is disabled (eval.js = false).");
+	if (!allowJs) throw new ToolError("JavaScript backend is disabled (PI_JS=0 or eval.js = false).");
 	return { backend: jsBackend };
 }
 
@@ -313,16 +314,13 @@ export class EvalTool implements AgentTool<typeof evalSchema> {
 				for (let i = 0; i < cells.length; i++) {
 					const cell = cells[i];
 					const backend = cell.resolved.backend;
-					// The per-cell `timeout` is a wall-clock budget on the cell's *own*
-					// work, but it is paused while a host-side `agent()`/`llm()` bridge
-					// call is in flight: those calls pump a heartbeat (see
-					// `withBridgeHeartbeat`) that re-arms the watchdog, so a long fanout
-					// or a slow completion runs to completion. Nothing else re-arms it —
-					// compute, stdout, `log()`/`phase()`, and ordinary tool calls all
-					// count against the budget — so a cell that is not delegating to an
-					// agent/llm is bounded by a plain wall-clock timeout. The watchdog
-					// drives `combinedSignal`; we pass no wall-clock deadline downstream
-					// so the backends never arm a competing fixed timer.
+					// The per-cell `timeout` is a budget on the cell runtime's *own*
+					// work. Host-side `agent()`/`parallel()`/`llm()` bridge calls suspend
+					// that budget entirely and restart a fresh timeout window when control
+					// returns to Python/JS. Compute, stdout, `log()`/`phase()`, and
+					// ordinary tool calls all count against the budget. The watchdog drives
+					// `combinedSignal`; we pass no wall-clock deadline downstream so the
+					// backends never arm a competing fixed timer.
 					const idleTimeoutMs = timeoutSecondsFromMs(cell.timeoutMs) * 1000;
 					const idle = new IdleTimeout(idleTimeoutMs);
 					const combinedSignal = signal
@@ -355,14 +353,12 @@ export class EvalTool implements AgentTool<typeof evalSchema> {
 								outputSink!.push(chunk);
 							},
 							onStatus: event => {
-								// Only a bridge heartbeat re-arms the watchdog: it is the
-								// keepalive `agent()`/`llm()` pump while a host-side call is
-								// in flight, so those calls effectively pause the budget. It
-								// carries no payload — bump and drop it. Every other event
-								// (compute helpers, log()/phase(), tool results) renders but
-								// counts against the plain wall-clock budget.
-								if (event.op === EVAL_HEARTBEAT_OP) {
-									idle.bump();
+								if (event.op === EVAL_TIMEOUT_PAUSE_OP) {
+									idle.pause();
+									return;
+								}
+								if (event.op === EVAL_TIMEOUT_RESUME_OP) {
+									idle.resume();
 									return;
 								}
 								cellResult.statusEvents ??= [];
diff --git a/packages/coding-agent/src/tools/fetch.ts b/packages/coding-agent/src/tools/fetch.ts
index 5fc359646..4d16ced81 100644
--- a/packages/coding-agent/src/tools/fetch.ts
+++ b/packages/coding-agent/src/tools/fetch.ts
@@ -1,4 +1,6 @@
+import { Database } from "bun:sqlite";
 import * as fs from "node:fs/promises";
+import * as os from "node:os";
 import * as path from "node:path";
 import type { AgentToolResult } from "@oh-my-pi/pi-agent-core";
 import type { ImageContent, TextContent } from "@oh-my-pi/pi-ai";
@@ -8,13 +10,14 @@ import { $which, ptree, truncate } from "@oh-my-pi/pi-utils";
 import { parseHTML } from "linkedom";
 import { LRUCache } from "lru-cache/raw";
 import type { Settings } from "../config/settings";
+import { readEditableNotebookText } from "../edit/notebook";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import { type Theme, theme } from "../modes/theme/theme";
 import type { ToolSession } from "../sdk";
 import type { AgentStorage } from "../session/agent-storage";
 import { DEFAULT_MAX_BYTES, truncateHead } from "../session/streaming-output";
 import { renderStatusLine, urlHyperlink } from "../tui";
-import { CachedOutputBlock } from "../tui/output-block";
+import { CachedOutputBlock, markFramedBlockComponent } from "../tui/output-block";
 import { formatDimensionNote, resizeImage } from "../utils/image-resize";
 import { ensureTool } from "../utils/tools-manager";
 import { extractWithParallel, findParallelApiKey, getParallelExtractContent } from "../web/parallel";
@@ -22,10 +25,12 @@ import { specialHandlers } from "../web/scrapers";
 import type { RenderResult } from "../web/scrapers/types";
 import { finalizeOutput, loadPage, looksLikeHtml, MAX_OUTPUT_CHARS } from "../web/scrapers/types";
 import { convertWithMarkit, fetchBinary } from "../web/scrapers/utils";
+import { type ArchiveFormat, listArchiveRoot, sniffArchiveFormat } from "./archive-reader";
 import { applyListLimit } from "./list-limit";
 import { formatStyledArtifactReference, type OutputMeta } from "./output-meta";
 import { type LineRange, parseLineRanges } from "./path-utils";
-import { formatExpandHint, getDomain, replaceTabs } from "./render-utils";
+import { formatBytes, formatExpandHint, getDomain, replaceTabs } from "./render-utils";
+import { listTables, looksLikeSqlite, renderTableList } from "./sqlite-reader";
 import { ToolAbortError, ToolError } from "./tool-errors";
 import { toolResult } from "./tool-result";
 import { clampTimeout } from "./tool-timeouts";
@@ -46,8 +51,6 @@ const CONVERTIBLE_MIMES = new Set([
 	"application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
 	"application/rtf",
 	"application/epub+zip",
-	"application/x-ipynb+json",
-	"application/zip",
 	"image/png",
 	"image/jpeg",
 	"image/gif",
@@ -67,7 +70,6 @@ const CONVERTIBLE_EXTENSIONS = new Set([
 	".xlsx",
 	".rtf",
 	".epub",
-	".ipynb",
 	".png",
 	".jpg",
 	".jpeg",
@@ -78,6 +80,27 @@ const CONVERTIBLE_EXTENSIONS = new Set([
 	".ogg",
 ]);
 
+const NOTEBOOK_MIMES = new Set(["application/x-ipynb+json"]);
+const NOTEBOOK_EXTENSIONS = new Set([".ipynb"]);
+
+const SQLITE_MIMES = new Set([
+	"application/vnd.sqlite3",
+	"application/x-sqlite3",
+	"application/sqlite3",
+	"application/sqlite",
+]);
+const SQLITE_EXTENSIONS = new Set([".sqlite", ".sqlite3", ".db", ".db3"]);
+
+const ARCHIVE_MIMES = new Set([
+	"application/zip",
+	"application/x-zip-compressed",
+	"application/x-tar",
+	"application/tar",
+	"application/gzip",
+	"application/x-gzip",
+]);
+const ARCHIVE_EXTENSIONS = new Set([".zip", ".tar", ".tar.gz", ".tgz", ".gz"]);
+
 const IMAGE_MIME_BY_EXTENSION = new Map<string, string>([
 	[".png", "image/png"],
 	[".jpg", "image/jpeg"],
@@ -261,6 +284,12 @@ function normalizeMime(contentType: string): string {
 	return contentType.split(";")[0].trim().toLowerCase();
 }
 
+function getFilenameExtensionHint(filename: string): string {
+	const lower = filename.toLowerCase();
+	if (lower.endsWith(".tar.gz")) return ".tar.gz";
+	return path.extname(filename).toLowerCase();
+}
+
 /**
  * Get extension from URL or Content-Disposition
  */
@@ -269,7 +298,7 @@ function getExtensionHint(url: string, contentDisposition?: string): string {
 	if (contentDisposition) {
 		const match = contentDisposition.match(/filename[*]?=["']?([^"';\n]+)/i);
 		if (match) {
-			const ext = path.extname(match[1]).toLowerCase();
+			const ext = getFilenameExtensionHint(match[1]);
 			if (ext) return ext;
 		}
 	}
@@ -277,7 +306,7 @@ function getExtensionHint(url: string, contentDisposition?: string): string {
 	// Fall back to URL path
 	try {
 		const pathname = new URL(url).pathname;
-		const ext = path.extname(pathname).toLowerCase();
+		const ext = getFilenameExtensionHint(pathname);
 		if (ext) return ext;
 	} catch {}
 
@@ -581,14 +610,25 @@ function parseFeedToMarkdown(content: string, maxItems = 10): string {
  */
 const REMOTE_READER_MAX_MS = 10_000;
 
+/** Reader backends for {@link renderHtmlToText}, in default priority order. */
+export type FetchProvider = "native" | "trafilatura" | "lynx" | "parallel" | "jina";
+
+const FETCH_PROVIDER_ORDER: readonly FetchProvider[] = ["native", "trafilatura", "lynx", "parallel", "jina"];
+
 /**
- * Render HTML to markdown using Parallel, jina, trafilatura, lynx, then the
- * in-process native converter. The overall `timeout` budget bounds the call,
- * but remote reader requests are additionally capped at `REMOTE_READER_MAX_MS`
- * so that a hung remote endpoint cannot prevent local fallbacks from running.
- * Only a real `userSignal` cancellation aborts the chain — remote per-attempt
- * timeouts and the overall reader-mode timeout still allow later renderers
- * (especially the purely-local native converter) to be tried.
+ * Render HTML to markdown by trying reader backends in priority order: native
+ * (in-process), trafilatura, lynx, Parallel, then Jina. The `providers.fetch`
+ * setting picks the order — `auto` uses the default above; any specific backend
+ * is tried first, then the remaining backends as fallbacks. Every backend's
+ * output must clear the same quality gate (>100 non-whitespace chars and not
+ * {@link isLowQualityOutput}) before it is accepted, otherwise the next backend
+ * is tried.
+ *
+ * The overall `timeout` budget bounds the whole call; remote backends (Parallel,
+ * Jina) are additionally capped at `REMOTE_READER_MAX_MS` so a hung endpoint
+ * cannot starve later renderers — especially the purely-local native converter,
+ * which always works on already-loaded HTML. Only a real `userSignal`
+ * cancellation aborts the chain (#1449).
  */
 export async function renderHtmlToText(
 	url: string,
@@ -607,92 +647,74 @@ export async function renderHtmlToText(
 		signal: overallSignal,
 	};
 	const remoteBudgetMs = Math.min(timeout * 1000, REMOTE_READER_MAX_MS);
+	// Per-attempt budget for remote endpoints so one stall cannot consume the
+	// whole reader-mode budget and starve the local fallbacks.
+	const remoteSignal = () => ptree.combineSignals(userSignal, remoteBudgetMs);
 
-	// Try Parallel extract first when credentials are configured
-	if (settings.get("providers.parallelFetch") && findParallelApiKey(storage)) {
-		try {
+	const runners: Record<FetchProvider, () => Promise<string | null>> = {
+		// Purely local, no network/subprocess: still works on already-loaded HTML
+		// even after remote/subprocess attempts are aborted by the budget.
+		native: () => htmlToMarkdown(html, { cleanContent: true }),
+		trafilatura: async () => {
+			const trafilatura = await ensureTool("trafilatura", { signal: overallSignal, silent: true });
+			if (!trafilatura) return null;
+			const result = await ptree.exec([trafilatura, "-u", url, "--output-format", "markdown"], execOptions);
+			return result.ok ? result.stdout : null;
+		},
+		lynx: async () => {
+			if (!hasCommand("lynx")) return null;
+			const result = await ptree.exec(["lynx", "-dump", "-nolist", "-width", "250", url], execOptions);
+			return result.ok ? result.stdout : null;
+		},
+		parallel: async () => {
+			if (!findParallelApiKey(storage)) return null;
 			const parallelResult = await extractWithParallel(
 				[url],
-				{
-					objective: "Extract the main content",
-					excerpts: true,
-					fullContent: false,
-					signal: ptree.combineSignals(userSignal, remoteBudgetMs),
-				},
+				{ objective: "Extract the main content", excerpts: true, fullContent: false, signal: remoteSignal() },
 				storage,
 			);
 			const firstDocument = parallelResult.results[0];
-			if (firstDocument) {
-				const content = getParallelExtractContent(firstDocument);
-				if (content.trim().length > 100 && !isLowQualityOutput(content)) {
-					return { content, ok: true, method: "parallel" };
-				}
+			return firstDocument ? getParallelExtractContent(firstDocument) : null;
+		},
+		jina: async () => {
+			const response = await fetch(`https://r.jina.ai/${url}`, {
+				headers: { Accept: "text/markdown" },
+				signal: remoteSignal(),
+			});
+			return response.ok ? await response.text() : null;
+		},
+	};
+
+	const preference = settings.get("providers.fetch");
+	const order: readonly FetchProvider[] =
+		preference === "auto"
+			? FETCH_PROVIDER_ORDER
+			: [preference, ...FETCH_PROVIDER_ORDER.filter(method => method !== preference)];
+
+	// Highest-priority output that is substantial but fails the low-quality gate.
+	// Surfaced (ok: true) only when no backend clears the gate, so the caller's
+	// targeted fallbacks (llms.txt / document extraction) still run and we beat
+	// returning the unrendered raw HTML.
+	let lowQuality: { content: string; method: FetchProvider } | null = null;
+
+	for (const method of order) {
+		// Honour real user cancellation between attempts; remote per-attempt and
+		// overall-budget timeouts still fall through to later (local) renderers.
+		userSignal?.throwIfAborted();
+		try {
+			const content = await runners[method]();
+			if (!content || content.trim().length <= 100) continue;
+			if (!isLowQualityOutput(content)) {
+				return { content, ok: true, method };
 			}
+			lowQuality ??= { content, method };
 		} catch {
-			// Parallel extract failed or stalled; honour real cancellation only.
 			userSignal?.throwIfAborted();
 		}
 	}
 
-	// Try jina reader API with its own sub-budget so a stall cannot starve
-	// later fallbacks (#1449).
-	try {
-		const jinaUrl = `https://r.jina.ai/${url}`;
-		const response = await fetch(jinaUrl, {
-			headers: { Accept: "text/markdown" },
-			signal: ptree.combineSignals(userSignal, remoteBudgetMs),
-		});
-		if (response.ok) {
-			const content = await response.text();
-			if (content.trim().length > 100 && !isLowQualityOutput(content)) {
-				return { content, ok: true, method: "jina" };
-			}
-		}
-	} catch {
-		// Jina failed or stalled; honour real cancellation only.
-		userSignal?.throwIfAborted();
-	}
-
-	// Try trafilatura (auto-install via uv/pip)
-	try {
-		const trafilatura = await ensureTool("trafilatura", { signal: overallSignal, silent: true });
-		if (trafilatura) {
-			const result = await ptree.exec([trafilatura, "-u", url, "--output-format", "markdown"], execOptions);
-			if (result.ok && result.stdout.trim().length > 100) {
-				return { content: result.stdout, ok: true, method: "trafilatura" };
-			}
-		}
-	} catch {
-		// trafilatura unavailable or stalled; continue to next method.
-		userSignal?.throwIfAborted();
-	}
-
-	// Try lynx (can't auto-install, system package)
-	try {
-		const lynx = hasCommand("lynx");
-		if (lynx) {
-			const result = await ptree.exec(["lynx", "-dump", "-nolist", "-width", "250", url], execOptions);
-			if (result.ok) {
-				return { content: result.stdout, ok: true, method: "lynx" };
-			}
-		}
-	} catch {
-		// lynx failed or stalled; continue to native converter.
-		userSignal?.throwIfAborted();
-	}
-
-	// Fall back to native converter (purely local, no network/subprocess).
-	// Always attempted: even if remote renderers and subprocesses were aborted
-	// by the overall reader-mode timeout, this still works on already-loaded
-	// HTML (#1449).
-	try {
-		const content = await htmlToMarkdown(html, { cleanContent: true });
-		if (content.trim().length > 100 && !isLowQualityOutput(content)) {
-			return { content, ok: true, method: "native" };
-		}
-	} catch {
-		// Native converter failed; nothing else to try.
-		userSignal?.throwIfAborted();
+	if (lowQuality) {
+		return { content: lowQuality.content, ok: true, method: lowQuality.method };
 	}
 	return { content: "", ok: false, method: "none" };
 }
@@ -745,6 +767,254 @@ type FetchRenderResult = RenderResult & {
 	image?: FetchImagePayload;
 };
 
+const BINARY_SAMPLE_CHARS = 4096;
+const URL_ARCHIVE_LIST_LIMIT = 500;
+const URL_SQLITE_LIST_LIMIT = 500;
+
+function sampleLooksBinary(text: string): boolean {
+	const limit = Math.min(text.length, BINARY_SAMPLE_CHARS);
+	if (limit === 0) return false;
+
+	let replacementCount = 0;
+	for (let index = 0; index < limit; index++) {
+		const code = text.charCodeAt(index);
+		if (code === 0) return true;
+		if (code === 0xfffd) replacementCount++;
+	}
+
+	return replacementCount >= 3 && replacementCount / limit > 0.01;
+}
+
+function isNotebookHint(mime: string, extensionHint: string): boolean {
+	return NOTEBOOK_MIMES.has(mime) || NOTEBOOK_EXTENSIONS.has(extensionHint);
+}
+
+function isSqliteHint(mime: string, extensionHint: string): boolean {
+	return SQLITE_MIMES.has(mime) || SQLITE_EXTENSIONS.has(extensionHint);
+}
+
+function isArchiveHint(mime: string, extensionHint: string): boolean {
+	return ARCHIVE_MIMES.has(mime) || ARCHIVE_EXTENSIONS.has(extensionHint);
+}
+
+function getArchiveFormatHint(mime: string, extensionHint: string): ArchiveFormat | undefined {
+	if (extensionHint === ".zip" || mime === "application/zip" || mime === "application/x-zip-compressed") {
+		return "zip";
+	}
+	if (extensionHint === ".tar" || mime === "application/x-tar" || mime === "application/tar") {
+		return "tar";
+	}
+	if (
+		extensionHint === ".tar.gz" ||
+		extensionHint === ".tgz" ||
+		extensionHint === ".gz" ||
+		mime === "application/gzip" ||
+		mime === "application/x-gzip"
+	) {
+		return "tar.gz";
+	}
+	return undefined;
+}
+
+function formatErrorMessage(error: unknown): string {
+	return error instanceof Error ? error.message : String(error);
+}
+
+function binaryContentType(mime: string): string {
+	return mime || "application/octet-stream";
+}
+
+function buildBinaryNotice(finalUrl: string, mime: string, byteLength?: number): string {
+	const size = byteLength === undefined ? "unknown size" : formatBytes(byteLength);
+	return `[Binary content: ${binaryContentType(mime)}, ${size}] ${finalUrl}`;
+}
+
+function buildBinaryPayloadResult(
+	url: string,
+	finalUrl: string,
+	mime: string,
+	method: string,
+	content: string,
+	fetchedAt: string,
+	notes: string[],
+): FetchRenderResult {
+	const output = finalizeOutput(content);
+	return {
+		url,
+		finalUrl,
+		contentType: binaryContentType(mime),
+		method,
+		content: output.content,
+		fetchedAt,
+		truncated: output.truncated,
+		notes,
+	};
+}
+
+async function withTempBinaryFile<T>(
+	prefix: string,
+	extension: string,
+	bytes: Uint8Array,
+	readTempFile: (tempPath: string) => Promise<T>,
+): Promise<T> {
+	const tempDir = await fs.mkdtemp(path.join(os.tmpdir(), prefix));
+	const tempPath = path.join(tempDir, `payload${extension}`);
+	try {
+		await Bun.write(tempPath, bytes);
+		return await readTempFile(tempPath);
+	} finally {
+		await fs.rm(tempDir, { recursive: true, force: true });
+	}
+}
+
+async function renderNotebookPayload(bytes: Uint8Array, displayUrl: string): Promise<string> {
+	return withTempBinaryFile("omp-url-notebook-", ".ipynb", bytes, tempPath =>
+		readEditableNotebookText(tempPath, displayUrl),
+	);
+}
+
+async function renderSqlitePayload(bytes: Uint8Array): Promise<string> {
+	return withTempBinaryFile("omp-url-sqlite-", ".sqlite", bytes, async tempPath => {
+		let db: Database | null = null;
+		try {
+			db = new Database(tempPath, { readonly: true, strict: true });
+			db.run("PRAGMA busy_timeout = 3000");
+			const listLimit = applyListLimit(listTables(db), { limit: URL_SQLITE_LIST_LIMIT });
+			return renderTableList(listLimit.items);
+		} finally {
+			db?.close();
+		}
+	});
+}
+
+async function tryRenderBinaryPayload(
+	url: string,
+	finalUrl: string,
+	mime: string,
+	extHint: string,
+	rawContent: string,
+	timeout: number,
+	signal: AbortSignal | undefined,
+	fetchedAt: string,
+	notes: readonly string[],
+): Promise<FetchRenderResult | null> {
+	const hasNotebookHint = isNotebookHint(mime, extHint);
+	const hasSqliteHint = isSqliteHint(mime, extHint);
+	const hasArchiveHint = isArchiveHint(mime, extHint);
+	const rawLooksBinary = sampleLooksBinary(rawContent);
+	if (!hasNotebookHint && !hasSqliteHint && !hasArchiveHint && !rawLooksBinary) {
+		return null;
+	}
+
+	const resultNotes = [...notes];
+	const binary = await fetchBinary(finalUrl, timeout, signal);
+	if (!binary.ok) {
+		resultNotes.push(binary.error ? `Binary fetch failed: ${binary.error}` : "Binary fetch failed");
+		return buildBinaryPayloadResult(
+			url,
+			finalUrl,
+			mime,
+			"binary",
+			buildBinaryNotice(finalUrl, mime),
+			fetchedAt,
+			resultNotes,
+		);
+	}
+
+	const binaryExtHint = getExtensionHint(finalUrl, binary.contentDisposition) || extHint;
+	if (isNotebookHint(mime, binaryExtHint)) {
+		try {
+			return buildBinaryPayloadResult(
+				url,
+				finalUrl,
+				mime,
+				"notebook",
+				await renderNotebookPayload(binary.buffer, finalUrl),
+				fetchedAt,
+				resultNotes,
+			);
+		} catch (error) {
+			resultNotes.push(`Notebook rendering failed: ${formatErrorMessage(error)}`);
+			return buildBinaryPayloadResult(
+				url,
+				finalUrl,
+				mime,
+				"binary",
+				buildBinaryNotice(finalUrl, mime, binary.buffer.byteLength),
+				fetchedAt,
+				resultNotes,
+			);
+		}
+	}
+
+	if (isSqliteHint(mime, binaryExtHint) || looksLikeSqlite(binary.buffer)) {
+		try {
+			return buildBinaryPayloadResult(
+				url,
+				finalUrl,
+				mime,
+				"sqlite",
+				await renderSqlitePayload(binary.buffer),
+				fetchedAt,
+				resultNotes,
+			);
+		} catch (error) {
+			resultNotes.push(`SQLite rendering failed: ${formatErrorMessage(error)}`);
+			return buildBinaryPayloadResult(
+				url,
+				finalUrl,
+				mime,
+				"binary",
+				buildBinaryNotice(finalUrl, mime, binary.buffer.byteLength),
+				fetchedAt,
+				resultNotes,
+			);
+		}
+	}
+
+	const hintedArchiveFormat = getArchiveFormatHint(mime, binaryExtHint);
+	const shouldArchiveSniff = hintedArchiveFormat !== undefined || !isConvertible(mime, binaryExtHint);
+	const archiveFormat = hintedArchiveFormat ?? (shouldArchiveSniff ? sniffArchiveFormat(binary.buffer) : undefined);
+	if (archiveFormat) {
+		try {
+			return buildBinaryPayloadResult(
+				url,
+				finalUrl,
+				mime,
+				"archive",
+				await listArchiveRoot(binary.buffer, archiveFormat, { limit: URL_ARCHIVE_LIST_LIMIT }),
+				fetchedAt,
+				resultNotes,
+			);
+		} catch (error) {
+			resultNotes.push(`Archive rendering failed: ${formatErrorMessage(error)}`);
+			return buildBinaryPayloadResult(
+				url,
+				finalUrl,
+				mime,
+				"binary",
+				buildBinaryNotice(finalUrl, mime, binary.buffer.byteLength),
+				fetchedAt,
+				resultNotes,
+			);
+		}
+	}
+
+	if (rawLooksBinary) {
+		return buildBinaryPayloadResult(
+			url,
+			finalUrl,
+			mime,
+			"binary",
+			buildBinaryNotice(finalUrl, mime, binary.buffer.byteLength),
+			fetchedAt,
+			resultNotes,
+		);
+	}
+
+	return null;
+}
+
 // =============================================================================
 // Unified Special Handler Dispatch
 // =============================================================================
@@ -991,6 +1261,19 @@ async function renderUrl(
 		}
 	}
 
+	const binaryPayloadResult = await tryRenderBinaryPayload(
+		url,
+		finalUrl,
+		mime,
+		extHint,
+		rawContent,
+		timeout,
+		signal,
+		fetchedAt,
+		notes,
+	);
+	if (binaryPayloadResult) return binaryPayloadResult;
+
 	// Step 4: Handle non-HTML text content
 	const isHtml = mime.includes("html") || mime.includes("xhtml");
 	const isJson = mime.includes("json");
@@ -999,7 +1282,7 @@ async function renderUrl(
 	const isFeed = mime.includes("rss") || mime.includes("atom") || mime.includes("feed");
 
 	// Raw mode skips every text-shaping branch below (JSON pretty-print, feed-to-markdown,
-	// HTML extraction) and returns the response body verbatim. The image/markit branches
+	// HTML extraction) and returns the response body verbatim. Binary-oriented branches
 	// above already ran because raw isn't useful for binary payloads.
 	if (raw) {
 		const output = finalizeOutput(rawContent);
@@ -1141,10 +1424,27 @@ async function renderUrl(
 			throw new ToolAbortError();
 		}
 
-		// 5E: Render HTML with lynx or html2text
+		// 5E: Render HTML via the reader-backend chain (native/trafilatura/lynx/parallel/jina)
 		const htmlResult = await renderHtmlToText(finalUrl, rawContent, timeout, settings, signal, storage);
 		if (!htmlResult.ok) {
-			notes.push("html rendering failed (lynx/html2text unavailable)");
+			notes.push("html rendering failed (no reader backend produced usable output)");
+
+			const llmResult = await tryLlmEndpoints(finalUrl, timeout, signal);
+			if (llmResult) {
+				notes.push(`Used llms.txt fallback: ${llmResult.endpoint}`);
+				const output = finalizeOutput(llmResult.content);
+				return {
+					url,
+					finalUrl,
+					contentType: "text/plain",
+					method: "llms.txt",
+					content: output.content,
+					fetchedAt,
+					truncated: output.truncated,
+					notes,
+				};
+			}
+
 			const output = finalizeOutput(rawContent);
 			return {
 				url,
@@ -1488,11 +1788,11 @@ export function renderReadUrlResult(
 		const header = renderStatusLine({ icon: "error", title: "Read", description }, uiTheme);
 		const errorLines = errorText.split("\n").map(line => uiTheme.fg("error", replaceTabs(line)));
 		const outputBlock = new CachedOutputBlock();
-		return {
+		return markFramedBlockComponent({
 			render: (width: number) =>
 				outputBlock.render({ header, state: "error", sections: [{ lines: errorLines }], width }, uiTheme),
 			invalidate: () => outputBlock.invalidate(),
-		};
+		});
 	}
 
 	const description = formatReadUrlDescription(details.finalUrl);
@@ -1542,7 +1842,7 @@ export function renderReadUrlResult(
 	let lastExpanded: boolean | undefined;
 	let contentPreviewLines: string[] | undefined;
 
-	return {
+	return markFramedBlockComponent({
 		render: (width: number) => {
 			const { expanded } = options;
 
@@ -1582,5 +1882,5 @@ export function renderReadUrlResult(
 			contentPreviewLines = undefined;
 			lastExpanded = undefined;
 		},
-	};
+	});
 }
diff --git a/packages/coding-agent/src/tools/find.ts b/packages/coding-agent/src/tools/find.ts
index 85b7bdaf0..8854a297a 100644
--- a/packages/coding-agent/src/tools/find.ts
+++ b/packages/coding-agent/src/tools/find.ts
@@ -13,6 +13,7 @@ import findDescription from "../prompts/tools/find.md" with { type: "text" };
 import { type TruncationResult, truncateHead } from "../session/streaming-output";
 import { Ellipsis, fileHyperlink, renderFileList, renderStatusLine, renderTreeList, truncateToWidth } from "../tui";
 import type { ToolSession } from ".";
+import { buildPathTree, walkPathTree } from "./grouped-file-output";
 import { applyListLimit } from "./list-limit";
 import { formatFullOutputReference, type OutputMeta } from "./output-meta";
 import {
@@ -54,34 +55,27 @@ const MIN_GLOB_TIMEOUT_MS = 500;
 const MAX_GLOB_TIMEOUT_MS = 60_000;
 
 /**
- * Group find matches by their directory so the model doesn't pay repeated
- * tokens for shared path prefixes. Preserves the input order: groups appear in
- * the order their first member was emitted (mtime-desc for native glob), and
- * within a group entries keep their relative order.
+ * Group find matches into a multi-level directory tree so the model doesn't pay
+ * repeated tokens for shared path prefixes. Single-child directory chains fold
+ * into one header (`# a/b/c/`), so a common prefix — including an absolute root
+ * for out-of-cwd results — collapses to a single line. Each level adds one `#`;
+ * files are listed bare under the deepest directory header that owns them.
+ *
+ * Order follows the input (mtime-desc for native glob): a directory appears when
+ * its first member is emitted, and a node's own files precede its subdirectories.
  */
 export function formatFindGroupedOutput(paths: readonly string[]): string {
 	if (paths.length === 0) return "";
-	const groups = new Map<string, string[]>();
-	for (const entry of paths) {
-		const hasTrailingSlash = entry.endsWith("/");
-		const trimmed = hasTrailingSlash ? entry.slice(0, -1) : entry;
-		const slash = trimmed.lastIndexOf("/");
-		const dir = slash === -1 ? "" : trimmed.slice(0, slash);
-		const base = slash === -1 ? trimmed : trimmed.slice(slash + 1);
-		const label = hasTrailingSlash ? `${base}/` : base;
-		const list = groups.get(dir);
-		if (list) list.push(label);
-		else groups.set(dir, [label]);
-	}
-	const sections: string[] = [];
-	for (const [dir, entries] of groups) {
-		if (dir === "") {
-			sections.push(entries.join("\n"));
+	const tree = buildPathTree(paths.map(entry => ({ path: entry, isDir: entry.endsWith("/") })));
+	const lines: string[] = [];
+	for (const event of walkPathTree(tree)) {
+		if (event.kind === "dir") {
+			lines.push(`${"#".repeat(event.depth + 1)} ${event.name}/`);
 		} else {
-			sections.push(`# ${dir}/\n${entries.join("\n")}`);
+			lines.push(event.name);
 		}
 	}
-	return sections.join("\n\n");
+	return lines.join("\n");
 }
 
 export interface FindToolDetails {
@@ -434,6 +428,10 @@ function formatFindRenderPaths(paths: FindRenderArgs["paths"]): string | undefin
 
 const COLLAPSED_LIST_LIMIT = PREVIEW_LIMITS.COLLAPSED_ITEMS;
 
+function findStatusIcon(uiTheme: Theme): string {
+	return uiTheme.fg("toolTitle", uiTheme.symbol("icon.search"));
+}
+
 export const findToolRenderer = {
 	inline: true,
 	renderCall(args: FindRenderArgs, _options: RenderResultOptions, uiTheme: Theme): Component {
@@ -441,10 +439,16 @@ export const findToolRenderer = {
 		if (args.limit !== undefined) meta.push(`limit:${args.limit}`);
 
 		const text = renderStatusLine(
-			{ icon: "pending", title: "Find", description: formatFindRenderPaths(args.paths) || "*", meta },
+			{
+				icon: "pending",
+				title: "Find",
+				titleColor: "toolTitle",
+				description: formatFindRenderPaths(args.paths) || "*",
+				meta,
+			},
 			uiTheme,
 		);
-		return new Text(text, 0, 0);
+		return new Text(text, 1, 0);
 	},
 
 	renderResult(
@@ -457,7 +461,7 @@ export const findToolRenderer = {
 
 		if (result.isError || details?.error) {
 			const errorText = details?.error || result.content?.find(c => c.type === "text")?.text || "Unknown error";
-			return new Text(formatErrorMessage(errorText, uiTheme), 0, 0);
+			return new Text(formatErrorMessage(errorText, uiTheme), 1, 0);
 		}
 
 		const hasDetailedData = details?.fileCount !== undefined;
@@ -470,14 +474,15 @@ export const findToolRenderer = {
 				textContent.includes("No files found") ||
 				textContent.trim() === ""
 			) {
-				return new Text(formatEmptyMessage("No files found", uiTheme), 0, 0);
+				return new Text(formatEmptyMessage("No files found", uiTheme), 1, 0);
 			}
 
 			const lines = textContent.split("\n").filter(l => l.trim());
 			const header = renderStatusLine(
 				{
-					icon: "success",
+					iconOverride: findStatusIcon(uiTheme),
 					title: "Find",
+					titleColor: "toolTitle",
 					description: formatFindRenderPaths(args?.paths),
 					meta: [formatCount("file", lines.length)],
 				},
@@ -498,6 +503,7 @@ export const findToolRenderer = {
 					);
 					return [header, ...listLines].map(l => truncateToWidth(l, width, Ellipsis.Omit));
 				},
+				{ paddingX: 1 },
 			);
 		}
 
@@ -513,20 +519,27 @@ export const findToolRenderer = {
 
 		if (fileCount === 0) {
 			const header = renderStatusLine(
-				{ icon: "warning", title: "Find", description: formatFindRenderPaths(args?.paths), meta: ["0 files"] },
+				{
+					icon: "warning",
+					title: "Find",
+					titleColor: "toolTitle",
+					description: formatFindRenderPaths(args?.paths),
+					meta: ["0 files"],
+				},
 				uiTheme,
 			);
 			const lines = [header, formatEmptyMessage("No files found", uiTheme)];
 			if (missingNote) lines.push(missingNote);
-			return new Text(lines.join("\n"), 0, 0);
+			return new Text(lines.join("\n"), 1, 0);
 		}
 		const meta: string[] = [formatCount("file", fileCount)];
 		if (details?.scopePath) meta.push(`in ${details.scopePath}`);
 		if (truncated) meta.push(uiTheme.fg("warning", "truncated"));
 		const header = renderStatusLine(
 			{
-				icon: truncated ? "warning" : "success",
+				...(truncated ? { icon: "warning" as const } : { iconOverride: findStatusIcon(uiTheme) }),
 				title: "Find",
+				titleColor: "toolTitle",
 				description: formatFindRenderPaths(args?.paths),
 				meta,
 			},
@@ -565,6 +578,7 @@ export const findToolRenderer = {
 				);
 				return [header, ...fileLines, ...extraLines].map(l => truncateToWidth(l, width, Ellipsis.Omit));
 			},
+			{ paddingX: 1 },
 		);
 	},
 	mergeCallAndResult: true,
diff --git a/packages/coding-agent/src/tools/gh-renderer.ts b/packages/coding-agent/src/tools/gh-renderer.ts
index cb01261ea..d34d8d9f0 100644
--- a/packages/coding-agent/src/tools/gh-renderer.ts
+++ b/packages/coding-agent/src/tools/gh-renderer.ts
@@ -1,7 +1,7 @@
-import { type Component, padding, Text, truncateToWidth, visibleWidth } from "@oh-my-pi/pi-tui";
+import { type Component, padding, Text, visibleWidth } from "@oh-my-pi/pi-tui";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import type { Theme, ThemeColor } from "../modes/theme/theme";
-import { renderStatusLine } from "../tui";
+import { framedBlock, renderStatusLine } from "../tui";
 import type {
 	GhRunWatchFailedLogDetails,
 	GhRunWatchJobDetails,
@@ -239,7 +239,7 @@ function renderFailedLogs(
 		return [];
 	}
 
-	const lines = ["", theme.fg("error", "failed logs")];
+	const lines: string[] = [];
 	for (const entry of failedLogs) {
 		const context = entry.workflowName ? `${entry.workflowName}  #${entry.runId}` : `run #${entry.runId}`;
 		lines.push(
@@ -268,36 +268,45 @@ function renderFailedLogs(
 	return lines;
 }
 
-function buildWatchLines(
+function buildWatchSections(
 	watch: GhRunWatchViewDetails,
 	theme: Theme,
 	options: RenderResultOptions,
 	width: number,
-): string[] {
-	const lines = [theme.fg("muted", getWatchHeader(watch))];
+): Array<{ label?: string; lines: string[] }> {
+	const main: string[] = [];
 
 	if (watch.note) {
-		lines.push(theme.fg("dim", replaceTabs(watch.note)));
+		main.push(theme.fg("dim", replaceTabs(watch.note)));
 	}
 
 	if (watch.mode === "run" && watch.run) {
-		lines.push(...renderRunBlock(watch.run, width, theme));
+		main.push(...renderRunBlock(watch.run, width, theme));
 	} else if (watch.mode === "commit") {
 		const runs = watch.runs ?? [];
 		if (runs.length === 0) {
-			lines.push(theme.fg("dim", "waiting for workflow runs..."));
+			main.push(theme.fg("dim", "waiting for workflow runs..."));
 		} else {
 			runs.forEach((run, index) => {
 				if (index > 0) {
-					lines.push("");
+					main.push("");
 				}
-				lines.push(...renderRunBlock(run, width, theme));
+				main.push(...renderRunBlock(run, width, theme));
 			});
 		}
 	}
 
-	lines.push(...renderFailedLogs(watch.failedLogs ?? [], width, theme, options.expanded));
-	return lines;
+	const sections: Array<{ label?: string; lines: string[] }> = [];
+	if (main.length > 0) {
+		sections.push({ lines: main });
+	}
+
+	const failed = renderFailedLogs(watch.failedLogs ?? [], width, theme, options.expanded);
+	if (failed.length > 0) {
+		sections.push({ label: "failed logs", lines: failed });
+	}
+
+	return sections;
 }
 
 function extractText(content: Array<{ type: string; text?: string }>): string {
@@ -335,29 +344,44 @@ function renderFallbackComponent(
 	}
 
 	const allLines = replaceTabs(text).split("\n");
+	while (allLines.length > 0 && allLines[0].trim() === "") allLines.shift();
+	while (allLines.length > 0 && allLines[allLines.length - 1].trim() === "") allLines.pop();
 
-	return {
-		render(width: number): string[] {
-			const lineWidth = Math.max(24, width || FALLBACK_WIDTH);
-			const expanded = options.expanded;
-			const limit = expanded ? allLines.length : Math.min(allLines.length, PREVIEW_LIMITS.OUTPUT_EXPANDED);
-			const visible = allLines.slice(0, limit);
-			const remaining = allLines.length - visible.length;
+	// Trivial one-line *success* result: a clean status line beats an almost-empty box.
+	// Errors always frame so the message reads as a structured block, never a raw red wrap.
+	if (allLines.length <= 1 && !isError) {
+		const body = allLines[0];
+		if (!body) return new Text(header, 0, 0);
+		const colored = isError ? theme.fg("error", body) : theme.fg("toolOutput", body);
+		return new Text(`${header}\n${colored}`, 0, 0);
+	}
 
-			const out: string[] = [header];
-			for (const line of visible) {
-				const colored = isError ? theme.fg("error", line) : theme.fg("toolOutput", line);
-				out.push(truncateVisualWidth(colored, lineWidth));
-			}
-			if (!expanded && remaining > 0) {
-				const hint = formatExpandHint(theme, expanded, true);
-				const more = `${formatMoreItems(remaining, "line")}${hint ? ` ${hint}` : ""}`;
-				out.push(theme.fg("dim", more));
-			}
-			return out.map(line => truncateToWidth(line, lineWidth));
-		},
-		invalidate() {},
-	};
+	return framedBlock(theme, width => {
+		const lineWidth = Math.max(1, (width || FALLBACK_WIDTH) - 3);
+		const expanded = options.expanded;
+		const limit = expanded ? allLines.length : Math.min(allLines.length, PREVIEW_LIMITS.OUTPUT_EXPANDED);
+		const visible = allLines.slice(0, limit);
+		const remaining = allLines.length - visible.length;
+
+		const out: string[] = [];
+		for (const line of visible) {
+			const colored = isError ? theme.fg("error", line) : theme.fg("toolOutput", line);
+			out.push(truncateVisualWidth(colored, lineWidth));
+		}
+		if (!expanded && remaining > 0) {
+			const hint = formatExpandHint(theme, expanded, true);
+			const more = `${formatMoreItems(remaining, "line")}${hint ? ` ${hint}` : ""}`;
+			out.push(theme.fg("dim", more));
+		}
+		return {
+			header,
+			sections: out.length > 0 ? [{ lines: out }] : [],
+			state: isError ? "error" : "success",
+			borderColor: isError ? "error" : "borderMuted",
+			applyBg: false,
+			width,
+		};
+	});
 }
 
 function renderWatchCall(args: GithubToolRenderArgs, options: RenderResultOptions, theme: Theme): Component {
@@ -380,7 +404,7 @@ function renderWatchCall(args: GithubToolRenderArgs, options: RenderResultOption
 	}
 
 	const header = `${icon} ${titleText}  ${metaText}`;
-	const wait = theme.fg("dim", "  waiting for workflow data...");
+	const wait = theme.fg("dim", "waiting for workflow data...");
 	return new Text(`${header}\n${wait}`, 0, 0);
 }
 
@@ -412,13 +436,28 @@ export const githubToolRenderer = {
 	): Component {
 		const watch = result.details?.watch;
 		if (watch) {
-			return {
-				render(width: number): string[] {
-					const lineWidth = Math.max(24, width || FALLBACK_WIDTH);
-					return buildWatchLines(watch, uiTheme, options, lineWidth).map(line => truncateToWidth(line, lineWidth));
+			const isError = result.isError === true;
+			const header = renderStatusLine(
+				{
+					icon: isError ? "error" : "success",
+					title: "GitHub Run Watch",
+					titleColor: isError ? "error" : "accent",
+					meta: [getWatchHeader(watch)],
 				},
-				invalidate() {},
-			};
+				uiTheme,
+			);
+			return framedBlock(uiTheme, width => {
+				const innerWidth = Math.max(1, (width || FALLBACK_WIDTH) - 3);
+				const sections = buildWatchSections(watch, uiTheme, options, innerWidth);
+				return {
+					header,
+					sections,
+					state: isError ? "error" : "success",
+					borderColor: isError ? "error" : "borderMuted",
+					applyBg: false,
+					width,
+				};
+			});
 		}
 
 		return renderFallbackComponent(result, options, uiTheme, args ?? {});
diff --git a/packages/coding-agent/src/tools/grouped-file-output.ts b/packages/coding-agent/src/tools/grouped-file-output.ts
index 6f27fb9dc..3027693a1 100644
--- a/packages/coding-agent/src/tools/grouped-file-output.ts
+++ b/packages/coding-agent/src/tools/grouped-file-output.ts
@@ -6,10 +6,127 @@ function isUrlLikePath(filePath: string): boolean {
 	return URL_LIKE_PATH_RE.test(filePath);
 }
 
+// =============================================================================
+// Multi-level path tree
+// =============================================================================
+//
+// File listings (grep / ast-grep / ast-edit / lsp diagnostics / find) used to
+// group by the *immediate* parent directory and print the full directory path in
+// every header. For results spread across a deep tree — or rooted outside cwd,
+// where paths stay absolute — that repeated the shared prefix on every line. The
+// tree below folds single-child directory chains (so the common prefix collapses
+// into one header) and nests the rest, charging the model one token per path
+// segment instead of one per file.
+
+interface PathTreeNode {
+	/** Direct file leaves, in first-seen order. */
+	files: Array<{ name: string; key: string }>;
+	/** Dedup set for `files` (a glob can surface the same path twice on retry). */
+	fileNames: Set<string>;
+	/** Child directories, in first-seen order. */
+	subdirs: Array<{ name: string; node: PathTreeNode }>;
+	/** Dedup index for `subdirs`. */
+	dirIndex: Map<string, PathTreeNode>;
+}
+
+export interface PathTreeInput {
+	/** Path string; absolute, cwd-relative, or url-like. Backslashes are normalized. */
+	path: string;
+	/** Whether the leaf itself is a directory (trailing-slash match from find). */
+	isDir: boolean;
+	/** Opaque key carried onto file events for section lookup. Defaults to `path`. */
+	key?: string;
+}
+
+/** One node emitted while walking the tree: a folded directory or a file leaf. */
+export interface GroupedTreeEvent {
+	kind: "dir" | "file";
+	/** 0-based nesting depth (root children are depth 0). */
+	depth: number;
+	/** Folded chain for dirs (e.g. `a/b/c`, no trailing slash); basename for files. */
+	name: string;
+	/** File key for `kind === "file"`; empty string for directories. */
+	key: string;
+}
+
+function createNode(): PathTreeNode {
+	return { files: [], fileNames: new Set(), subdirs: [], dirIndex: new Map() };
+}
+
+function addFile(node: PathTreeNode, name: string, key: string): void {
+	if (node.fileNames.has(name)) return;
+	node.fileNames.add(name);
+	node.files.push({ name, key });
+}
+
+/**
+ * Build a directory tree from a flat list of paths. URL-like entries are kept
+ * whole as root-level file leaves (they have no meaningful directory structure).
+ * Absolute paths carry a leading empty segment so they share a common `/` root
+ * and fold like any other prefix.
+ */
+export function buildPathTree(entries: Iterable<PathTreeInput>): PathTreeNode {
+	const root = createNode();
+	for (const { path: rawPath, isDir, key } of entries) {
+		const normalized = rawPath.replace(/\\/g, "/");
+		const fileKey = key ?? rawPath;
+		if (isUrlLikePath(normalized)) {
+			addFile(root, normalized, fileKey);
+			continue;
+		}
+		const trimmed = normalized.endsWith("/") ? normalized.slice(0, -1) : normalized;
+		if (trimmed.length === 0) continue;
+		const segments = trimmed.split("/");
+		const dirCount = isDir ? segments.length : segments.length - 1;
+		let node = root;
+		for (let i = 0; i < dirCount; i++) {
+			const segment = segments[i]!;
+			let child = node.dirIndex.get(segment);
+			if (!child) {
+				child = createNode();
+				node.dirIndex.set(segment, child);
+				node.subdirs.push({ name: segment, node: child });
+			}
+			node = child;
+		}
+		if (!isDir) {
+			addFile(node, segments[segments.length - 1]!, fileKey);
+		}
+	}
+	return root;
+}
+
+/**
+ * Depth-first walk yielding directory and file events. Directories collapse their
+ * single-child chains (`a` → `a/b` → `a/b/c`) so a shared prefix becomes one
+ * header. Each node's direct files are emitted before its subdirectories, keeping
+ * a file unambiguously attached to the header above it.
+ */
+export function* walkPathTree(node: PathTreeNode, depth = 0): Generator<GroupedTreeEvent> {
+	for (const file of node.files) {
+		yield { kind: "file", depth, name: file.name, key: file.key };
+	}
+	for (const subdir of node.subdirs) {
+		let dirNode = subdir.node;
+		const parts = [subdir.name];
+		while (dirNode.files.length === 0 && dirNode.subdirs.length === 1) {
+			const only = dirNode.subdirs[0]!;
+			parts.push(only.name);
+			dirNode = only.node;
+		}
+		yield { kind: "dir", depth, name: parts.join("/"), key: "" };
+		yield* walkPathTree(dirNode, depth + 1);
+	}
+}
+
+// =============================================================================
+// Grouped file output (grep / ast-grep / ast-edit / lsp diagnostics)
+// =============================================================================
+
 /**
  * One file's contribution to a grouped file output. The header itself is generated
- * by `formatGroupedFiles` (single `#` for root files, `##` for files inside a dir);
- * use `headerSuffix` to tack on extras like ` (1 replacement)`.
+ * by `formatGroupedFiles` (one `#` per nesting level); use `headerSuffix` to tack
+ * on extras like ` (1 replacement)`.
  */
 export interface GroupedFileSection {
 	/** Optional suffix appended to the file header. */
@@ -28,76 +145,183 @@ export interface GroupedFilesOutput {
 }
 
 /**
- * Render a list of files as directory-grouped sections shared by grep, ast-grep,
- * ast-edit, and the LSP diagnostic formatter.
+ * Render a list of files as a multi-level, prefix-folded directory tree shared by
+ * grep, ast-grep, ast-edit, and the LSP diagnostic formatter.
  *
- * Layout:
- *   # dir/
- *   ## file.ts
+ * Layout (one `#` per level; the shared prefix folds into the top header):
+ *   # packages/pkg/src/
+ *   ## root.ts
+ *   …body…
+ *   ## nested/
+ *   ### child.ts
  *   …body…
  *
- *   # otherdir/
- *   ## other.ts
- *   …body…
- *
- * Files in the project root (directory `.`) become single-`#` headers without a
- * `## file` line, matching the existing convention.
+ * Files in the (folded) project root become single-`#` headers with no parent
+ * directory line. A blank line precedes every directory header and every
+ * root-level file so the renderers can split the output into collapsible groups.
  */
 export function formatGroupedFiles(
 	files: string[],
 	renderFile: (filePath: string) => GroupedFileSection,
 ): GroupedFilesOutput {
-	const filesByDirectory = new Map<string, string[]>();
+	const sections = new Map<string, GroupedFileSection>();
+	const inputs: PathTreeInput[] = [];
 	for (const filePath of files) {
-		const directory = isUrlLikePath(filePath) ? "." : path.dirname(filePath).replace(/\\/g, "/");
-		if (!filesByDirectory.has(directory)) {
-			filesByDirectory.set(directory, []);
-		}
-		filesByDirectory.get(directory)!.push(filePath);
+		if (sections.has(filePath)) continue;
+		const section = renderFile(filePath);
+		if (section.skip) continue;
+		sections.set(filePath, section);
+		inputs.push({ path: filePath, isDir: false, key: filePath });
 	}
 
+	const tree = buildPathTree(inputs);
 	const model: string[] = [];
 	const display: string[] = [];
+	let emitted = false;
 
-	const pushSeparatorIfNeeded = () => {
-		if (model.length > 0) {
+	for (const event of walkPathTree(tree)) {
+		const hashes = "#".repeat(event.depth + 1);
+		const needsSeparator = emitted && (event.depth === 0 || event.kind === "dir");
+		if (needsSeparator) {
 			model.push("");
 			display.push("");
 		}
-	};
-
-	for (const [directory, dirFiles] of filesByDirectory) {
-		if (directory === ".") {
-			for (const filePath of dirFiles) {
-				const section = renderFile(filePath);
-				if (section.skip) continue;
-				pushSeparatorIfNeeded();
-				const headerName = isUrlLikePath(filePath) ? filePath : path.basename(filePath);
-				const header = `# ${headerName}${section.headerSuffix ?? ""}`;
-				model.push(header, ...section.modelLines);
-				display.push(header, ...(section.displayLines ?? section.modelLines));
-			}
+		emitted = true;
+		if (event.kind === "dir") {
+			const header = `${hashes} ${event.name}/`;
+			model.push(header);
+			display.push(header);
 			continue;
 		}
-
-		const sections: Array<{ filePath: string; section: GroupedFileSection }> = [];
-		for (const filePath of dirFiles) {
-			const section = renderFile(filePath);
-			if (section.skip) continue;
-			sections.push({ filePath, section });
-		}
-		if (sections.length === 0) continue;
-
-		pushSeparatorIfNeeded();
-		const dirHeader = `# ${directory}/`;
-		model.push(dirHeader);
-		display.push(dirHeader);
-		for (const { filePath, section } of sections) {
-			const fileHeader = `## ${path.basename(filePath)}${section.headerSuffix ?? ""}`;
-			model.push(fileHeader, ...section.modelLines);
-			display.push(fileHeader, ...(section.displayLines ?? section.modelLines));
-		}
+		const section = sections.get(event.key)!;
+		const header = `${hashes} ${event.name}${section.headerSuffix ?? ""}`;
+		model.push(header, ...section.modelLines);
+		display.push(header, ...(section.displayLines ?? section.modelLines));
 	}
 
 	return { model, display };
 }
+
+// =============================================================================
+// Parsing grouped output back into per-line context (TUI renderers)
+// =============================================================================
+
+const GROUPED_HEADER_RE = /^(#+)\s+(.*)$/;
+const HEADER_SUFFIX_RE = /\s+\([^)]*\)\s*$/;
+const HEADER_HASH_TAG_RE = /#[0-9a-f]+$/i;
+
+/** Per-line classification of grouped output, used by renderers for hyperlinks. */
+export interface GroupedLineContext {
+	/** Directory header, file header, or any non-header body/content line. */
+	kind: "dir" | "file" | "content";
+	/** Number of leading `#` for headers; 0 for content lines. */
+	depth: number;
+	/** Resolved absolute path of the dir/file a header points at (when resolvable). */
+	headerPath?: string;
+	/** For content lines, the absolute path of the owning file (line hyperlinks). */
+	filePath?: string;
+	/** Header is an internal/url-like target the caller resolves itself. */
+	isUrl?: boolean;
+}
+
+function resolveGroupedPath(parent: string | undefined, name: string): string | undefined {
+	if (parent === undefined) return undefined;
+	if (name === "" || name === ".") return parent;
+	// `path.resolve` keeps an absolute `name` intact (out-of-cwd results) while
+	// joining a relative folded chain (`packages/pkg/src`) onto the parent.
+	return path.resolve(parent, name);
+}
+
+/**
+ * Walk grouped output lines, tracking a directory stack keyed by header depth, so
+ * each header and body line can be linked back to its absolute filesystem path.
+ * Reconstruction is stack-based (not per-blank-group) so nested directory headers
+ * resolve correctly across the whole output.
+ *
+ * `headerBase` is the directory the displayed (folded) header paths are relative
+ * to — for grep/ast tools that is the session cwd, since display paths are
+ * formatted relative to cwd regardless of the (sub)directory the search was
+ * scoped to. `fileScope` is the initial owning file for body lines that appear
+ * before any header (single-file scopes have no `#` headers); it defaults to
+ * `headerBase` and should be passed the scoped file's absolute path.
+ */
+export function classifyGroupedLines(
+	lines: readonly string[],
+	headerBase: string | undefined,
+	fileScope: string | undefined = headerBase,
+): GroupedLineContext[] {
+	const result: GroupedLineContext[] = [];
+	const dirAtDepth = new Map<number, string>();
+	// Body lines before any header (single-file scopes) link to the scoped file.
+	let currentFile = fileScope;
+
+	const clearDeeper = (depth: number) => {
+		for (const key of dirAtDepth.keys()) {
+			if (key >= depth) dirAtDepth.delete(key);
+		}
+	};
+
+	for (const line of lines) {
+		const match = GROUPED_HEADER_RE.exec(line);
+		if (!match) {
+			result.push({ kind: "content", depth: 0, filePath: currentFile });
+			continue;
+		}
+		const depth = match[1]!.length;
+		const rest = match[2]!.trimEnd();
+		if (isUrlLikePath(rest)) {
+			clearDeeper(depth);
+			currentFile = undefined;
+			result.push({ kind: "file", depth, isUrl: true });
+			continue;
+		}
+		const parent = depth > 1 ? dirAtDepth.get(depth - 1) : headerBase;
+		if (rest.endsWith("/")) {
+			const name = rest.slice(0, -1).replace(HEADER_SUFFIX_RE, "");
+			const abs = resolveGroupedPath(parent, name);
+			clearDeeper(depth);
+			if (abs !== undefined) dirAtDepth.set(depth, abs);
+			currentFile = undefined;
+			result.push({ kind: "dir", depth, headerPath: abs });
+			continue;
+		}
+		const name = rest.replace(HEADER_SUFFIX_RE, "").replace(HEADER_HASH_TAG_RE, "");
+		const abs = name ? resolveGroupedPath(parent, name) : undefined;
+		currentFile = abs;
+		result.push({ kind: "file", depth, headerPath: abs });
+	}
+
+	return result;
+}
+
+/**
+ * Split line indices into blank-line-separated groups, mirroring
+ * `splitGroupsByBlankLine`: when any blank line is present, break on runs of
+ * blanks; otherwise return a single group of the non-empty lines. Returning
+ * indices lets callers slice parallel arrays (raw lines, styled lines, contexts).
+ */
+export function groupLineIndicesByBlank(rawLines: readonly string[]): number[][] {
+	const hasSeparators = rawLines.some(line => line.trim().length === 0);
+	const groups: number[][] = [];
+	if (hasSeparators) {
+		let current: number[] = [];
+		for (let i = 0; i < rawLines.length; i++) {
+			if (rawLines[i]!.trim().length === 0) {
+				if (current.length > 0) {
+					groups.push(current);
+					current = [];
+				}
+				continue;
+			}
+			current.push(i);
+		}
+		if (current.length > 0) groups.push(current);
+	} else {
+		const current: number[] = [];
+		for (let i = 0; i < rawLines.length; i++) {
+			if (rawLines[i]!.trim().length > 0) current.push(i);
+		}
+		if (current.length > 0) groups.push(current);
+	}
+	return groups;
+}
diff --git a/packages/coding-agent/src/tools/image-gen.ts b/packages/coding-agent/src/tools/image-gen.ts
index 521fb94e6..ee90c5ca0 100644
--- a/packages/coding-agent/src/tools/image-gen.ts
+++ b/packages/coding-agent/src/tools/image-gen.ts
@@ -1,6 +1,6 @@
 import * as os from "node:os";
 import * as path from "node:path";
-import { getAntigravityUserAgent, getEnvApiKey, type Model } from "@oh-my-pi/pi-ai";
+import { type ApiKey, getAntigravityUserAgent, getEnvApiKey, type Model, withAuth } from "@oh-my-pi/pi-ai";
 import {
 	CODEX_BASE_URL,
 	getCodexAccountId,
@@ -20,6 +20,7 @@ import {
 } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
 import packageJson from "../../package.json" with { type: "json" };
+
 import { isAuthenticated, type ModelRegistry } from "../config/model-registry";
 import type { CustomTool } from "../extensibility/custom-tools/types";
 import { ohMyPiXAIUserAgent, resolveXAIHttpCredentials } from "../lib/xai-http";
@@ -864,7 +865,10 @@ async function generateOpenAIHostedImage(
 
 	if (!response.ok) {
 		const errorText = await response.text();
-		throw new Error(`OpenAI image request failed (${response.status}): ${getOpenAIResponseErrorMessage(errorText)}`);
+		throw Object.assign(
+			new Error(`OpenAI image request failed (${response.status}): ${getOpenAIResponseErrorMessage(errorText)}`),
+			{ status: response.status },
+		);
 	}
 
 	const contentType = response.headers.get("content-type") ?? "";
@@ -1037,13 +1041,16 @@ export const imageGenTool: CustomTool<typeof imageGenSchema, ImageGenToolDetails
 					throw new Error("Missing active GPT model for OpenAI image generation");
 				}
 
-				const parsed = await generateOpenAIHostedImage(
-					apiKey.apiKey,
-					apiKey.model,
-					params,
-					resolvedImages,
-					requestSignal,
+				const hostedModel = apiKey.model;
+				const hostedKey: ApiKey = ctx.modelRegistry.resolver(hostedModel.provider, {
 					sessionId,
+					baseUrl: hostedModel.baseUrl,
+				});
+
+				const parsed = await withAuth(
+					hostedKey,
+					key => generateOpenAIHostedImage(key, hostedModel, params, resolvedImages, requestSignal, sessionId),
+					{ signal: requestSignal },
 				);
 
 				if (parsed.images.length === 0) {
@@ -1088,38 +1095,57 @@ export const imageGenTool: CustomTool<typeof imageGenSchema, ImageGenToolDetails
 				}
 
 				const prompt = assemblePrompt(params);
-				const requestBody = buildAntigravityRequest(
-					prompt,
-					model,
-					apiKey.projectId,
-					params.aspect_ratio,
-					params.image_size,
-					resolvedImages,
-				);
-
-				const response = await fetch(`${ANTIGRAVITY_ENDPOINT}/v1internal:streamGenerateContent?alt=sse`, {
-					method: "POST",
-					headers: {
-						Authorization: `Bearer ${apiKey.apiKey}`,
-						"Content-Type": "application/json",
-						Accept: "text/event-stream",
-						"User-Agent": getAntigravityUserAgent(),
-					},
-					body: JSON.stringify(requestBody),
-					signal: requestSignal,
+				const antigravityKey: ApiKey = ctx.modelRegistry.resolver("google-antigravity", {
+					sessionId,
 				});
 
-				if (!response.ok) {
-					const errorText = await response.text();
-					let message = errorText;
-					try {
-						const parsed = JSON.parse(errorText) as { error?: { message?: string } };
-						message = parsed.error?.message ?? message;
-					} catch {
-						// Keep raw text.
-					}
-					throw new Error(`Antigravity image request failed (${response.status}): ${message}`);
-				}
+				const response = await withAuth(
+					antigravityKey,
+					async key => {
+						// On a retry the resolver yields the raw stored credential JSON
+						// ({ token, projectId }); the initial seed is the already-parsed
+						// access token. Tolerate both, falling back to the seed projectId.
+						const rotated = parseAntigravityCredentials(key);
+						const bearer = rotated?.accessToken ?? key;
+						const projectId = rotated?.projectId ?? apiKey.projectId!;
+						const requestBody = buildAntigravityRequest(
+							prompt,
+							model,
+							projectId,
+							params.aspect_ratio,
+							params.image_size,
+							resolvedImages,
+						);
+
+						const resp = await fetch(`${ANTIGRAVITY_ENDPOINT}/v1internal:streamGenerateContent?alt=sse`, {
+							method: "POST",
+							headers: {
+								Authorization: `Bearer ${bearer}`,
+								"Content-Type": "application/json",
+								Accept: "text/event-stream",
+								"User-Agent": getAntigravityUserAgent(),
+							},
+							body: JSON.stringify(requestBody),
+							signal: requestSignal,
+						});
+
+						if (!resp.ok) {
+							const errorText = await resp.text();
+							let message = errorText;
+							try {
+								const parsedErr = JSON.parse(errorText) as { error?: { message?: string } };
+								message = parsedErr.error?.message ?? message;
+							} catch {
+								// Keep raw text.
+							}
+							throw Object.assign(new Error(`Antigravity image request failed (${resp.status}): ${message}`), {
+								status: resp.status,
+							});
+						}
+						return resp;
+					},
+					{ signal: requestSignal },
+				);
 
 				const parsed = await parseAntigravitySseForImage(response, requestSignal);
 				const responseText = parsed.text.length > 0 ? parsed.text.join(" ") : undefined;
@@ -1191,28 +1217,41 @@ export const imageGenTool: CustomTool<typeof imageGenSchema, ImageGenToolDetails
 					: xaiBaseBody;
 				const xaiEndpoint = isEdit ? "/images/edits" : "/images/generations";
 
-				const xaiResponse = await fetch(`${xaiCreds.baseURL}${xaiEndpoint}`, {
-					method: "POST",
-					headers: {
-						Authorization: `Bearer ${xaiCreds.apiKey}`,
-						"Content-Type": "application/json",
-						"User-Agent": ohMyPiXAIUserAgent(),
-					},
-					body: JSON.stringify(xaiBody),
-					signal: requestSignal,
+				const xaiKey: ApiKey = ctx.modelRegistry.resolver(xaiCreds.provider, {
+					sessionId,
+					baseUrl: xaiCreds.baseURL,
 				});
 
-				const xaiRawText = await xaiResponse.text();
-				if (!xaiResponse.ok) {
-					let message = xaiRawText;
-					try {
-						const parsedErr = JSON.parse(xaiRawText) as { error?: { message?: string } };
-						message = parsedErr.error?.message ?? message;
-					} catch {
-						// Keep raw text.
-					}
-					throw new Error(`xAI image request failed (${xaiResponse.status}): ${message}`);
-				}
+				const xaiRawText = await withAuth(
+					xaiKey,
+					async key => {
+						const resp = await fetch(`${xaiCreds.baseURL}${xaiEndpoint}`, {
+							method: "POST",
+							headers: {
+								Authorization: `Bearer ${key}`,
+								"Content-Type": "application/json",
+								"User-Agent": ohMyPiXAIUserAgent(),
+							},
+							body: JSON.stringify(xaiBody),
+							signal: requestSignal,
+						});
+						const rawText = await resp.text();
+						if (!resp.ok) {
+							let message = rawText;
+							try {
+								const parsedErr = JSON.parse(rawText) as { error?: { message?: string } };
+								message = parsedErr.error?.message ?? message;
+							} catch {
+								// Keep raw text.
+							}
+							throw Object.assign(new Error(`xAI image request failed (${resp.status}): ${message}`), {
+								status: resp.status,
+							});
+						}
+						return rawText;
+					},
+					{ signal: requestSignal },
+				);
 
 				const xaiData = JSON.parse(xaiRawText) as {
 					data?: Array<{ b64_json?: string; url?: string }>;
@@ -1269,30 +1308,34 @@ export const imageGenTool: CustomTool<typeof imageGenSchema, ImageGenToolDetails
 					messages: [{ role: "user" as const, content: contentParts }],
 				};
 
-				const response = await fetch("https://openrouter.ai/api/v1/chat/completions", {
-					method: "POST",
-					headers: {
-						"Content-Type": "application/json",
-						Authorization: `Bearer ${apiKey.apiKey}`,
-						"HTTP-Referer": "https://omp.sh/",
-						"X-OpenRouter-Title": "Oh-My-Pi",
-						"X-OpenRouter-Categories": "cli-agent",
-					},
-					body: JSON.stringify(requestBody),
-					signal: requestSignal,
-				});
-
-				const rawText = await response.text();
-				if (!response.ok) {
-					let message = rawText;
-					try {
-						const parsed = JSON.parse(rawText) as { error?: { message?: string } };
-						message = parsed.error?.message ?? message;
-					} catch {
-						// Keep raw text.
+				const rawText = await withAuth(apiKey.apiKey, async key => {
+					const resp = await fetch("https://openrouter.ai/api/v1/chat/completions", {
+						method: "POST",
+						headers: {
+							"Content-Type": "application/json",
+							Authorization: `Bearer ${key}`,
+							"HTTP-Referer": "https://omp.sh/",
+							"X-OpenRouter-Title": "Oh-My-Pi",
+							"X-OpenRouter-Categories": "cli-agent",
+						},
+						body: JSON.stringify(requestBody),
+						signal: requestSignal,
+					});
+					const text = await resp.text();
+					if (!resp.ok) {
+						let message = text;
+						try {
+							const parsed = JSON.parse(text) as { error?: { message?: string } };
+							message = parsed.error?.message ?? message;
+						} catch {
+							// Keep raw text.
+						}
+						throw Object.assign(new Error(`OpenRouter image request failed (${resp.status}): ${message}`), {
+							status: resp.status,
+						});
 					}
-					throw new Error(`OpenRouter image request failed (${response.status}): ${message}`);
-				}
+					return text;
+				});
 
 				const data = JSON.parse(rawText) as OpenRouterResponse;
 				const message = data.choices?.[0]?.message;
@@ -1360,30 +1403,34 @@ export const imageGenTool: CustomTool<typeof imageGenSchema, ImageGenToolDetails
 				generationConfig,
 			};
 
-			const response = await fetch(
-				`https://generativelanguage.googleapis.com/v1beta/models/${encodeURIComponent(model)}:generateContent`,
-				{
-					method: "POST",
-					headers: {
-						"Content-Type": "application/json",
-						"x-goog-api-key": apiKey.apiKey,
+			const rawText = await withAuth(apiKey.apiKey, async key => {
+				const resp = await fetch(
+					`https://generativelanguage.googleapis.com/v1beta/models/${encodeURIComponent(model)}:generateContent`,
+					{
+						method: "POST",
+						headers: {
+							"Content-Type": "application/json",
+							"x-goog-api-key": key,
+						},
+						body: JSON.stringify(requestBody),
+						signal: requestSignal,
 					},
-					body: JSON.stringify(requestBody),
-					signal: requestSignal,
-				},
-			);
-
-			const rawText = await response.text();
-			if (!response.ok) {
-				let message = rawText;
-				try {
-					const parsed = JSON.parse(rawText) as { error?: { message?: string } };
-					message = parsed.error?.message ?? message;
-				} catch {
-					// Keep raw text.
+				);
+				const text = await resp.text();
+				if (!resp.ok) {
+					let message = text;
+					try {
+						const parsed = JSON.parse(text) as { error?: { message?: string } };
+						message = parsed.error?.message ?? message;
+					} catch {
+						// Keep raw text.
+					}
+					throw Object.assign(new Error(`Gemini image request failed (${resp.status}): ${message}`), {
+						status: resp.status,
+					});
 				}
-				throw new Error(`Gemini image request failed (${response.status}): ${message}`);
-			}
+				return text;
+			});
 
 			const data = JSON.parse(rawText) as GeminiGenerateContentResponse;
 			const responseParts = combineParts(data);
diff --git a/packages/coding-agent/src/tools/index.ts b/packages/coding-agent/src/tools/index.ts
index baf336e4c..13788ebfd 100644
--- a/packages/coding-agent/src/tools/index.ts
+++ b/packages/coding-agent/src/tools/index.ts
@@ -23,6 +23,7 @@ import type { CustomMessage } from "../session/messages";
 import type { ToolChoiceQueue } from "../session/tool-choice-queue";
 import { TaskTool } from "../task";
 import type { AgentOutputManager } from "../task/output-manager";
+import { countToolsForAutoDiscovery, resolveEffectiveToolDiscoveryMode } from "../tool-discovery/mode";
 import type { DiscoverableTool, DiscoverableToolSearchIndex } from "../tool-discovery/tool-index";
 import type { EventBus } from "../utils/event-bus";
 import { WebSearchTool } from "../web/search";
@@ -420,14 +421,11 @@ export async function createTools(session: ToolSession, toolNames?: string[]): P
 		}
 	}
 	// Resolve effective tool discovery mode.
-	// tools.discoveryMode takes precedence; mcp.discoveryMode is a back-compat alias for "mcp-only".
-	const toolsDiscoveryMode = session.settings.get("tools.discoveryMode");
-	const effectiveDiscoveryMode: "off" | "mcp-only" | "all" =
-		toolsDiscoveryMode !== "off"
-			? (toolsDiscoveryMode as "off" | "mcp-only" | "all")
-			: session.settings.get("mcp.discoveryMode")
-				? "mcp-only"
-				: "off";
+	// tools.discoveryMode controls the new modes; mcp.discoveryMode remains a back-compat alias for "mcp-only".
+	const effectiveDiscoveryMode = resolveEffectiveToolDiscoveryMode(
+		session.settings,
+		countToolsForAutoDiscovery(requestedTools ?? Object.keys(BUILTIN_TOOLS)),
+	);
 	const discoveryActive = effectiveDiscoveryMode !== "off";
 
 	const allTools: Record<string, ToolFactory> = { ...BUILTIN_TOOLS, ...HIDDEN_TOOLS };
diff --git a/packages/coding-agent/src/tools/inspect-image-renderer.ts b/packages/coding-agent/src/tools/inspect-image-renderer.ts
index 75a80e0ea..604a04d52 100644
--- a/packages/coding-agent/src/tools/inspect-image-renderer.ts
+++ b/packages/coding-agent/src/tools/inspect-image-renderer.ts
@@ -2,8 +2,8 @@ import type { Component } from "@oh-my-pi/pi-tui";
 import { Text } from "@oh-my-pi/pi-tui";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import type { Theme } from "../modes/theme/theme";
-import { renderStatusLine } from "../tui";
-import { formatExpandHint, replaceTabs, shortenPath, truncateToWidth } from "./render-utils";
+import { framedBlock, renderStatusLine } from "../tui";
+import { formatErrorDetail, formatExpandHint, replaceTabs, shortenPath, truncateToWidth } from "./render-utils";
 
 interface InspectImageRenderArgs {
 	path?: string;
@@ -27,17 +27,21 @@ const INSPECT_OUTPUT_COLLAPSED_LINES = 4;
 const INSPECT_OUTPUT_EXPANDED_LINES = 16;
 const INSPECT_OUTPUT_LINE_WIDTH = 120;
 
+function questionLine(question: string, uiTheme: Theme): string {
+	return `${uiTheme.fg("dim", "Question:")} ${uiTheme.fg("accent", truncateToWidth(replaceTabs(question), INSPECT_QUESTION_PREVIEW_WIDTH))}`;
+}
+
 export const inspectImageToolRenderer = {
 	renderCall(args: InspectImageRenderArgs, _options: RenderResultOptions, uiTheme: Theme): Component {
 		const rawPath = args.path ?? "";
 		const pathDisplay = rawPath ? shortenPath(rawPath) : "…";
-		const header = renderStatusLine({ icon: "pending", title: "Inspect Image", description: pathDisplay }, uiTheme);
+		const header = renderStatusLine({ icon: "pending", title: "Inspect", description: pathDisplay }, uiTheme);
 		const question = args.question?.trim();
-		if (!question) {
-			return new Text(header, 0, 0);
-		}
-		const questionLine = ` ${uiTheme.fg("dim", uiTheme.tree.last)} ${uiTheme.fg("dim", "Question:")} ${uiTheme.fg("accent", truncateToWidth(replaceTabs(question), INSPECT_QUESTION_PREVIEW_WIDTH))}`;
-		return new Text(`${header}\n${questionLine}`, 0, 0);
+		// Call is at most a status line plus a one-line question — too small to box.
+		// The container renders a lone Text cleanly with no chrome.
+		if (!question) return new Text(header, 0, 0);
+		const tree = ` ${uiTheme.fg("dim", uiTheme.tree.last)} ${questionLine(question, uiTheme)}`;
+		return new Text(`${header}\n${tree}`, 0, 0);
 	},
 
 	renderResult(
@@ -49,55 +53,73 @@ export const inspectImageToolRenderer = {
 		const details = result.details;
 		const rawPath = details?.imagePath ?? args?.path ?? "";
 		const pathDisplay = rawPath ? shortenPath(rawPath) : "image";
-		const metaParts: string[] = [];
-		if (details?.model) metaParts.push(details.model);
-		if (details?.mimeType) metaParts.push(details.mimeType);
 		const header = renderStatusLine(
 			{
 				icon: result.isError ? "error" : "success",
-				title: "Inspect Image",
+				title: "Inspect",
 				description: pathDisplay,
 			},
 			uiTheme,
 		);
 
-		const lines: string[] = [header];
 		const question = args?.question?.trim();
-		if (question) {
-			lines.push(
-				` ${uiTheme.fg("dim", uiTheme.tree.branch)} ${uiTheme.fg("dim", "Question:")} ${uiTheme.fg("accent", truncateToWidth(replaceTabs(question), INSPECT_QUESTION_PREVIEW_WIDTH))}`,
-			);
-		}
-
 		const outputText = result.content.find(content => content.type === "text")?.text?.trimEnd() ?? "";
+
+		if (result.isError) {
+			return framedBlock(uiTheme, width => {
+				const bodyLines: string[] = [];
+				if (question) bodyLines.push(questionLine(question, uiTheme));
+				bodyLines.push(formatErrorDetail(outputText || "inspection failed", uiTheme));
+				return {
+					header,
+					sections: [{ lines: bodyLines }],
+					state: "error",
+					borderColor: "error",
+					applyBg: false,
+					width,
+				};
+			});
+		}
+
+		const metaParts: string[] = [];
+		if (details?.model) metaParts.push(details.model);
+		if (details?.mimeType) metaParts.push(details.mimeType);
+		const metaLine = metaParts.length > 0 ? uiTheme.fg("dim", metaParts.join(" · ")) : "";
+
+		// No answer text: nothing worth boxing — keep it to a clean status line
+		// (plus a trailing meta line, when present).
 		if (!outputText) {
-			lines.push(uiTheme.fg("dim", "(no output)"));
-			if (metaParts.length > 0) {
-				lines.push("");
-				lines.push(uiTheme.fg("dim", metaParts.join(" · ")));
+			return new Text(metaLine ? `${header}\n${metaLine}` : header, 0, 0);
+		}
+
+		return framedBlock(uiTheme, width => {
+			const bodyLines: string[] = [];
+			if (question) {
+				bodyLines.push(questionLine(question, uiTheme));
+				bodyLines.push("");
 			}
-			return new Text(lines.join("\n"), 0, 0);
-		}
 
-		lines.push("");
-		const outputLines = replaceTabs(outputText).split("\n");
-		const maxLines = options.expanded ? INSPECT_OUTPUT_EXPANDED_LINES : INSPECT_OUTPUT_COLLAPSED_LINES;
-		for (const line of outputLines.slice(0, maxLines)) {
-			lines.push(uiTheme.fg("toolOutput", truncateToWidth(line, INSPECT_OUTPUT_LINE_WIDTH)));
-		}
+			const outputLines = replaceTabs(outputText).split("\n");
+			const maxLines = options.expanded ? INSPECT_OUTPUT_EXPANDED_LINES : INSPECT_OUTPUT_COLLAPSED_LINES;
+			for (const line of outputLines.slice(0, maxLines)) {
+				bodyLines.push(uiTheme.fg("toolOutput", truncateToWidth(line, INSPECT_OUTPUT_LINE_WIDTH)));
+			}
+			if (outputLines.length > maxLines) {
+				const remaining = outputLines.length - maxLines;
+				const hint = formatExpandHint(uiTheme, options.expanded, true);
+				bodyLines.push(`${uiTheme.fg("dim", `… ${remaining} more lines`)}${hint ? ` ${hint}` : ""}`);
+			}
 
-		if (outputLines.length > maxLines) {
-			const remaining = outputLines.length - maxLines;
-			const hint = formatExpandHint(uiTheme, options.expanded, true);
-			lines.push(`${uiTheme.fg("dim", `… ${remaining} more lines`)}${hint ? ` ${hint}` : ""}`);
-		}
-
-		if (metaParts.length > 0) {
-			lines.push("");
-			lines.push(uiTheme.fg("dim", metaParts.join(" · ")));
-		}
-
-		return new Text(lines.join("\n"), 0, 0);
+			return {
+				header,
+				headerMeta: metaLine || undefined,
+				sections: [{ lines: bodyLines }],
+				state: "success",
+				borderColor: "borderMuted",
+				applyBg: false,
+				width,
+			};
+		});
 	},
 	mergeCallAndResult: true,
 };
diff --git a/packages/coding-agent/src/tools/inspect-image.ts b/packages/coding-agent/src/tools/inspect-image.ts
index 3759e5ed1..e7db9fc3c 100644
--- a/packages/coding-agent/src/tools/inspect-image.ts
+++ b/packages/coding-agent/src/tools/inspect-image.ts
@@ -4,6 +4,7 @@ import { type Api, completeSimple, type Model } from "@oh-my-pi/pi-ai";
 import { prompt } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
 import { extractTextContent } from "../commit/utils";
+
 import { expandRoleAlias, resolveModelFromString } from "../config/model-resolver";
 import inspectImageDescription from "../prompts/tools/inspect-image.md" with { type: "text" };
 import inspectImageSystemPromptTemplate from "../prompts/tools/inspect-image-system.md" with { type: "text" };
@@ -136,7 +137,13 @@ export class InspectImageTool implements AgentTool<typeof inspectImageSchema, In
 					},
 				],
 			},
-			{ apiKey, signal },
+			{
+				apiKey: modelRegistry.resolver(model.provider, {
+					sessionId: this.session.getSessionId?.() ?? undefined,
+					baseUrl: model.baseUrl,
+				}),
+				signal,
+			},
 			{ telemetry, oneshotKind: "inspect_image", completeImpl: this.completeImageRequest },
 		);
 
diff --git a/packages/coding-agent/src/tools/job.ts b/packages/coding-agent/src/tools/job.ts
index 18defb3f7..c0f326d5d 100644
--- a/packages/coding-agent/src/tools/job.ts
+++ b/packages/coding-agent/src/tools/job.ts
@@ -396,7 +396,7 @@ export const jobToolRenderer = {
 	inline: true,
 
 	renderCall(args: JobRenderArgs, _options: RenderResultOptions, uiTheme: Theme): Component {
-		const text = renderStatusLine({ icon: "pending", title: "Job", description: describeTarget(args) }, uiTheme);
+		const text = renderStatusLine({ icon: "pending", title: describeTarget(args) || "Job" }, uiTheme);
 		return new Text(text, 0, 0);
 	},
 
@@ -410,7 +410,7 @@ export const jobToolRenderer = {
 
 		if (jobs.length === 0) {
 			const fallback = result.content?.find(c => c.type === "text")?.text || "No jobs to process";
-			const header = renderStatusLine({ icon: "warning", title: "Job", description: describeTarget(args) }, uiTheme);
+			const header = renderStatusLine({ icon: "warning", title: describeTarget(args) || "Job" }, uiTheme);
 			return new Text([header, formatEmptyMessage(fallback, uiTheme)].join("\n"), 0, 0);
 		}
 
@@ -433,8 +433,7 @@ export const jobToolRenderer = {
 			{
 				icon: headerIcon,
 				spinnerFrame: counts.running > 0 ? options.spinnerFrame : undefined,
-				title: "Job",
-				description,
+				title: description,
 				meta,
 			},
 			uiTheme,
diff --git a/packages/coding-agent/src/tools/memory-render.ts b/packages/coding-agent/src/tools/memory-render.ts
index 79e7da17a..2430eef6c 100644
--- a/packages/coding-agent/src/tools/memory-render.ts
+++ b/packages/coding-agent/src/tools/memory-render.ts
@@ -4,7 +4,10 @@
  *
  * These keep the transcript terse — one status line plus, for `retain`, one
  * `Remember: …` line per stored item — instead of the generic JSON arg tree,
- * which exploded multi-line memory blobs into an unreadable wall.
+ * which exploded multi-line memory blobs into an unreadable wall. The tool
+ * container is a transparent passthrough, so these renderers stay frameless:
+ * a status line with a couple of dim bullets reads far cleaner than boxing a
+ * one-line memory note.
  */
 import type { Component } from "@oh-my-pi/pi-tui";
 import { Text } from "@oh-my-pi/pi-tui";
diff --git a/packages/coding-agent/src/tools/plan-mode-guard.ts b/packages/coding-agent/src/tools/plan-mode-guard.ts
index 5212a2906..c27fe2fea 100644
--- a/packages/coding-agent/src/tools/plan-mode-guard.ts
+++ b/packages/coding-agent/src/tools/plan-mode-guard.ts
@@ -1,4 +1,3 @@
-import * as path from "node:path";
 import { resolveLocalUrlToPath, resolveVaultUrlToPath } from "../internal-urls";
 import type { ToolSession } from ".";
 import { normalizeLocalScheme, resolveToCwd } from "./path-utils";
@@ -6,10 +5,19 @@ import { ToolError } from "./tool-errors";
 
 const VAULT_SCHEME_PREFIX = "vault:";
 const LOCAL_SCHEME_PREFIX = "local:";
-const PLAN_ALIAS_FILE = "PLAN.md";
-const LOCAL_PLAN_ALIAS = "local://PLAN.md";
 
-function resolveRawPath(session: ToolSession, targetPath: string): string {
+/** True when `targetPath` addresses the session-local artifact sandbox
+ *  (`local://…`). Those files are not part of the working tree, so plan mode
+ *  treats them as freely writable scratch/plan space. */
+function targetsLocalSandbox(targetPath: string): boolean {
+	return normalizeLocalScheme(targetPath).startsWith(LOCAL_SCHEME_PREFIX);
+}
+
+/**
+ * Resolve a write/edit target to its absolute filesystem path, honoring the
+ * `local://` and `vault://` schemes. Plain paths resolve against the session cwd.
+ */
+export function resolvePlanPath(session: ToolSession, targetPath: string): string {
 	const normalized = normalizeLocalScheme(targetPath);
 	if (normalized.startsWith(LOCAL_SCHEME_PREFIX)) {
 		return resolveLocalUrlToPath(normalized, {
@@ -25,37 +33,12 @@ function resolveRawPath(session: ToolSession, targetPath: string): string {
 	return resolveToCwd(normalized, session.cwd);
 }
 
-function isPlanAliasTarget(session: ToolSession, targetPath: string, resolved: string): boolean {
-	const normalized = normalizeLocalScheme(targetPath);
-	if (normalized === LOCAL_PLAN_ALIAS) return true;
-	return resolved === resolveToCwd(PLAN_ALIAS_FILE, session.cwd);
-}
-
 /**
- * Resolve a write/edit target to its absolute filesystem path.
- *
- * In plan mode, transparently redirects `PLAN.md` aliases and targets whose
- * basename matches the plan file's basename to the canonical plan file
- * location at `state.planFilePath`. This lets `write` and `edit` accept the
- * habitual plan filename after approval even when the active artifact has a
- * titled path such as `local://APPROVED.md`.
- *
- * Outside plan mode (or when the basename does not match) this is a no-op.
+ * Plan mode keeps the working tree read-only while letting the agent draft its
+ * plan. Writes and edits to the `local://` artifact sandbox are allowed (that is
+ * where the plan and any scratch notes live); anything that would touch the
+ * working tree — or rename/delete a file — is rejected.
  */
-export function resolvePlanPath(session: ToolSession, targetPath: string): string {
-	const resolved = resolveRawPath(session, targetPath);
-
-	const state = session.getPlanModeState?.();
-	if (!state?.enabled) return resolved;
-
-	const planResolved = resolveRawPath(session, state.planFilePath);
-	if (resolved === planResolved) return resolved;
-	if (isPlanAliasTarget(session, targetPath, resolved)) return planResolved;
-	if (path.basename(resolved) !== path.basename(planResolved)) return resolved;
-
-	return planResolved;
-}
-
 export function enforcePlanModeWrite(
 	session: ToolSession,
 	targetPath: string,
@@ -64,9 +47,6 @@ export function enforcePlanModeWrite(
 	const state = session.getPlanModeState?.();
 	if (!state?.enabled) return;
 
-	const resolvedTarget = resolvePlanPath(session, targetPath);
-	const resolvedPlan = resolvePlanPath(session, state.planFilePath);
-
 	if (options?.move) {
 		throw new ToolError("Plan mode: renaming files is not allowed.");
 	}
@@ -75,7 +55,9 @@ export function enforcePlanModeWrite(
 		throw new ToolError("Plan mode: deleting files is not allowed.");
 	}
 
-	if (resolvedTarget !== resolvedPlan) {
-		throw new ToolError(`Plan mode: only the plan file may be modified (${state.planFilePath}).`);
-	}
+	if (targetsLocalSandbox(targetPath)) return;
+
+	throw new ToolError(
+		"Plan mode: the working tree is read-only. Write your plan to a local://<slug>-plan.md file instead.",
+	);
 }
diff --git a/packages/coding-agent/src/tools/read.ts b/packages/coding-agent/src/tools/read.ts
index aec92c365..020ac82b4 100644
--- a/packages/coding-agent/src/tools/read.ts
+++ b/packages/coding-agent/src/tools/read.ts
@@ -29,12 +29,12 @@ import {
 	truncateLine,
 } from "../session/streaming-output";
 import { fileHyperlink, renderCodeCell, renderMarkdownCell, renderStatusLine, tryResolveInternalUrlSync } from "../tui";
-import { CachedOutputBlock } from "../tui/output-block";
+import { CachedOutputBlock, markFramedBlockComponent } from "../tui/output-block";
 import { resolveFileDisplayMode } from "../utils/file-display-mode";
 import { ImageInputTooLargeError, loadImageInput, MAX_IMAGE_INPUT_BYTES } from "../utils/image-loading";
 import { convertFileWithMarkit } from "../utils/markit";
 import { buildDirectoryTree, type DirectoryTree } from "../workspace-tree";
-import { type ArchiveReader, openArchive, parseArchivePathCandidates } from "./archive-reader";
+import { type ArchiveReader, formatArchiveEntryLines, openArchive, parseArchivePathCandidates } from "./archive-reader";
 import {
 	type ConflictEntry,
 	type ConflictScope,
@@ -1154,17 +1154,10 @@ export class ReadTool implements AgentTool<typeof readSchema, ReadToolDetails> {
 		const limitedEntries = listLimit.items;
 		const limitMeta = listLimit.meta;
 
-		const results: string[] = [];
-		for (const entry of limitedEntries) {
+		for (let index = 0; index < limitedEntries.length; index++) {
 			throwIfAborted(signal);
-			if (entry.isDirectory) {
-				results.push(`${entry.name}/`);
-				continue;
-			}
-
-			const sizeSuffix = entry.size > 0 ? ` (${formatBytes(entry.size)})` : "";
-			results.push(`${entry.name}${sizeSuffix}`);
 		}
+		const results = formatArchiveEntryLines(limitedEntries);
 
 		const output = results.length > 0 ? results.join("\n") : "(empty archive directory)";
 		const text = prependSuffixResolutionNotice(output, details.suffixResolution);
@@ -2337,10 +2330,20 @@ function firstReadSelectorLine(sel: string | undefined): number | undefined {
 	}
 }
 
+/** Absolute fs path the read result actually resolved to, used as the OSC 8 link
+ * target when the structured `resolvedPath` isn't set (the common plain-file and
+ * image reads only record the path in `meta.source`). URL/internal sources are
+ * not fs paths, so only `type: "path"` qualifies. */
+function readSourceFsPath(details: ReadToolDetails | undefined): string | undefined {
+	const source = details?.meta?.source;
+	return source?.type === "path" ? source.value : undefined;
+}
+
 function formatReadPathLink(
 	rawPath: string,
 	options: {
 		resolvedPath?: string;
+		sourcePath?: string;
 		suffixResolution?: { from: string; to: string };
 		offset?: number;
 		fallbackLabel?: string;
@@ -2352,7 +2355,7 @@ function formatReadPathLink(
 	const plainDisplayPath = options.suffixResolution
 		? shortenPath(options.suffixResolution.to)
 		: shortenPath(basePath || options.resolvedPath || options.fallbackLabel || rawPath);
-	const target = options.resolvedPath ?? tryResolveInternalUrlSync(basePath);
+	const target = options.resolvedPath ?? options.sourcePath ?? tryResolveInternalUrlSync(basePath);
 	const line = firstReadSelectorLine(split.sel) ?? options.offset;
 	const linkOptions = line !== undefined ? { line } : undefined;
 	const displayPath = target ? fileHyperlink(target, plainDisplayPath, linkOptions) : plainDisplayPath;
@@ -2403,7 +2406,9 @@ export const readToolRenderer = {
 			const rawErrorText = result.content?.find(c => c.type === "text")?.text ?? "";
 			const errorText = (rawErrorText || "Unknown error").replace(/^Error:\s*/, "");
 			const rawPath = args?.file_path || args?.path || "";
-			const filePath = formatReadPathLink(rawPath, { offset: args?.offset }) || shortenPath(rawPath);
+			const filePath =
+				formatReadPathLink(rawPath, { offset: args?.offset, sourcePath: readSourceFsPath(result.details) }) ||
+				shortenPath(rawPath);
 			let title = filePath ? `Read ${filePath}` : "Read";
 			if (args?.offset !== undefined || args?.limit !== undefined) {
 				const startLine = args.offset ?? 1;
@@ -2413,11 +2418,11 @@ export const readToolRenderer = {
 			const header = renderStatusLine({ icon: "error", title }, uiTheme);
 			const errorLines = errorText.split("\n").map(line => uiTheme.fg("error", replaceTabs(line)));
 			const outputBlock = new CachedOutputBlock();
-			return {
+			return markFramedBlockComponent({
 				render: (width: number) =>
 					outputBlock.render({ header, state: "error", sections: [{ lines: errorLines }], width }, uiTheme),
 				invalidate: () => outputBlock.invalidate(),
-			};
+			});
 		}
 		const details = result.details;
 		const rawText = result.content?.find(c => c.type === "text")?.text ?? "";
@@ -2454,6 +2459,7 @@ export const readToolRenderer = {
 			const suffix = details?.suffixResolution;
 			const displayPath = formatReadPathLink(rawPath, {
 				resolvedPath: details?.resolvedPath,
+				sourcePath: readSourceFsPath(details),
 				suffixResolution: suffix,
 				fallbackLabel: "image",
 			});
@@ -2465,7 +2471,7 @@ export const readToolRenderer = {
 			const detailLines = contentText ? contentText.split("\n").map(line => uiTheme.fg("toolOutput", line)) : [];
 			const lines = [...detailLines, ...warningLines];
 			const outputBlock = new CachedOutputBlock();
-			return {
+			return markFramedBlockComponent({
 				render: (width: number) =>
 					outputBlock.render(
 						{
@@ -2482,16 +2488,17 @@ export const readToolRenderer = {
 						uiTheme,
 					),
 				invalidate: () => outputBlock.invalidate(),
-			};
+			});
 		}
 
 		const suffix = details?.suffixResolution;
-		// resolvedPath is the absolute fs path for fs-backed reads (regular files plus
-		// local:// / memory:// / skill:// / artifact:// resources). Fall back to a sync
-		// resolver for fs-backed internal URLs so the title is clickable even before the
-		// result lands or if the handler didn't populate resolvedPath.
+		// resolvedPath is the absolute fs path when a read resolved/corrected the
+		// input (suffix match, internal URL, archive/sqlite/notebook); plain file
+		// reads only record the absolute path in meta.source, so fall back to that
+		// (and then to a sync internal-URL resolver) to keep the title clickable.
 		const displayPath = formatReadPathLink(rawPath, {
 			resolvedPath: details?.resolvedPath,
+			sourcePath: readSourceFsPath(details),
 			suffixResolution: suffix,
 			offset: args?.offset,
 		});
@@ -2514,7 +2521,7 @@ export const readToolRenderer = {
 		let cachedWidth: number | undefined;
 		let cachedExpanded: boolean | undefined;
 		let cachedLines: string[] | undefined;
-		return {
+		return markFramedBlockComponent({
 			render: (width: number) => {
 				const expanded = options.expanded;
 				if (cachedLines && cachedWidth === width && cachedExpanded === expanded) return cachedLines;
@@ -2551,7 +2558,7 @@ export const readToolRenderer = {
 				cachedExpanded = undefined;
 				cachedLines = undefined;
 			},
-		};
+		});
 	},
 	mergeCallAndResult: true,
 };
diff --git a/packages/coding-agent/src/tools/render-utils.ts b/packages/coding-agent/src/tools/render-utils.ts
index 5ade92d95..7179670c4 100644
--- a/packages/coding-agent/src/tools/render-utils.ts
+++ b/packages/coding-agent/src/tools/render-utils.ts
@@ -10,8 +10,9 @@ import * as path from "node:path";
 import type { ToolCallContext } from "@oh-my-pi/pi-agent-core";
 import type { Ellipsis } from "@oh-my-pi/pi-natives";
 import type { Component } from "@oh-my-pi/pi-tui";
-import { replaceTabs, truncateToWidth } from "@oh-my-pi/pi-tui";
+import { getKeybindings, replaceTabs, truncateToWidth } from "@oh-my-pi/pi-tui";
 import { pluralize } from "@oh-my-pi/pi-utils";
+import { formatKeyHints, type KeyId } from "../config/keybindings";
 import { settings } from "../config/settings";
 import type { Theme } from "../modes/theme/theme";
 import { Hasher } from "../tui/utils";
@@ -75,8 +76,16 @@ export const TRUNCATE_LENGTHS = {
 	SHORT: 40,
 } as const;
 
-/** Standard expand hint text */
-export const EXPAND_HINT = "(Ctrl+O for more)";
+/** Keybinding action that toggles tool-output expansion. */
+const EXPAND_ACTION = "app.tools.expand";
+/** Fallback key when no binding is resolvable (e.g. outside an interactive session). */
+const DEFAULT_EXPAND_KEY: KeyId = "ctrl+o";
+
+/** Human-readable key currently bound to tool-output expansion, e.g. `Ctrl+O`. */
+export function expandKeyHint(): string {
+	const keys = getKeybindings().getKeys(EXPAND_ACTION);
+	return formatKeyHints(keys.length > 0 ? keys : [DEFAULT_EXPAND_KEY]);
+}
 
 // =============================================================================
 // Text Truncation Utilities
@@ -150,7 +159,7 @@ export function formatStatusIcon(status: ToolUIStatus, theme: Theme, spinnerFram
 export function formatExpandHint(theme: Theme, expanded?: boolean, hasMore?: boolean): string {
 	if (expanded) return "";
 	if (hasMore === false) return "";
-	return theme.fg("dim", wrapBrackets(EXPAND_HINT, theme));
+	return theme.fg("dim", wrapBrackets(`${expandKeyHint()}: Expand`, theme));
 }
 
 /**
@@ -171,14 +180,74 @@ export function formatMoreItems(remaining: number, itemType: string): string {
 	return `… ${safeRemaining} more ${pluralize(itemType, safeRemaining)}`;
 }
 
+/**
+ * Maximum rows a tool's streaming/pending *call* preview may render before it is
+ * capped. This is intentionally conservative: the preview still sits inside a
+ * transcript that already consumed some viewport rows, and tool blocks carry
+ * extra chrome (status/header/border/"more lines"), so a "reasonable" raw code
+ * or command preview like 10-12 lines can still overflow and strand its top
+ * while the block is volatile. Keeping the live call window short avoids that
+ * across terminals without turning the transcript into an interactive scroller.
+ */
+export const CALL_PREVIEW_MAX_LINES = 6;
+
+/**
+ * Cap a pre-rendered pending/call preview to a bounded window. When truncated,
+ * show both the head and the live tail so the user can still see what the tool
+ * is currently writing while the volatile block stays short enough not to strand
+ * its top above the viewport. `Ctrl+O` widens the bounded window, but does not
+ * fully uncap live tool previews for the same reason.
+ *
+ * `prefix` (raw, e.g. a dim tree gutter) is prepended to the summary line so
+ * nested previews stay aligned.
+ */
+export function capPreviewLines(
+	lines: string[],
+	theme: Theme,
+	options: { max?: number; expanded?: boolean; prefix?: string } = {},
+): string[] {
+	const max = options.max ?? (options.expanded ? PREVIEW_LIMITS.EXPANDED_LINES : CALL_PREVIEW_MAX_LINES);
+	if (lines.length <= max) return lines;
+	if (max <= 1) {
+		const hint = formatExpandHint(theme, options.expanded, true);
+		const moreLine = `${formatMoreItems(lines.length, "line")}${hint ? ` ${hint}` : ""}`;
+		return [`${options.prefix ?? ""}${theme.fg("dim", moreLine)}`];
+	}
+	const bodyBudget = max - 1; // reserve one summary row
+	const headCount = Math.max(1, Math.ceil(bodyBudget / 2));
+	const tailCount = Math.max(1, bodyBudget - headCount);
+	const hidden = Math.max(0, lines.length - headCount - tailCount);
+	const hint = formatExpandHint(theme, options.expanded, true);
+	const moreLine = `${formatMoreItems(hidden, "line")}${hint ? ` ${hint}` : ""}`;
+	return [
+		...lines.slice(0, headCount),
+		`${options.prefix ?? ""}${theme.fg("dim", moreLine)}`,
+		...lines.slice(lines.length - tailCount),
+	];
+}
+
 export function formatMeta(meta: string[], theme: Theme): string {
 	return meta.length > 0 ? ` ${theme.fg("muted", meta.join(theme.sep.dot))}` : "";
 }
 
-export function formatErrorMessage(message: string | undefined, theme: Theme): string {
+function sanitizeErrorText(message: string | undefined): string {
 	const clean = (message ?? "").replace(/^Error:\s*/, "").trim();
-	const safe = clean ? replaceTabs(truncateToWidth(clean, TRUNCATE_LENGTHS.LINE)) : "Unknown error";
-	return `${theme.styledSymbol("status.error", "error")} ${theme.fg("error", `Error: ${safe}`)}`;
+	return clean ? replaceTabs(truncateToWidth(clean, TRUNCATE_LENGTHS.LINE)) : "Unknown error";
+}
+
+export function formatErrorMessage(message: string | undefined, theme: Theme): string {
+	return `${theme.styledSymbol("status.error", "error")} ${theme.fg("error", `Error: ${sanitizeErrorText(message)}`)}`;
+}
+
+/**
+ * Error message rendered as a subordinate detail line beneath a status header
+ * that already carries the error icon (e.g. `✘ Write: <path>`). The header's
+ * icon already signals failure, so this omits the redundant error symbol and
+ * "Error:" prefix that `formatErrorMessage` adds for standalone single-line
+ * errors, indenting two columns to sit under the header title instead.
+ */
+export function formatErrorDetail(message: string | undefined, theme: Theme): string {
+	return `  ${theme.fg("error", sanitizeErrorText(message))}`;
 }
 
 export function formatEmptyMessage(message: string, theme: Theme): string {
@@ -676,36 +745,6 @@ export function capParseErrors(
 // Renderer helpers shared by search / find / ast tools
 // =============================================================================
 
-/**
- * Group `rawLines` by blank-line separators, mirroring the historical search /
- * ast-grep / ast-edit renderer behavior: if any blank line is present, splits on
- * runs of blank lines; otherwise collapses non-empty lines into a single group.
- */
-export function splitGroupsByBlankLine(rawLines: string[]): string[][] {
-	const hasSeparators = rawLines.some(line => line.trim().length === 0);
-	const groups: string[][] = [];
-	if (hasSeparators) {
-		let current: string[] = [];
-		for (const line of rawLines) {
-			if (line.trim().length === 0) {
-				if (current.length > 0) {
-					groups.push(current);
-					current = [];
-				}
-				continue;
-			}
-			current.push(line);
-		}
-		if (current.length > 0) groups.push(current);
-	} else {
-		const nonEmpty = rawLines.filter(line => line.trim().length > 0);
-		if (nonEmpty.length > 0) {
-			groups.push(nonEmpty);
-		}
-	}
-	return groups;
-}
-
 /**
  * Standard width+expand keyed render cache used by every search-style tool
  * renderer. `compute` re-runs only when the cache key changes; the returned
@@ -714,6 +753,7 @@ export function splitGroupsByBlankLine(rawLines: string[]): string[][] {
 export function createCachedComponent(
 	getExpanded: () => boolean,
 	compute: (width: number, expanded: boolean) => string[],
+	options: { paddingX?: number } = {},
 ): Component {
 	let cached: { key: bigint; lines: string[] } | undefined;
 	return {
@@ -721,9 +761,13 @@ export function createCachedComponent(
 			const expanded = getExpanded();
 			const key = new Hasher().bool(expanded).u32(width).digest();
 			if (cached?.key === key) return cached.lines;
-			const lines = compute(width, expanded);
-			cached = { key, lines };
-			return lines;
+			const paddingX = Math.max(0, options.paddingX ?? 0);
+			const innerWidth = Math.max(1, width - paddingX * 2);
+			const lines = compute(innerWidth, expanded);
+			const pad = paddingX === 0 ? "" : " ".repeat(paddingX);
+			const paddedLines = paddingX === 0 ? lines : lines.map(line => `${pad}${line}${pad}`);
+			cached = { key, lines: paddedLines };
+			return paddedLines;
 		},
 		invalidate() {
 			cached = undefined;
diff --git a/packages/coding-agent/src/tools/renderers.ts b/packages/coding-agent/src/tools/renderers.ts
index eebbe57a1..4f74bd090 100644
--- a/packages/coding-agent/src/tools/renderers.ts
+++ b/packages/coding-agent/src/tools/renderers.ts
@@ -31,7 +31,7 @@ import { sshToolRenderer } from "./ssh";
 import { todoToolRenderer } from "./todo";
 import { writeToolRenderer } from "./write";
 
-type ToolRenderer = {
+export type ToolRenderer = {
 	renderCall: (args: unknown, options: RenderResultOptions, theme: Theme) => Component;
 	renderResult: (
 		result: { content: Array<{ type: string; text?: string }>; details?: unknown; isError?: boolean },
@@ -40,6 +40,21 @@ type ToolRenderer = {
 		args?: unknown,
 	) => Component;
 	mergeCallAndResult?: boolean;
+	/**
+	 * While a tool's preview is still streaming, report whether the
+	 * currently-rendered preview is append-only: its rows only grow at the bottom
+	 * and never re-layout above the bottom live region (a full, top-anchored
+	 * content/code preview). The transcript reports this up to the TUI so a
+	 * streaming preview taller than the viewport commits its scrolled-off head to
+	 * native scrollback instead of dropping it (see
+	 * `ToolExecutionComponent.isTranscriptBlockAppendOnly`). `result` is the
+	 * latest (possibly partial) tool result, or `undefined` before one exists —
+	 * `eval`/`bash` use its presence to defer committing until the streamed input
+	 * (code) has finalized. Omit (or return `false`) for previews that slide a
+	 * tail window or later collapse to a compact result — committing their head
+	 * would strand stale rows.
+	 */
+	isStreamingPreviewAppendOnly?: (args: unknown, options: RenderResultOptions, result?: unknown) => boolean;
 	/** Render without background box, inline in the response flow */
 	inline?: boolean;
 };
diff --git a/packages/coding-agent/src/tools/report-tool-issue.ts b/packages/coding-agent/src/tools/report-tool-issue.ts
index 0526612ed..e723ff491 100644
--- a/packages/coding-agent/src/tools/report-tool-issue.ts
+++ b/packages/coding-agent/src/tools/report-tool-issue.ts
@@ -41,7 +41,7 @@ function buildReportToolIssueParams(activeBuiltinNames: readonly string[]) {
 }
 
 export function isAutoQaEnabled(settings?: Settings): boolean {
-	return $flag("PI_AUTO_QA") || !!settings?.get("dev.autoqa");
+	return $flag("PI_AUTO_QA", !!settings?.get("dev.autoqa"));
 }
 
 // ───────────────────────────────────────────────────────────────────────────
diff --git a/packages/coding-agent/src/tools/resolve.ts b/packages/coding-agent/src/tools/resolve.ts
index 9899e9376..15f22d406 100644
--- a/packages/coding-agent/src/tools/resolve.ts
+++ b/packages/coding-agent/src/tools/resolve.ts
@@ -187,6 +187,20 @@ export class ResolveTool implements AgentTool<typeof resolveSchema, ResolveToolD
 		return untilAborted(signal, async () => {
 			const invoker = this.session.peekQueueInvoker?.() ?? this.session.peekStandingResolveHandler?.();
 			if (!invoker) {
+				// `discard` is a request to cancel/abort a staged action. When nothing is
+				// pending, the desired end-state (no staged change) already holds, so honor
+				// it as a successful cancellation instead of surfacing a hard error to the
+				// model. `apply` still errors — there is nothing to apply.
+				if (params.action === "discard") {
+					return {
+						content: [{ type: "text" as const, text: "Nothing to discard; no pending action remains." }],
+						details: {
+							action: "discard",
+							reason: params.reason,
+							...(params.extra != null ? { extra: params.extra } : {}),
+						},
+					};
+				}
 				throw new ToolError("No pending action to resolve. Nothing to apply or discard.");
 			}
 			const result = (await invoker(params)) as AgentToolResult<ResolveToolDetails>;
diff --git a/packages/coding-agent/src/tools/search-tool-bm25.ts b/packages/coding-agent/src/tools/search-tool-bm25.ts
index d2b37450e..466b5e02a 100644
--- a/packages/coding-agent/src/tools/search-tool-bm25.ts
+++ b/packages/coding-agent/src/tools/search-tool-bm25.ts
@@ -6,6 +6,7 @@ import * as z from "zod/v4";
 import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import type { Theme } from "../modes/theme/theme";
 import searchToolBm25Description from "../prompts/tools/search-tool-bm25.md" with { type: "text" };
+import { resolveEffectiveToolDiscoveryMode } from "../tool-discovery/mode";
 import {
 	buildDiscoverableToolSearchIndex,
 	type DiscoverableTool,
@@ -15,9 +16,9 @@ import {
 	searchDiscoverableTools,
 	summarizeDiscoverableTools,
 } from "../tool-discovery/tool-index";
-import { renderStatusLine, renderTreeList, truncateToWidth } from "../tui";
+import { framedBlock, renderStatusLine, truncateToWidth } from "../tui";
 import type { ToolSession } from ".";
-import { formatCount, replaceTabs, TRUNCATE_LENGTHS } from "./render-utils";
+import { formatCount, formatExpandHint, formatMoreItems, replaceTabs, TRUNCATE_LENGTHS } from "./render-utils";
 import { ToolError } from "./tool-errors";
 
 const DEFAULT_LIMIT = 8;
@@ -170,6 +171,25 @@ function renderMatchLines(match: SearchToolBm25Match, theme: Theme): string[] {
 	return lines;
 }
 
+function renderMatchBullets(tools: SearchToolBm25Match[], expanded: boolean, theme: Theme): string[] {
+	const shown = expanded ? tools.length : Math.min(tools.length, COLLAPSED_MATCH_LIMIT);
+	const bullet = theme.fg("dim", theme.format.bullet);
+	const lines: string[] = [];
+	for (let i = 0; i < shown; i++) {
+		const itemLines = renderMatchLines(tools[i]!, theme);
+		lines.push(`${bullet} ${itemLines[0]}`);
+		for (let j = 1; j < itemLines.length; j++) {
+			lines.push(`  ${itemLines[j]}`);
+		}
+	}
+	const remaining = tools.length - shown;
+	if (remaining > 0) {
+		const hint = formatExpandHint(theme, expanded, true);
+		lines.push(`${theme.fg("muted", formatMoreItems(remaining, "tool"))}${hint ? ` ${hint}` : ""}`);
+	}
+	return lines;
+}
+
 function renderFallbackResult(text: string, theme: Theme): Component {
 	const header = renderStatusLine({ icon: "warning", title: TOOL_DISCOVERY_TITLE }, theme);
 	const bodyLines = (text || "Tool discovery completed")
@@ -198,12 +218,9 @@ export class SearchToolBm25Tool implements AgentTool<typeof searchToolBm25Schema
 	constructor(private readonly session: ToolSession) {}
 
 	static createIf(session: ToolSession): SearchToolBm25Tool | null {
-		// Active when new tools.discoveryMode is non-"off" or legacy mcp.discoveryMode is true
-		const toolsDiscoveryMode = session.settings.get("tools.discoveryMode");
-		const active =
-			(toolsDiscoveryMode !== undefined && toolsDiscoveryMode !== "off") ||
-			session.settings.get("mcp.discoveryMode") === true;
-		if (!active) return null;
+		// Direct createTools() calls do not know the final MCP/extension catalog yet, so
+		// auto mode is activated later by createAgentSession after the full registry exists.
+		if (resolveEffectiveToolDiscoveryMode(session.settings, 0) === "off") return null;
 		return supportsToolDiscoveryExecution(session) ? new SearchToolBm25Tool(session) : null;
 	}
 
@@ -273,14 +290,11 @@ export const searchToolBm25Renderer = {
 	renderCall(args: SearchToolBm25Params, _options: RenderResultOptions, uiTheme: Theme): Component {
 		const query = typeof args.query === "string" ? replaceTabs(args.query.trim()) : "";
 		const meta = args.limit ? [`limit:${args.limit}`] : [];
-		return new Text(
-			renderStatusLine(
-				{ icon: "pending", title: TOOL_DISCOVERY_TITLE, description: query || "(empty query)", meta },
-				uiTheme,
-			),
-			0,
-			0,
+		const header = renderStatusLine(
+			{ icon: "pending", title: TOOL_DISCOVERY_TITLE, description: query || "(empty query)", meta },
+			uiTheme,
 		);
+		return new Text(header, 0, 0);
 	},
 
 	renderResult(
@@ -307,7 +321,9 @@ export const searchToolBm25Renderer = {
 		const safeQuery = replaceTabs(details.query);
 		const header = renderStatusLine(
 			{
-				icon: details.tools.length > 0 ? "success" : "warning",
+				...(details.tools.length > 0
+					? { iconOverride: uiTheme.fg("accent", uiTheme.symbol("icon.search")) }
+					: { icon: "warning" as const }),
 				title: TOOL_DISCOVERY_TITLE,
 				description: truncateToWidth(safeQuery, MATCH_LABEL_LEN),
 				meta,
@@ -320,19 +336,14 @@ export const searchToolBm25Renderer = {
 			return new Text(`${header}\n${uiTheme.fg("muted", emptyMessage)}`, 0, 0);
 		}
 
-		const lines = [header];
-		const treeLines = renderTreeList(
-			{
-				items: details.tools,
-				expanded: options.expanded,
-				maxCollapsed: COLLAPSED_MATCH_LIMIT,
-				itemType: "tool",
-				renderItem: match => renderMatchLines(match, uiTheme),
-			},
-			uiTheme,
-		);
-		lines.push(...treeLines);
-		return new Text(lines.join("\n"), 0, 0);
+		return framedBlock(uiTheme, width => ({
+			header,
+			sections: [{ lines: renderMatchBullets(details.tools, options.expanded ?? false, uiTheme) }],
+			state: "success",
+			borderColor: "borderMuted",
+			applyBg: false,
+			width,
+		}));
 	},
 
 	mergeCallAndResult: true,
diff --git a/packages/coding-agent/src/tools/search.ts b/packages/coding-agent/src/tools/search.ts
index 9cc55f67c..0215ff88f 100644
--- a/packages/coding-agent/src/tools/search.ts
+++ b/packages/coding-agent/src/tools/search.ts
@@ -19,6 +19,8 @@ import { DEFAULT_MAX_COLUMN, type TruncationResult, truncateHead, truncateLine }
 import {
 	Ellipsis,
 	fileHyperlink,
+	getTreeBranch,
+	getTreeContinuePrefix,
 	renderStatusLine,
 	renderTreeList,
 	truncateToWidth,
@@ -34,9 +36,9 @@ import {
 	parseArchivePathCandidates,
 } from "./archive-reader";
 import { createFileRecorder, formatResultPath } from "./file-recorder";
-import { formatGroupedFiles } from "./grouped-file-output";
+import { classifyGroupedLines, formatGroupedFiles, groupLineIndicesByBlank } from "./grouped-file-output";
 import { formatMatchLine } from "./match-line-format";
-import { formatFullOutputReference, type OutputMeta } from "./output-meta";
+import type { OutputMeta } from "./output-meta";
 import {
 	expandDelimitedPathEntries,
 	hasGlobPathChars,
@@ -56,8 +58,9 @@ import {
 	formatCount,
 	formatEmptyMessage,
 	formatErrorMessage,
+	formatMoreItems,
 	PREVIEW_LIMITS,
-	splitGroupsByBlankLine,
+	replaceTabs,
 } from "./render-utils";
 import { ToolError } from "./tool-errors";
 import { toolResult } from "./tool-result";
@@ -283,7 +286,6 @@ interface IndexedContentLines {
 	starts: number[];
 }
 
-const INTERNAL_URL_DISPLAY_RE = /^[a-z][a-z0-9+.-]*:\/\//i;
 const OMP_ROOT_URL_RE = /^omp:\/\/(?:\/?|docs\/?)$/i;
 
 function normalizeSearchLine(line: string): string {
@@ -619,6 +621,10 @@ export interface SearchToolDetails {
 	/** Absolute base directory used during search. Used by the renderer to resolve
 	 * display-relative paths to absolute paths for OSC 8 hyperlinks. */
 	searchPath?: string;
+	/** Session cwd at search time. The renderer resolves the display-relative
+	 * (cwd-relative) header/match paths against this for OSC 8 hyperlinks;
+	 * `searchPath` is the scope label target, not the display-path base. */
+	cwd?: string;
 	/** User-supplied paths whose base directory was missing on disk. The tool
 	 * skipped these and continued with the surviving entries; surfaced as a
 	 * non-fatal warning in the renderer and in the model-facing text. */
@@ -1001,6 +1007,7 @@ export class SearchTool implements AgentTool<typeof searchSchema, SearchToolDeta
 					const details: SearchToolDetails = {
 						scopePath,
 						searchPath,
+						cwd: this.session.cwd,
 						matchCount: 0,
 						fileCount: 0,
 						files: [],
@@ -1127,6 +1134,7 @@ export class SearchTool implements AgentTool<typeof searchSchema, SearchToolDeta
 				const details: SearchToolDetails = {
 					scopePath,
 					searchPath,
+					cwd: this.session.cwd,
 					matchCount: selectedMatches.length,
 					fileCount: fileList.length,
 					files: fileList,
@@ -1169,6 +1177,10 @@ interface SearchRenderArgs {
 }
 
 const COLLAPSED_TEXT_LIMIT = PREVIEW_LIMITS.COLLAPSED_LINES * 2;
+/** Line budget for the expanded view. Larger than collapsed so expanding
+ * reveals more matches with context, but still bounded so a single hot file
+ * whose matches span the whole file can't dump its entire length. */
+const EXPANDED_TEXT_LIMIT = PREVIEW_LIMITS.EXPANDED_LINES * 2;
 
 const SEARCH_CODE_FRAME_LINE_RE = /^\s*\*?(\d+)│/;
 
@@ -1190,6 +1202,140 @@ function parseSearchDisplayLineNumber(line: string): number | undefined {
 	return Number.parseInt(match[1]!, 10);
 }
 
+const SEARCH_MATCH_LINE_RE = /^\s*\*\d+(?:│|[:|])/;
+
+interface RenderedSearchLine {
+	raw: string;
+	styled: string;
+}
+
+function isSearchMatchLine(line: string): boolean {
+	return SEARCH_MATCH_LINE_RE.test(line);
+}
+
+function isSearchHeaderLine(line: string): boolean {
+	return /^#+ /.test(line);
+}
+
+const URL_HEADER_PREFIX_RE = /^#+\s+/;
+
+function renderSearchDisplayLines(
+	lines: readonly string[],
+	headerBase: string | undefined,
+	fileScope: string | undefined,
+	uiTheme: Theme,
+): RenderedSearchLine[] {
+	const contexts = classifyGroupedLines(lines, headerBase, fileScope);
+	// `classifyGroupedLines` can't resolve internal URLs (TUI-only), so track the
+	// resolved URL target here and use it for the body lines that follow.
+	let urlFile: string | undefined;
+	return lines.map((line, index) => {
+		const ctx = contexts[index]!;
+		if (ctx.kind === "dir") {
+			urlFile = undefined;
+			const styled = uiTheme.fg("accent", line);
+			return { raw: line, styled: ctx.headerPath ? fileHyperlink(ctx.headerPath, styled) : styled };
+		}
+		if (ctx.kind === "file") {
+			if (ctx.isUrl) {
+				const raw = line
+					.replace(URL_HEADER_PREFIX_RE, "")
+					.trimEnd()
+					.replace(/\s+\([^)]*\)\s*$/, "");
+				const linked = linkUrlLikeSearchHeader(raw, uiTheme.fg("accent", line));
+				urlFile = linked.absPath;
+				return { raw: line, styled: linked.line };
+			}
+			urlFile = undefined;
+			// Root-level files keep the bright accent; nested file headers are dimmed.
+			const styled = uiTheme.fg(ctx.depth === 1 ? "accent" : "dim", line);
+			return { raw: line, styled: ctx.headerPath ? fileHyperlink(ctx.headerPath, styled) : styled };
+		}
+		const styled = uiTheme.fg("toolOutput", line);
+		const lineNumber = parseSearchDisplayLineNumber(line);
+		const filePath = ctx.filePath ?? urlFile;
+		return {
+			raw: line,
+			styled: filePath && lineNumber !== undefined ? fileHyperlink(filePath, styled, { line: lineNumber }) : styled,
+		};
+	});
+}
+
+function compactSearchPreviewGroup(group: RenderedSearchLine[]): RenderedSearchLine[] {
+	const compact = group.filter(line => isSearchHeaderLine(line.raw) || isSearchMatchLine(line.raw));
+	return compact.length > 0 ? compact : group;
+}
+
+function countPreviewMatches(lines: readonly RenderedSearchLine[], hasMarkedMatches: boolean): number {
+	if (hasMarkedMatches) return lines.reduce((count, line) => count + (isSearchMatchLine(line.raw) ? 1 : 0), 0);
+	return lines.reduce((count, line) => count + (!isSearchHeaderLine(line.raw) && line.raw.length > 0 ? 1 : 0), 0);
+}
+
+function renderBudgetedSearchGroups(
+	groups: RenderedSearchLine[][],
+	maxLines: number,
+	matchCount: number,
+	uiTheme: Theme,
+	compact: boolean,
+): string[] {
+	if (maxLines <= 0) return [];
+	const renderedGroups = groups
+		.map(group => (compact ? compactSearchPreviewGroup(group) : group))
+		.filter(group => group.length > 0);
+	if (renderedGroups.length === 0) return [];
+
+	let totalLines = 0;
+	let totalMarkedMatches = 0;
+	let totalFallbackMatches = 0;
+	for (const group of renderedGroups) {
+		totalLines += group.length;
+		totalMarkedMatches += countPreviewMatches(group, true);
+		totalFallbackMatches += countPreviewMatches(group, false);
+	}
+	const hasMarkedMatches = totalMarkedMatches > 0;
+	const needsSummary = totalLines > maxLines;
+	const contentBudget = needsSummary ? Math.max(maxLines - 1, 0) : maxLines;
+	const visibleGroups: RenderedSearchLine[][] = [];
+	let visibleLineCount = 0;
+	let visibleMatches = 0;
+	for (const group of renderedGroups) {
+		if (visibleLineCount >= contentBudget) break;
+		const available = contentBudget - visibleLineCount;
+		const take = Math.min(group.length, available);
+		if (take <= 0) break;
+		const visibleGroup = group.slice(0, take);
+		visibleGroups.push(visibleGroup);
+		visibleLineCount += visibleGroup.length;
+		visibleMatches += countPreviewMatches(visibleGroup, hasMarkedMatches);
+	}
+
+	const totalMatches = hasMarkedMatches ? totalMarkedMatches : Math.max(matchCount, totalFallbackMatches);
+	const hiddenMatches = Math.max(totalMatches - visibleMatches, 0);
+	const hiddenLines = Math.max(totalLines - visibleLineCount, 0);
+	const hasSummary = needsSummary && (hiddenMatches > 0 || hiddenLines > 0);
+	const lines: string[] = [];
+	for (let i = 0; i < visibleGroups.length; i++) {
+		const group = visibleGroups[i]!;
+		const isLast = !hasSummary && i === visibleGroups.length - 1;
+		const prefix = `${uiTheme.fg("dim", getTreeBranch(isLast, uiTheme))} `;
+		const continuePrefix = uiTheme.fg("dim", getTreeContinuePrefix(isLast, uiTheme));
+		lines.push(`${prefix}${replaceTabs(group[0]!.styled)}`);
+		for (let j = 1; j < group.length; j++) {
+			lines.push(`${continuePrefix}${replaceTabs(group[j]!.styled)}`);
+		}
+	}
+	if (hasSummary) {
+		const hiddenLabel =
+			hiddenMatches > 0 ? formatMoreItems(hiddenMatches, "match") : formatMoreItems(hiddenLines, "line");
+		lines.push(`${uiTheme.fg("dim", uiTheme.tree.last)} ${uiTheme.fg("muted", hiddenLabel)}`);
+	}
+	return lines;
+}
+
+function searchStatusIcon(uiTheme: Theme): string {
+	return uiTheme.fg("toolTitle", uiTheme.symbol("icon.search"));
+}
+
 export const searchToolRenderer = {
 	inline: true,
 	renderCall(args: SearchRenderArgs, _options: RenderResultOptions, uiTheme: Theme): Component {
@@ -1201,10 +1347,10 @@ export const searchToolRenderer = {
 		if (args.skip !== undefined && args.skip > 0) meta.push(`skip:${args.skip}`);
 
 		const text = renderStatusLine(
-			{ icon: "pending", title: "Search", description: args.pattern || "?", meta },
+			{ icon: "pending", title: "Search", titleColor: "toolTitle", description: args.pattern || "?", meta },
 			uiTheme,
 		);
-		return new Text(text, 0, 0);
+		return new Text(text, 1, 0);
 	},
 
 	renderResult(
@@ -1217,7 +1363,7 @@ export const searchToolRenderer = {
 
 		if (result.isError || details?.error) {
 			const errorText = details?.error || result.content?.find(c => c.type === "text")?.text || "Unknown error";
-			return new Text(formatErrorMessage(errorText, uiTheme), 0, 0);
+			return new Text(formatErrorMessage(errorText, uiTheme), 1, 0);
 		}
 
 		const hasDetailedData = details?.matchCount !== undefined || details?.fileCount !== undefined;
@@ -1225,12 +1371,18 @@ export const searchToolRenderer = {
 		if (!hasDetailedData) {
 			const textContent = result.details?.displayContent ?? result.content?.find(c => c.type === "text")?.text;
 			if (!textContent || textContent === "No matches found") {
-				return new Text(formatEmptyMessage("No matches found", uiTheme), 0, 0);
+				return new Text(formatEmptyMessage("No matches found", uiTheme), 1, 0);
 			}
 			const lines = textContent.split("\n").filter(line => line.trim() !== "");
 			const description = args?.pattern ?? undefined;
 			const header = renderStatusLine(
-				{ icon: "success", title: "Search", description, meta: [formatCount("item", lines.length)] },
+				{
+					iconOverride: searchStatusIcon(uiTheme),
+					title: "Search",
+					titleColor: "toolTitle",
+					description,
+					meta: [formatCount("item", lines.length)],
+				},
 				uiTheme,
 			);
 			return createCachedComponent(
@@ -1249,6 +1401,7 @@ export const searchToolRenderer = {
 					);
 					return [header, ...listLines].map(l => truncateToWidth(l, width, Ellipsis.Omit));
 				},
+				{ paddingX: 1 },
 			);
 		}
 
@@ -1269,12 +1422,12 @@ export const searchToolRenderer = {
 			const scopeMeta = searchScopeMeta(details);
 			if (scopeMeta) meta.push(scopeMeta);
 			const header = renderStatusLine(
-				{ icon: "warning", title: "Search", description: args?.pattern, meta },
+				{ icon: "warning", title: "Search", titleColor: "toolTitle", description: args?.pattern, meta },
 				uiTheme,
 			);
 			const lines = [header, formatEmptyMessage("No matches found", uiTheme)];
 			if (missingNote) lines.push(missingNote);
-			return new Text(lines.join("\n"), 0, 0);
+			return new Text(lines.join("\n"), 1, 0);
 		}
 
 		const summaryParts = [formatCount("match", matchCount), formatCount("file", fileCount)];
@@ -1284,104 +1437,45 @@ export const searchToolRenderer = {
 		if (truncated) meta.push(uiTheme.fg("warning", "truncated"));
 		const description = args?.pattern ?? undefined;
 		const header = renderStatusLine(
-			{ icon: truncated ? "warning" : "success", title: "Search", description, meta },
+			{
+				...(truncated ? { icon: "warning" as const } : { iconOverride: searchStatusIcon(uiTheme) }),
+				title: "Search",
+				titleColor: "toolTitle",
+				description,
+				meta,
+			},
 			uiTheme,
 		);
 
 		const textContent = result.details?.displayContent ?? result.content?.find(c => c.type === "text")?.text ?? "";
-		const matchGroups = splitGroupsByBlankLine(textContent.split("\n"));
-
-		const renderedFileLimit = details?.fileLimitReached;
-		const renderedPerFileLimit = details?.perFileLimitReached;
-		const truncationReasons: string[] = [];
-		if (renderedFileLimit) truncationReasons.push(`first ${renderedFileLimit} files (skip to paginate)`);
-		if (renderedPerFileLimit) truncationReasons.push(`first ${renderedPerFileLimit} matches per file`);
-		if (truncation) truncationReasons.push(truncation.truncatedBy === "lines" ? "line limit" : "size limit");
-		if (limits?.columnTruncated) truncationReasons.push(`line length ${limits.columnTruncated.maxColumn}`);
-		if (truncation?.artifactId) truncationReasons.push(formatFullOutputReference(truncation.artifactId));
+		const allLines = textContent.split("\n");
+		// Resolve hyperlinks once over the whole output so a nested directory stack
+		// reconstructs correctly across blank-line group boundaries.
+		// Header/match display paths are cwd-relative, so resolve them against cwd
+		// (falling back to searchPath for legacy results that predate `cwd`); the
+		// scoped file's absolute path seeds body lines in single-file searches.
+		const renderedLines = renderSearchDisplayLines(
+			allLines,
+			details?.cwd ?? details?.searchPath,
+			details?.searchPath,
+			uiTheme,
+		);
+		const matchGroups = groupLineIndicesByBlank(allLines).map(indices => indices.map(i => renderedLines[i]!));
 
 		const extraLines: string[] = [];
-		if (truncationReasons.length > 0) {
-			extraLines.push(uiTheme.fg("warning", `truncated: ${truncationReasons.join(", ")}`));
-		}
 		if (missingNote) extraLines.push(missingNote);
 
 		return createCachedComponent(
 			() => options.expanded,
 			width => {
-				const collapsedMatchLineBudget = Math.max(COLLAPSED_TEXT_LIMIT - extraLines.length, 0);
-				const searchBase = details?.searchPath;
-				const matchLines = renderTreeList(
-					{
-						items: matchGroups,
-						expanded: options.expanded,
-						maxCollapsed: matchGroups.length,
-						maxCollapsedLines: collapsedMatchLineBudget,
-						itemType: "match",
-						renderItem: group => {
-							// Track directory/file context within a group so headers and code-frame
-							// lines link to the backing file, with line-specific links for matches.
-							let contextDir = searchBase ?? "";
-							const hasFileHeader = group.some(line => line.startsWith("# "));
-							let currentFilePath: string | undefined = hasFileHeader ? undefined : searchBase;
-							return group.map(line => {
-								if (line.startsWith("## ")) {
-									// Strip optional ` (suffix)` and `#hash` before resolving.
-									const fileName = line
-										.slice(3)
-										.trimEnd()
-										.replace(/\s+\([^)]*\)\s*$/, "")
-										.replace(/#[0-9a-f]+$/, "");
-									const absPath = contextDir && fileName ? path.join(contextDir, fileName) : undefined;
-									currentFilePath = absPath;
-									const styled = uiTheme.fg("dim", line);
-									return absPath ? fileHyperlink(absPath, styled) : styled;
-								}
-								if (line.startsWith("# ")) {
-									const raw = line
-										.slice(2)
-										.trimEnd()
-										.replace(/\s+\([^)]*\)\s*$/, "");
-									if (INTERNAL_URL_DISPLAY_RE.test(raw)) {
-										contextDir = "";
-										const styled = uiTheme.fg("accent", line);
-										const linked = linkUrlLikeSearchHeader(raw, styled);
-										currentFilePath = linked.absPath;
-										return linked.line;
-									}
-									const isDirectory = raw.endsWith("/");
-									const name = isDirectory ? raw.replace(/\/$/, "") : raw.replace(/#[0-9a-f]+$/, "");
-									if (isDirectory) {
-										const absPath = searchBase
-											? name === "."
-												? searchBase
-												: path.join(searchBase, name)
-											: undefined;
-										if (absPath) {
-											contextDir = absPath;
-										}
-										currentFilePath = undefined;
-										const styled = uiTheme.fg("accent", line);
-										return absPath ? fileHyperlink(absPath, styled) : styled;
-									}
-									// Root-level file emitted by formatGroupedFiles when the directory is `.`.
-									const absPath = searchBase && name ? path.join(searchBase, name) : undefined;
-									currentFilePath = absPath;
-									const styled = uiTheme.fg("accent", line);
-									return absPath ? fileHyperlink(absPath, styled) : styled;
-								}
-								const styled = uiTheme.fg("toolOutput", line);
-								const lineNumber = parseSearchDisplayLineNumber(line);
-								return currentFilePath && lineNumber !== undefined
-									? fileHyperlink(currentFilePath, styled, { line: lineNumber })
-									: styled;
-							});
-						},
-					},
-					uiTheme,
+				const budget = Math.max(
+					(options.expanded ? EXPANDED_TEXT_LIMIT : COLLAPSED_TEXT_LIMIT) - extraLines.length,
+					0,
 				);
+				const matchLines = renderBudgetedSearchGroups(matchGroups, budget, matchCount, uiTheme, !options.expanded);
 				return [header, ...matchLines, ...extraLines].map(l => truncateToWidth(l, width, Ellipsis.Omit));
 			},
+			{ paddingX: 1 },
 		);
 	},
 	mergeCallAndResult: true,
diff --git a/packages/coding-agent/src/tools/sqlite-reader.ts b/packages/coding-agent/src/tools/sqlite-reader.ts
index 37948a48d..710c7fa66 100644
--- a/packages/coding-agent/src/tools/sqlite-reader.ts
+++ b/packages/coding-agent/src/tools/sqlite-reader.ts
@@ -5,6 +5,14 @@ import { ToolError } from "./tool-errors";
 const SQLITE_MAGIC = new Uint8Array([
 	0x53, 0x51, 0x4c, 0x69, 0x74, 0x65, 0x20, 0x66, 0x6f, 0x72, 0x6d, 0x61, 0x74, 0x20, 0x33, 0x00,
 ]);
+
+export function looksLikeSqlite(bytes: Uint8Array): boolean {
+	if (bytes.byteLength < SQLITE_MAGIC.byteLength) return false;
+	for (const [index, byte] of SQLITE_MAGIC.entries()) {
+		if (bytes[index] !== byte) return false;
+	}
+	return true;
+}
 const SQLITE_PATH_PATTERN = /\.(?:sqlite3?|db3?)(?=(?::|\?|$))/gi;
 const DEFAULT_QUERY_LIMIT = 20;
 const DEFAULT_SCHEMA_SAMPLE_LIMIT = 5;
@@ -443,18 +451,7 @@ export function parseSqlitePathCandidates(filePath: string): SqlitePathCandidate
 
 export async function isSqliteFile(absolutePath: string): Promise<boolean> {
 	try {
-		const bytes = await Bun.file(absolutePath).slice(0, SQLITE_MAGIC.byteLength).bytes();
-		if (bytes.length !== SQLITE_MAGIC.byteLength) {
-			return false;
-		}
-
-		for (const [index, byte] of SQLITE_MAGIC.entries()) {
-			if (bytes[index] !== byte) {
-				return false;
-			}
-		}
-
-		return true;
+		return looksLikeSqlite(await Bun.file(absolutePath).slice(0, SQLITE_MAGIC.byteLength).bytes());
 	} catch {
 		return false;
 	}
diff --git a/packages/coding-agent/src/tools/ssh.ts b/packages/coding-agent/src/tools/ssh.ts
index 9f821beb1..553f8e0f0 100644
--- a/packages/coding-agent/src/tools/ssh.ts
+++ b/packages/coding-agent/src/tools/ssh.ts
@@ -13,11 +13,11 @@ import type { SSHHostInfo } from "../ssh/connection-manager";
 import { ensureHostInfo, getHostInfoForHost } from "../ssh/connection-manager";
 import { executeSSH } from "../ssh/ssh-executor";
 import { renderStatusLine } from "../tui";
-import { CachedOutputBlock } from "../tui/output-block";
+import { CachedOutputBlock, markFramedBlockComponent } from "../tui/output-block";
 import type { ToolSession } from ".";
 import { truncateForPrompt } from "./approval";
 import { formatStyledTruncationWarning, type OutputMeta, stripOutputNotice } from "./output-meta";
-import { replaceTabs } from "./render-utils";
+import { capPreviewLines, replaceTabs } from "./render-utils";
 import { ToolError } from "./tool-errors";
 import { toolResult } from "./tool-result";
 import { clampTimeout } from "./tool-timeouts";
@@ -244,16 +244,22 @@ export const sshToolRenderer = {
 		const header = renderStatusLine({ icon: "pending", title: "SSH", description: `[${host}]` }, uiTheme);
 		const cmdLines = formatSshCommandLines(command, uiTheme);
 		const outputBlock = new CachedOutputBlock();
-		return {
+		return markFramedBlockComponent({
 			render: (width: number): string[] =>
 				outputBlock.render(
-					{ header, state: "pending", sections: [{ lines: cmdLines }], width, animate: true },
+					{
+						header,
+						state: "pending",
+						sections: [{ lines: capPreviewLines(cmdLines, uiTheme, { expanded: _options.expanded }) }],
+						width,
+						animate: true,
+					},
 					uiTheme,
 				),
 			invalidate: () => {
 				outputBlock.invalidate();
 			},
-		};
+		});
 	},
 
 	renderResult(
@@ -273,7 +279,7 @@ export const sshToolRenderer = {
 		const textContent = result.content?.find(c => c.type === "text")?.text ?? "";
 		const outputBlock = new CachedOutputBlock();
 
-		return {
+		return markFramedBlockComponent({
 			render: (width: number): string[] => {
 				// REACTIVE: read mutable options at render time
 				const { expanded, renderContext } = options;
@@ -319,7 +325,14 @@ export const sshToolRenderer = {
 					{
 						header,
 						state: "success",
-						sections: [{ lines: cmdLines }, { label: uiTheme.fg("toolTitle", "Output"), lines: outputLines }],
+						sections: [
+							{
+								lines: options.isPartial
+									? capPreviewLines(cmdLines, uiTheme, { expanded: options.expanded })
+									: cmdLines,
+							},
+							{ label: uiTheme.fg("toolTitle", "Output"), lines: outputLines },
+						],
 						width,
 					},
 					uiTheme,
@@ -328,7 +341,7 @@ export const sshToolRenderer = {
 			invalidate: () => {
 				outputBlock.invalidate();
 			},
-		};
+		});
 	},
 	mergeCallAndResult: true,
 };
diff --git a/packages/coding-agent/src/tools/todo.ts b/packages/coding-agent/src/tools/todo.ts
index 6d7c67ab0..174223cc3 100644
--- a/packages/coding-agent/src/tools/todo.ts
+++ b/packages/coding-agent/src/tools/todo.ts
@@ -9,8 +9,8 @@ import type { Theme } from "../modes/theme/theme";
 import todoDescription from "../prompts/tools/todo.md" with { type: "text" };
 import type { ToolSession } from "../sdk";
 import type { SessionEntry } from "../session/session-manager";
-import { renderStatusLine, renderTreeList } from "../tui";
-import { PREVIEW_LIMITS } from "./render-utils";
+import { framedBlock, renderStatusLine, renderTreeList } from "../tui";
+import { formatErrorDetail, PREVIEW_LIMITS } from "./render-utils";
 
 // =============================================================================
 // Types
@@ -755,19 +755,17 @@ function formatTodoLine(
 	}
 }
 
-function renderNoteAttachments(phases: TodoPhase[], uiTheme: Theme): string[] {
+function renderNoteAttachments(phases: TodoPhase[], uiTheme: Theme, indent: string): string[] {
 	const lines: string[] = [];
 	for (const phase of phases) {
 		for (const task of phase.tasks) {
 			if (task.status !== "in_progress" || !task.notes || task.notes.length === 0) continue;
-			const bar = uiTheme.fg("dim", uiTheme.tree.vertical);
-			const title = uiTheme.fg("dim", chalk.italic(`§ notes — ${task.content}`));
 			lines.push("");
-			lines.push(`  ${title}`);
+			lines.push(`${indent}${uiTheme.fg("dim", chalk.italic(`§ notes — ${task.content}`))}`);
 			for (let j = 0; j < task.notes.length; j++) {
-				if (j > 0) lines.push(`  ${bar}`);
+				if (j > 0) lines.push("");
 				for (const noteLine of task.notes[j].split("\n")) {
-					lines.push(`  ${bar} ${uiTheme.fg("dim", noteLine)}`);
+					lines.push(`${indent}  ${uiTheme.fg("dim", noteLine)}`);
 				}
 			}
 		}
@@ -775,25 +773,102 @@ function renderNoteAttachments(phases: TodoPhase[], uiTheme: Theme): string[] {
 	return lines;
 }
 
+/**
+ * Phases the latest update touched, plus the active (in_progress) phase.
+ * Returns `null` when there is no usable signal, meaning "render every phase
+ * fully" — this preserves the legacy view and the manual-expand path.
+ */
+function computeTouchedPhases(
+	args: TodoRenderArgs | undefined,
+	phases: TodoPhase[],
+	completedTasks: TodoCompletionTransition[],
+): Set<string> | null {
+	const touched = new Set<string>();
+	// The phase holding the in_progress task is where attention sits after the
+	// auto-promotion that follows every completion.
+	for (const phase of phases) {
+		if (phase.tasks.some(task => task.status === "in_progress")) touched.add(phase.name);
+	}
+	// Phases with a task that just transitioned to completed in this update.
+	for (const transition of completedTasks) touched.add(transition.phase);
+	// Phases explicitly named by the ops that ran. `init` replaces the whole
+	// list, so the entire plan is fresh and every phase counts as touched.
+	const ops = Array.isArray(args?.ops) ? args.ops : [];
+	for (const op of ops) {
+		if (!op || typeof op !== "object") continue;
+		if (op.op === "init") {
+			for (const phase of phases) touched.add(phase.name);
+			break;
+		}
+		if (typeof op.phase === "string" && op.phase) {
+			const named = phases.find(phase => phase.name === op.phase);
+			if (named) touched.add(named.name);
+		}
+		if (typeof op.task === "string" && op.task) {
+			const located = findTaskByContent(phases, op.task);
+			if (located) touched.add(located.phase.name);
+		}
+	}
+	return touched.size > 0 ? touched : null;
+}
+
+/** One-line summary for a collapsed (untouched) phase: dim header + progress. */
+function formatPhaseSummary(phase: TodoPhase, oneBasedIndex: number, uiTheme: Theme): string {
+	const total = phase.tasks.length;
+	const done = phase.tasks.filter(task => task.status === "completed").length;
+	const name = uiTheme.fg("dim", chalk.bold(formatPhaseDisplayName(phase.name, oneBasedIndex)));
+	return `${name}${uiTheme.fg("dim", `  ${done}/${total}`)}`;
+}
+
 export const todoToolRenderer = {
-	renderCall(args: TodoRenderArgs, _options: RenderResultOptions, uiTheme: Theme): Component {
-		const ops = args?.ops?.map(entry => {
-			const parts = [entry.op ?? "update"];
-			if (entry.task) parts.push(entry.task);
-			if (entry.phase) parts.push(entry.phase);
-			if (entry.items?.length) parts.push(`${entry.items.length} item${entry.items.length === 1 ? "" : "s"}`);
-			return parts.join(" ");
-		}) ?? ["update"];
-		const text = renderStatusLine({ icon: "pending", title: "Todo", meta: ops }, uiTheme);
-		return new Text(text, 0, 0);
+	renderCall(args: TodoRenderArgs, options: RenderResultOptions, uiTheme: Theme): Component {
+		// `args` here is the raw partially-parsed JSON from the streaming
+		// tool-call delta and may not satisfy `TodoRenderArgs` at runtime:
+		// `parseStreamingJson` can hand back `{ ops: "[" }` mid-delta, or
+		// entries that are `null` / strings before fields stream. Guard
+		// against non-array `ops` and non-object entries so a malformed
+		// delta never breaks the TUI render loop (#2005).
+		const opsList = Array.isArray(args?.ops) ? args.ops : [];
+		const ops =
+			opsList.length === 0
+				? ["update"]
+				: opsList.map(entry => {
+						const e = entry && typeof entry === "object" ? entry : ({} as NonNullable<typeof entry>);
+						const parts = [e.op ?? "update"];
+						if (e.task) parts.push(e.task);
+						if (e.phase) parts.push(e.phase);
+						if (Array.isArray(e.items) && e.items.length) {
+							parts.push(`${e.items.length} item${e.items.length === 1 ? "" : "s"}`);
+						}
+						return parts.join(" ");
+					});
+		// No body worth boxing while the call streams — a lone status line reads
+		// cleaner than an empty frame. The container renders it without chrome.
+		const header = renderStatusLine(
+			{ icon: "pending", spinnerFrame: options?.spinnerFrame, title: "Todo", meta: ops },
+			uiTheme,
+		);
+		return new Text(header, 0, 0);
 	},
 
 	renderResult(
-		result: { content: Array<{ type: string; text?: string }>; details?: TodoToolDetails },
+		result: { content: Array<{ type: string; text?: string }>; details?: TodoToolDetails; isError?: boolean },
 		options: RenderResultOptions,
 		uiTheme: Theme,
-		_args?: TodoRenderArgs,
+		args?: TodoRenderArgs,
 	): Component {
+		if (result.isError) {
+			const errorText = result.content?.find(content => content.type === "text")?.text ?? "Todo operation failed";
+			const header = renderStatusLine({ icon: "error", title: "Todo" }, uiTheme);
+			return framedBlock(uiTheme, width => ({
+				header,
+				sections: [{ lines: formatErrorDetail(errorText, uiTheme).split("\n") }],
+				state: "error",
+				borderColor: "error",
+				width,
+			}));
+		}
+
 		const phases = (result.details?.phases ?? []).filter(phase => phase.tasks.length > 0);
 		const completedTasks = result.details?.completedTasks ?? [];
 		const completionKeysByPhase = new Map<string, Set<string>>();
@@ -809,48 +884,52 @@ export const todoToolRenderer = {
 		const header = renderStatusLine({ icon: "success", title: "Todo", meta: [`${allTasks.length} tasks`] }, uiTheme);
 		if (allTasks.length === 0) {
 			const fallback = result.content?.find(content => content.type === "text")?.text ?? "No todos";
-			return new Text(`${header}\n${uiTheme.fg("dim", fallback)}`, 0, 0);
+			return new Text(`${header}\n  ${uiTheme.fg("dim", fallback)}`, 0, 0);
 		}
 
-		let cachedKey: string | undefined;
-		let cachedLines: string[] | undefined;
-		return {
-			invalidate(): void {
-				cachedKey = undefined;
-				cachedLines = undefined;
-			},
-			render(width: number): string[] {
-				const { expanded, spinnerFrame } = options;
-				const key = `${expanded ? 1 : 0}:${spinnerFrame ?? -1}:${width}`;
-				if (cachedKey === key && cachedLines) return cachedLines;
-
-				const lines: string[] = [header];
-				for (let p = 0; p < phases.length; p++) {
-					const phase = phases[p];
-					if (phases.length > 1) {
-						lines.push(uiTheme.fg("accent", chalk.bold(`  ${formatPhaseDisplayName(phase.name, p + 1)}`)));
-					}
-					const completionKeys = completionKeysByPhase.get(phase.name) ?? EMPTY_COMPLETION_KEYS;
-					const treeLines = renderTreeList(
-						{
-							items: phase.tasks,
-							expanded,
-							maxCollapsed: PREVIEW_LIMITS.COLLAPSED_ITEMS,
-							itemType: "todo",
-							renderItem: todo => formatTodoLine(todo, uiTheme, "", completionKeys, spinnerFrame),
-						},
-						uiTheme,
-					);
-					for (const line of treeLines) {
-						lines.push(`  ${line}`);
-					}
+		return framedBlock(uiTheme, width => {
+			const { expanded, spinnerFrame } = options;
+			const multiPhase = phases.length > 1;
+			const indent = multiPhase ? "  " : "";
+			// Collapse phases this update didn't touch down to a one-line summary so
+			// a single task flip doesn't redraw every phase's full task list. The
+			// manual expand toggle (and the no-signal fallback) still shows all.
+			const touched = expanded || !multiPhase ? null : computeTouchedPhases(args, phases, completedTasks);
+			const bodyLines: string[] = [];
+			for (let p = 0; p < phases.length; p++) {
+				const phase = phases[p];
+				if (touched && !touched.has(phase.name)) {
+					bodyLines.push(formatPhaseSummary(phase, p + 1, uiTheme));
+					continue;
 				}
-				lines.push(...renderNoteAttachments(phases, uiTheme));
-				cachedKey = key;
-				cachedLines = lines;
-				return lines;
-			},
-		};
+				if (multiPhase) {
+					bodyLines.push(uiTheme.fg("accent", chalk.bold(formatPhaseDisplayName(phase.name, p + 1))));
+				}
+				const completionKeys = completionKeysByPhase.get(phase.name) ?? EMPTY_COMPLETION_KEYS;
+				const treeLines = renderTreeList(
+					{
+						items: phase.tasks,
+						expanded,
+						maxCollapsed: PREVIEW_LIMITS.COLLAPSED_ITEMS,
+						itemType: "todo",
+						renderItem: todo => formatTodoLine(todo, uiTheme, "", completionKeys, spinnerFrame),
+					},
+					uiTheme,
+				);
+				for (const line of treeLines) {
+					bodyLines.push(`${indent}${line}`);
+				}
+			}
+			bodyLines.push(...renderNoteAttachments(phases, uiTheme, indent));
+			while (bodyLines.length > 0 && bodyLines[0].trim() === "") bodyLines.shift();
+			return {
+				header,
+				sections: bodyLines.length > 0 ? [{ lines: bodyLines }] : [],
+				state: options.isPartial ? "pending" : "success",
+				borderColor: "borderMuted",
+				width,
+			};
+		});
 	},
 	mergeCallAndResult: true,
 };
diff --git a/packages/coding-agent/src/tools/write.ts b/packages/coding-agent/src/tools/write.ts
index 409030bad..2a401d8af 100644
--- a/packages/coding-agent/src/tools/write.ts
+++ b/packages/coding-agent/src/tools/write.ts
@@ -5,7 +5,6 @@ import * as path from "node:path";
 import { formatHashlineHeader, stripHashlinePrefixes } from "@oh-my-pi/hashline";
 import type { AgentTool, AgentToolContext, AgentToolResult, AgentToolUpdateCallback } from "@oh-my-pi/pi-agent-core";
 import type { Component } from "@oh-my-pi/pi-tui";
-import { Text } from "@oh-my-pi/pi-tui";
 import { isEnoent, isRecord, prompt, untilAborted } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
 
@@ -19,7 +18,7 @@ import { getDiagnosticsLedger } from "../lsp/diagnostics-ledger";
 import { getLanguageFromPath, highlightCode, type Theme } from "../modes/theme/theme";
 import writeDescription from "../prompts/tools/write.md" with { type: "text" };
 import type { ToolSession } from "../sdk";
-import { Ellipsis, Hasher, type RenderCache, renderStatusLine, truncateToWidth } from "../tui";
+import { fileHyperlink, framedBlock, renderStatusLine } from "../tui";
 import { resolveFileDisplayMode } from "../utils/file-display-mode";
 import { truncateForPrompt } from "./approval";
 import { parseArchivePathCandidates } from "./archive-reader";
@@ -37,10 +36,9 @@ import { formatPathRelativeToCwd, isInternalUrlPath } from "./path-utils";
 import { enforcePlanModeWrite, resolvePlanPath } from "./plan-mode-guard";
 import {
 	formatDiagnostics,
+	formatErrorDetail,
 	formatExpandHint,
 	formatMoreItems,
-	formatStatusIcon,
-	formatTitle,
 	getLspBatchRequest,
 	replaceTabs,
 	shortenPath,
@@ -58,7 +56,7 @@ import {
 import { ToolError } from "./tool-errors";
 import { toolResult } from "./tool-result";
 
-const LOOSE_HASHLINE_HEADER_RE = /^\s*¶\S+#[^ \t\r\n]*\s*$/;
+const LOOSE_HASHLINE_HEADER_RE = /^\s*\[[^#\r\n]+#[^ \t\r\n]*\]\s*$/;
 
 let fflateModulePromise: Promise<typeof import("fflate")> | undefined;
 async function loadFflate(): Promise<typeof import("fflate")> {
@@ -79,6 +77,9 @@ export interface WriteToolDetails {
 	meta?: OutputMeta;
 	/** Set when the file was auto-chmod'd because content begins with a `#!` shebang. */
 	madeExecutable?: boolean;
+	/** Absolute filesystem path the write resolved to. Used by the renderer to wrap
+	 * the (possibly cwd-relative) header path in an OSC 8 `file://` hyperlink. */
+	resolvedPath?: string;
 }
 
 /**
@@ -109,7 +110,7 @@ function stripWriteContentWithPotentialLooseHeader(lines: string[]): { text: str
 /**
  * Strip hashline display prefixes from write content.
  *
- * Only active when hashline edit mode is enabled — the model sees `¶PATH#HASH`
+ * Only active when hashline edit mode is enabled — the model sees `[PATH#HASH]`
  * headers plus `LINE:` prefixes in read output and sometimes copies them into write content.
  */
 function stripWriteContent(session: ToolSession, content: string): { text: string; stripped: boolean } {
@@ -122,7 +123,7 @@ function stripWriteContent(session: ToolSession, content: string): { text: strin
 /**
  * Record a snapshot of the freshly-written `content` for `absolutePath`
  * so subsequent hashline edits address the new file with a current tag,
- * and return the matching `¶displayPath#TAG` header. Returns `undefined`
+ * and return the matching `[displayPath#TAG]` header. Returns `undefined`
  * when the session is not in hashline mode so callers can no-op cheaply.
  *
  * Mirrors the post-commit snapshot recording the hashline patcher performs
@@ -418,7 +419,7 @@ export class WriteTool implements AgentTool<typeof writeSchema, WriteToolDetails
 		}`;
 		return {
 			content: [{ type: "text", text: `Successfully wrote ${content.length} bytes to ${outputPath}` }],
-			details: {},
+			details: { resolvedPath: resolvedArchivePath.absolutePath },
 		};
 	}
 
@@ -534,7 +535,10 @@ export class WriteTool implements AgentTool<typeof writeSchema, WriteToolDetails
 			}
 
 			invalidateFsScanAfterWrite(resolvedSqlitePath.absolutePath);
-			return toolResult<WriteToolDetails>({}).text(resultText).sourcePath(resolvedSqlitePath.absolutePath).done();
+			return toolResult<WriteToolDetails>({ resolvedPath: resolvedSqlitePath.absolutePath })
+				.text(resultText)
+				.sourcePath(resolvedSqlitePath.absolutePath)
+				.done();
 		} catch (error) {
 			if (isEnoent(error)) {
 				throw new ToolError(`SQLite database '${displayPath}' not found`);
@@ -595,12 +599,13 @@ export class WriteTool implements AgentTool<typeof writeSchema, WriteToolDetails
 		if (!diagnostics) {
 			return {
 				content: [{ type: "text", text: resultText }],
-				details: {},
+				details: { resolvedPath: absolutePath },
 			};
 		}
 		return {
 			content: [{ type: "text", text: resultText }],
 			details: {
+				resolvedPath: absolutePath,
 				diagnostics,
 				meta: outputMeta()
 					.diagnostics(diagnostics.summary, diagnostics.messages ?? [])
@@ -770,7 +775,7 @@ export class WriteTool implements AgentTool<typeof writeSchema, WriteToolDetails
 		context?: AgentToolContext,
 	): Promise<AgentToolResult<WriteToolDetails>> {
 		return untilAborted(signal, async () => {
-			// Strip hashline display prefixes (¶PATH#HASH + LINE:) if the model copied them from read output
+			// Strip hashline display prefixes ([PATH#HASH] + LINE:) if the model copied them from read output
 			const { text: cleanContent, stripped } = stripWriteContent(this.session, content);
 			const internalRouter = InternalUrlRouter.instance();
 			if (internalRouter.canHandle(path)) {
@@ -873,7 +878,10 @@ export class WriteTool implements AgentTool<typeof writeSchema, WriteToolDetails
 				if (stripped) {
 					resultText += `\nNote: auto-stripped hashline display prefixes from content before writing.`;
 				}
-				return { content: [{ type: "text", text: resultText }], details: {} };
+				return {
+					content: [{ type: "text", text: resultText }],
+					details: { resolvedPath: absolutePath },
+				};
 			}
 
 			const diagnostics = await this.#writethrough(absolutePath, cleanContent, signal, undefined, batchRequest);
@@ -890,13 +898,14 @@ export class WriteTool implements AgentTool<typeof writeSchema, WriteToolDetails
 			if (!diagnostics) {
 				return {
 					content: [{ type: "text", text: resultText }],
-					details: { madeExecutable: madeExecutable || undefined },
+					details: { resolvedPath: absolutePath, madeExecutable: madeExecutable || undefined },
 				};
 			}
 
 			return {
 				content: [{ type: "text", text: resultText }],
 				details: {
+					resolvedPath: absolutePath,
 					diagnostics,
 					madeExecutable: madeExecutable || undefined,
 					meta: outputMeta()
@@ -935,11 +944,20 @@ function normalizeDisplayText(text: string): string {
 	return text.replace(/\r/g, "");
 }
 
-function formatStreamingContent(content: string, language: string | undefined, uiTheme: Theme): string {
+function formatStreamingContent(
+	content: string,
+	expanded: boolean,
+	language: string | undefined,
+	uiTheme: Theme,
+): string {
 	if (!content) return "";
 	const lines = normalizeDisplayText(content).split("\n");
 	const totalLines = lines.length;
-	const startIndex = Math.max(0, totalLines - WRITE_STREAMING_PREVIEW_LINES);
+	// Collapsed: follow the streaming edge with a bounded tail window so the box
+	// stays short enough not to strand its scrolled-off head above the viewport
+	// while the block is volatile. `Ctrl+O` (expanded) lifts the cap for a
+	// deliberate full view — matching the eval streaming preview.
+	const startIndex = expanded ? 0 : Math.max(0, totalLines - WRITE_STREAMING_PREVIEW_LINES);
 	const visibleLines = lines.slice(startIndex);
 	const hidden = startIndex;
 	const highlighted = highlightCode(visibleLines.join("\n"), language);
@@ -951,9 +969,9 @@ function formatStreamingContent(content: string, language: string | undefined, u
 	}
 	for (let i = 0; i < highlighted.length; i++) {
 		const lineNum = startIndex + i + 1;
-		const gutter = uiTheme.fg("dim", `${String(lineNum).padStart(lineNumberWidth, " ")}│`);
+		const gutter = uiTheme.fg("dim", `${String(lineNum).padStart(lineNumberWidth, " ")} `);
 		const body = replaceTabs(highlighted[i] ?? "");
-		text += ` ${gutter}${body}\n`;
+		text += `${gutter}${body}\n`;
 	}
 	text += uiTheme.fg("dim", `… (streaming)`);
 	return text;
@@ -977,9 +995,9 @@ function renderContentPreview(
 	let text = "\n\n";
 	for (let i = 0; i < highlighted.length; i++) {
 		const lineNum = i + 1;
-		const gutter = uiTheme.fg("dim", `${String(lineNum).padStart(lineNumberWidth, " ")}│`);
+		const gutter = uiTheme.fg("dim", `${String(lineNum).padStart(lineNumberWidth, " ")} `);
 		const body = replaceTabs(highlighted[i] ?? "");
-		text += ` ${gutter}${body}\n`;
+		text += `${gutter}${body}\n`;
 	}
 	if (!expanded && hidden > 0) {
 		const hint = formatExpandHint(uiTheme, expanded, hidden > 0);
@@ -996,23 +1014,44 @@ export const writeToolRenderer = {
 		const lang = getLanguageFromPath(rawPath) ?? "text";
 		const langIcon = uiTheme.fg("muted", uiTheme.getLangIcon(lang));
 		const pathDisplay = filePath ? uiTheme.fg("accent", filePath) : uiTheme.fg("toolOutput", "…");
-		const spinner =
-			options?.spinnerFrame !== undefined ? formatStatusIcon("running", uiTheme, options.spinnerFrame) : "";
+		const header = renderStatusLine(
+			{
+				icon: "pending",
+				spinnerFrame: options?.spinnerFrame,
+				title: "Write",
+				description: `${langIcon} ${pathDisplay}`,
+			},
+			uiTheme,
+		);
+		return framedBlock(uiTheme, width => {
+			const body = args.content
+				? formatStreamingContent(args.content, Boolean(options?.expanded), lang, uiTheme)
+				: "";
+			const bodyLines = body ? body.split("\n") : [];
+			while (bodyLines.length > 0 && bodyLines[0].trim() === "") bodyLines.shift();
+			return {
+				header,
+				sections: bodyLines.length > 0 ? [{ lines: bodyLines }] : [],
+				state: "pending",
+				borderColor: "borderMuted",
+				width,
+			};
+		});
+	},
 
-		let text = `${formatTitle("Write", uiTheme)} ${spinner ? `${spinner} ` : ""}${langIcon} ${pathDisplay}`;
-
-		if (!args.content) {
-			return new Text(text, 0, 0);
-		}
-
-		// Show streaming preview of content (tail)
-		text += formatStreamingContent(args.content, lang, uiTheme);
-
-		return new Text(text, 0, 0);
+	// Only the expanded (Ctrl+O) preview is append-only: it renders the whole
+	// content top-anchored, so streamed chunks only append rows at the bottom.
+	// The collapsed preview slides a bounded tail window (`formatStreamingContent`
+	// with `WRITE_STREAMING_PREVIEW_LINES`) whose visible rows re-layout as the
+	// window moves — not append-only, but it never overflows the viewport, so its
+	// head is never at risk of being dropped regardless. `write` has no partial
+	// result (content streams as args), so `result` is ignored here.
+	isStreamingPreviewAppendOnly(args: WriteRenderArgs, options: RenderResultOptions, _result?: unknown): boolean {
+		return Boolean(options?.expanded && args.content);
 	},
 
 	renderResult(
-		result: { content: Array<{ type: string; text?: string }>; details?: WriteToolDetails },
+		result: { content: Array<{ type: string; text?: string }>; details?: WriteToolDetails; isError?: boolean },
 		options: RenderResultOptions,
 		uiTheme: Theme,
 		args?: WriteRenderArgs,
@@ -1022,14 +1061,33 @@ export const writeToolRenderer = {
 		const fileContent = args?.content || "";
 		const lang = getLanguageFromPath(rawPath);
 		const langIcon = uiTheme.fg("muted", uiTheme.getLangIcon(lang));
-		const pathDisplay = filePath ? uiTheme.fg("accent", filePath) : uiTheme.fg("toolOutput", "…");
+		// The header shows the cwd-relative path but links to the absolute path the
+		// write resolved to (args.path may be relative, which would yield a broken
+		// `file://` URI). Falls back to plain text when the result lacks a path.
+		const linkTarget = result.details?.resolvedPath;
+		const styledPath = filePath ? uiTheme.fg("accent", filePath) : uiTheme.fg("toolOutput", "…");
+		const pathDisplay = filePath && linkTarget ? fileHyperlink(linkTarget, styledPath) : styledPath;
+
+		if (result.isError) {
+			const errorText = result.content?.find(c => c.type === "text")?.text ?? "";
+			const header = renderStatusLine(
+				{ icon: "error", title: "Write", description: `${langIcon} ${pathDisplay}` },
+				uiTheme,
+			);
+			return framedBlock(uiTheme, width => ({
+				header,
+				sections: [{ lines: formatErrorDetail(errorText, uiTheme).split("\n") }],
+				state: "error",
+				borderColor: "error",
+				width,
+			}));
+		}
+
 		const lineCount = countLines(fileContent);
 		const lineSuffix = formatLineCountSuffix(lineCount, uiTheme);
 		const execSuffix = result.details?.madeExecutable
 			? `${uiTheme.fg("dim", " · ")}${uiTheme.fg("success", "made executable!")}`
 			: "";
-
-		// Build header with status icon
 		const header = renderStatusLine(
 			{
 				icon: "success",
@@ -1040,38 +1098,29 @@ export const writeToolRenderer = {
 		);
 		const diagnostics = result.details?.diagnostics;
 
-		let cached: RenderCache | undefined;
-
-		return {
-			render(width: number) {
-				const { expanded } = options;
-				const key = new Hasher().bool(expanded).u32(width).digest();
-				if (cached?.key === key) return cached.lines;
-
-				let text = header;
-				text += renderContentPreview(fileContent, expanded, lang, uiTheme);
-
-				if (diagnostics) {
-					const diagText = formatDiagnostics(diagnostics, expanded, uiTheme, fp =>
-						uiTheme.getLangIcon(getLanguageFromPath(fp)),
-					);
-					if (diagText.trim()) {
-						const diagLines = diagText.split("\n");
-						const firstNonEmpty = diagLines.findIndex(line => line.trim());
-						if (firstNonEmpty >= 0) {
-							text += `\n${diagLines.slice(firstNonEmpty).join("\n")}`;
-						}
-					}
+		return framedBlock(uiTheme, width => {
+			const { expanded } = options;
+			let body = renderContentPreview(fileContent, expanded, lang, uiTheme);
+			if (diagnostics) {
+				const diagText = formatDiagnostics(diagnostics, expanded, uiTheme, fp =>
+					uiTheme.getLangIcon(getLanguageFromPath(fp)),
+				);
+				if (diagText.trim()) {
+					const diagLines = diagText.split("\n");
+					const firstNonEmpty = diagLines.findIndex(line => line.trim());
+					if (firstNonEmpty >= 0) body += `\n${diagLines.slice(firstNonEmpty).join("\n")}`;
 				}
-
-				const lines = text.split("\n").map(l => truncateToWidth(l, width, Ellipsis.Omit));
-				cached = { key, lines };
-				return lines;
-			},
-			invalidate() {
-				cached = undefined;
-			},
-		};
+			}
+			const bodyLines = body.split("\n");
+			while (bodyLines.length > 0 && bodyLines[0].trim() === "") bodyLines.shift();
+			return {
+				header,
+				sections: bodyLines.length > 0 ? [{ lines: bodyLines }] : [],
+				state: "success",
+				borderColor: "borderMuted",
+				width,
+			};
+		});
 	},
 	mergeCallAndResult: true,
 };
diff --git a/packages/coding-agent/src/tui/code-cell.ts b/packages/coding-agent/src/tui/code-cell.ts
index cead51153..c2f061807 100644
--- a/packages/coding-agent/src/tui/code-cell.ts
+++ b/packages/coding-agent/src/tui/code-cell.ts
@@ -25,6 +25,12 @@ export interface CodeCellOptions {
 	output?: string;
 	outputMaxLines?: number;
 	codeMaxLines?: number;
+	/**
+	 * Show the LAST `codeMaxLines` rows (the live streaming edge) instead of the
+	 * first, with a "… N earlier lines" marker on top. Lets a pending preview
+	 * follow code as it is written while staying bounded. Ignored when `expanded`.
+	 */
+	codeTail?: boolean;
 	expanded?: boolean;
 	/** Animate the cell border with a sweeping segment while pending/running. */
 	animate?: boolean;
@@ -102,13 +108,22 @@ export function renderCodeCell(options: CodeCellOptions, theme: Theme): string[]
 	const normalizedCode = replaceTabs(code ?? "");
 	const rawCodeLines = sanitizeTerminalLines(normalizedCode);
 	const maxCodeLines = expanded ? rawCodeLines.length : Math.min(rawCodeLines.length, codeMaxLines);
-	const visibleCode = rawCodeLines.slice(0, maxCodeLines).join("\n");
-	const codeLines = highlightCode(visibleCode, language);
 	const hiddenCodeLines = rawCodeLines.length - maxCodeLines;
+	const tail = options.codeTail === true && !expanded && hiddenCodeLines > 0;
+	const startIndex = tail ? rawCodeLines.length - maxCodeLines : 0;
+	const visibleCode = rawCodeLines.slice(startIndex, startIndex + maxCodeLines).join("\n");
+	const codeLines = highlightCode(visibleCode, language);
 	if (hiddenCodeLines > 0) {
 		const hint = formatExpandHint(theme, expanded, hiddenCodeLines > 0);
-		const moreLine = `${formatMoreItems(hiddenCodeLines, "line")}${hint ? ` ${hint}` : ""}`;
-		codeLines.push(theme.fg("dim", moreLine));
+		if (tail) {
+			// Earlier rows scrolled above the live tail window — mark them on top so
+			// the newest streamed line stays pinned to the bottom of the box.
+			const earlier = `… ${hiddenCodeLines} earlier line${hiddenCodeLines === 1 ? "" : "s"}${hint ? ` ${hint}` : ""}`;
+			codeLines.unshift(theme.fg("dim", earlier));
+		} else {
+			const moreLine = `${formatMoreItems(hiddenCodeLines, "line")}${hint ? ` ${hint}` : ""}`;
+			codeLines.push(theme.fg("dim", moreLine));
+		}
 	}
 
 	const outputLines: string[] = [];
diff --git a/packages/coding-agent/src/tui/output-block.ts b/packages/coding-agent/src/tui/output-block.ts
index 1973a5d39..d80344f68 100644
--- a/packages/coding-agent/src/tui/output-block.ts
+++ b/packages/coding-agent/src/tui/output-block.ts
@@ -1,8 +1,9 @@
 /**
  * Bordered output container with optional header and sections.
  */
+import type { Component } from "@oh-my-pi/pi-tui";
 import { ImageProtocol, padding, TERMINAL, visibleWidth, wrapTextWithAnsi } from "@oh-my-pi/pi-tui";
-import type { Theme } from "../modes/theme/theme";
+import type { Theme, ThemeColor } from "../modes/theme/theme";
 import { getSixelLineMask } from "../utils/sixel";
 import type { State } from "./types";
 import type { RenderCache } from "./utils";
@@ -12,14 +13,31 @@ export interface OutputBlockOptions {
 	header?: string;
 	headerMeta?: string;
 	state?: State;
-	sections?: Array<{ label?: string; lines: string[] }>;
+	sections?: Array<{ label?: string; lines: string[]; separator?: boolean }>;
 	width: number;
 	applyBg?: boolean;
+	contentPaddingLeft?: number;
 	/** Animate the border with a sweeping dark segment (pending/running state). */
 	animate?: boolean;
+	/** Override the state-derived border color. Used for muted "legacy" tool
+	 * frames that should not visually compete with framed-output tools. */
+	borderColor?: ThemeColor;
 }
 
-const BORDER_SHIMMER_TICK_MS = 16;
+const FRAMED_BLOCK_COMPONENT = Symbol("framedBlockComponent");
+
+export type FramedBlockComponent = Component & { [FRAMED_BLOCK_COMPONENT]?: true };
+
+export function markFramedBlockComponent<T extends Component>(component: T): T & FramedBlockComponent {
+	(component as T & FramedBlockComponent)[FRAMED_BLOCK_COMPONENT] = true;
+	return component as T & FramedBlockComponent;
+}
+
+export function isFramedBlockComponent(component: Component): boolean {
+	return (component as FramedBlockComponent)[FRAMED_BLOCK_COMPONENT] === true;
+}
+
+const BORDER_SHIMMER_TICK_MS = 1000 / 30;
 /** Duration of one full left↔right↔left bounce of the bottom-edge segment, in
  * ms. Position is derived from the wall clock against this fixed cycle so a
  * resize only nudges the segment proportionally instead of teleporting it. */
@@ -28,9 +46,9 @@ const BORDER_BOUNCE_MS = 3000;
 const BORDER_SEGMENT_LEN = 8;
 
 /**
- * Monotonic frame counter for animated borders, quantized to the TUI's ~16ms
- * render cap so the cache key advances once per ~60fps frame — fine enough for a
- * smooth segment sweep, coarse enough to coalesce multiple render passes that
+ * Monotonic frame counter for animated borders, quantized to the TUI's ~30fps
+ * render cap so the cache key advances once per animation frame — fine enough
+ * for a smooth segment sweep, coarse enough to coalesce multiple render passes
  * land inside the same frame.
  */
 export function borderShimmerTick(): number {
@@ -78,6 +96,11 @@ type BlockRow =
 	| { kind: "content"; inner: string }
 	| { kind: "sixel"; raw: string };
 
+function normalizeContentPaddingLeft(value: number | undefined): number {
+	if (value === undefined || !Number.isFinite(value)) return 1;
+	return Math.max(0, Math.floor(value));
+}
+
 export function renderOutputBlock(options: OutputBlockOptions, theme: Theme): string[] {
 	const { header, headerMeta, state, sections = [], width, applyBg = true } = options;
 	const h = theme.boxSharp.horizontal;
@@ -85,14 +108,15 @@ export function renderOutputBlock(options: OutputBlockOptions, theme: Theme): st
 	const cap = h.repeat(3);
 	const lineWidth = Math.max(0, width);
 	// Border colors: running/pending use accent, success uses dim (gray), error/warning keep their colors
-	const borderColor: "error" | "warning" | "accent" | "dim" =
-		state === "error"
+	const borderColor: ThemeColor =
+		options.borderColor ??
+		(state === "error"
 			? "error"
 			: state === "warning"
 				? "warning"
 				: state === "running" || state === "pending"
 					? "accent"
-					: "dim";
+					: "dim");
 	const border = (text: string) => theme.fg(borderColor, text);
 	const bgFn = (() => {
 		if (!state || !applyBg) return undefined;
@@ -107,7 +131,9 @@ export function renderOutputBlock(options: OutputBlockOptions, theme: Theme): st
 		};
 	})();
 
-	const contentWidth = Math.max(0, lineWidth - visibleWidth(`${v} `) - visibleWidth(v));
+	const contentPaddingLeft = normalizeContentPaddingLeft(options.contentPaddingLeft);
+	const contentWidth = Math.max(0, lineWidth - visibleWidth(v) - contentPaddingLeft - visibleWidth(v));
+	const contentLeftPadding = contentPaddingLeft > 0 ? padding(contentPaddingLeft) : "";
 
 	// ── Layout pass: collect row descriptors so the border perimeter length is
 	// known before the moving segment is positioned. ──
@@ -121,7 +147,11 @@ export function renderOutputBlock(options: OutputBlockOptions, theme: Theme): st
 	});
 
 	const normalizedSections = sections.length > 0 ? sections : [{ lines: [] as string[] }];
-	for (const section of normalizedSections) {
+	for (let sectionIndex = 0; sectionIndex < normalizedSections.length; sectionIndex++) {
+		const section = normalizedSections[sectionIndex]!;
+		// A labeled section always draws its titled separator bar. A label-less
+		// section can still request a plain divider via `separator`, but only
+		// between sections — leading with one would just double the header bar.
 		if (section.label) {
 			rows.push({
 				kind: "bar",
@@ -129,6 +159,12 @@ export function renderOutputBlock(options: OutputBlockOptions, theme: Theme): st
 				rightChar: theme.boxSharp.teeLeft,
 				label: section.label,
 			});
+		} else if (section.separator && sectionIndex > 0) {
+			rows.push({
+				kind: "bar",
+				leftChar: theme.boxSharp.teeRight,
+				rightChar: theme.boxSharp.teeLeft,
+			});
 		}
 		const allLines = section.lines.flatMap(l => l.split("\n"));
 		const sixelLineMask = TERMINAL.imageProtocol === ImageProtocol.Sixel ? getSixelLineMask(allLines) : undefined;
@@ -188,7 +224,12 @@ export function renderOutputBlock(options: OutputBlockOptions, theme: Theme): st
 		const rightGlyph = row.rightChar;
 		if (lineWidth <= 0) return border(leftGlyphs) + border(rightGlyph);
 		const labelText = [row.label, row.meta].filter(Boolean).join(theme.sep.dot);
-		const rawLabel = labelText ? ` ${labelText} ` : " ";
+		if (!labelText) {
+			// No header: draw a clean, continuous top/separator bar (no 1-col gap).
+			const fillCount = Math.max(0, lineWidth - visibleWidth(leftGlyphs) - visibleWidth(rightGlyph));
+			return `${border(leftGlyphs)}${border(h.repeat(fillCount))}${border(rightGlyph)}`;
+		}
+		const rawLabel = ` ${labelText} `;
 		const leftWidth = visibleWidth(leftGlyphs);
 		const rightWidth = visibleWidth(rightGlyph);
 		const maxLabelWidth = Math.max(0, lineWidth - leftWidth - rightWidth);
@@ -211,7 +252,7 @@ export function renderOutputBlock(options: OutputBlockOptions, theme: Theme): st
 		return `${leftStr}${fillStr}${rightStr}`;
 	};
 
-	const renderContent = (inner: string): string => `${border(`${v} `)}${inner}${border(v)}`;
+	const renderContent = (inner: string): string => `${border(v)}${contentLeftPadding}${inner}${border(v)}`;
 
 	const lines: string[] = [];
 	for (let r = 0; r < H; r++) {
@@ -255,15 +296,18 @@ export class CachedOutputBlock {
 	#buildKey(options: OutputBlockOptions): bigint {
 		const h = new Hasher();
 		h.u32(options.width);
+		h.u32(normalizeContentPaddingLeft(options.contentPaddingLeft));
 		h.optional(options.header);
 		h.optional(options.headerMeta);
 		h.optional(options.state);
+		h.optional(options.borderColor);
 		h.bool(options.applyBg ?? true);
 		h.bool(options.animate ?? false);
 		if (options.animate) h.u32(borderShimmerTick());
 		if (options.sections) {
 			for (const s of options.sections) {
 				h.optional(s.label);
+				h.bool(s.separator ?? false);
 				for (const line of s.lines) {
 					h.str(line);
 				}
@@ -272,3 +316,20 @@ export class CachedOutputBlock {
 		return h.digest();
 	}
 }
+
+/**
+ * Build a self-framing tool component backed by a cached output block. The
+ * `build` callback returns the block options for a given width; the cache
+ * dedupes re-renders. Pass `borderColor: "borderMuted"` for the dim "legacy"
+ * look that does not compete with the state-colored framed tools.
+ */
+export function framedBlock(theme: Theme, build: (width: number) => OutputBlockOptions): Component {
+	const block = new CachedOutputBlock();
+	// Marked so the tool-execution container treats it as self-framing (renders
+	// flush, no extra padding/background) the same way `markFramedBlockComponent`
+	// blocks are treated.
+	return markFramedBlockComponent({
+		render: (width: number): string[] => block.render(build(width), theme),
+		invalidate: () => block.invalidate(),
+	});
+}
diff --git a/packages/coding-agent/src/tui/status-line.ts b/packages/coding-agent/src/tui/status-line.ts
index 043b78e41..0a227a28b 100644
--- a/packages/coding-agent/src/tui/status-line.ts
+++ b/packages/coding-agent/src/tui/status-line.ts
@@ -7,6 +7,9 @@ import { formatStatusIcon } from "../tools/render-utils";
 
 export interface StatusLineOptions {
 	icon?: ToolUIStatus;
+	/** Pre-rendered glyph that replaces the status icon (e.g. a magnifier for
+	 * search-family tools). Takes precedence over `icon`. */
+	iconOverride?: string;
 	spinnerFrame?: number;
 	title: string;
 	titleColor?: ThemeColor;
@@ -27,7 +30,8 @@ function flattenForHeader(text: string): string {
 }
 
 export function renderStatusLine(options: StatusLineOptions, theme: Theme): string {
-	const icon = options.icon ? formatStatusIcon(options.icon, theme, options.spinnerFrame) : "";
+	const icon =
+		options.iconOverride ?? (options.icon ? formatStatusIcon(options.icon, theme, options.spinnerFrame) : "");
 	const titleColor = options.titleColor ?? "accent";
 	const title = theme.fg(titleColor, flattenForHeader(options.title));
 	let line = icon ? `${icon} ${title}` : title;
diff --git a/packages/coding-agent/src/utils/commit-message-generator.ts b/packages/coding-agent/src/utils/commit-message-generator.ts
index cdc16f491..7346895c2 100644
--- a/packages/coding-agent/src/utils/commit-message-generator.ts
+++ b/packages/coding-agent/src/utils/commit-message-generator.ts
@@ -6,6 +6,7 @@ import type { ThinkingLevel } from "@oh-my-pi/pi-agent-core";
 import type { Api, Model } from "@oh-my-pi/pi-ai";
 import { completeSimple } from "@oh-my-pi/pi-ai";
 import { logger, prompt } from "@oh-my-pi/pi-utils";
+
 import type { ModelRegistry } from "../config/model-registry";
 import { resolveModelRoleValue } from "../config/model-resolver";
 import type { Settings } from "../config/settings";
@@ -110,7 +111,14 @@ export async function generateCommitMessage(
 					systemPrompt: [COMMIT_SYSTEM_PROMPT],
 					messages: [{ role: "user", content: userMessage, timestamp: Date.now() }],
 				},
-				{ apiKey, maxTokens, reasoning: toReasoningEffort(candidate.thinkingLevel) },
+				{
+					apiKey: registry.resolver(candidate.model.provider, {
+						sessionId,
+						baseUrl: candidate.model.baseUrl,
+					}),
+					maxTokens,
+					reasoning: toReasoningEffort(candidate.thinkingLevel),
+				},
 			);
 
 			if (response.stopReason === "error") {
diff --git a/packages/coding-agent/src/utils/enhanced-paste.ts b/packages/coding-agent/src/utils/enhanced-paste.ts
new file mode 100644
index 000000000..16fdf416e
--- /dev/null
+++ b/packages/coding-agent/src/utils/enhanced-paste.ts
@@ -0,0 +1,202 @@
+import type { ImageContent } from "@oh-my-pi/pi-ai";
+
+const OSC5522_PREFIX = "\x1b]5522;";
+const OSC_TERMINATOR_ST = "\x1b\\";
+const OSC_TERMINATOR_BEL = "\x07";
+const PASTE_EVENT_NAME_BASE64 = Buffer.from("Paste event", "utf8").toString("base64");
+
+const IMAGE_MIME_PRIORITY = ["image/png", "image/jpeg", "image/webp", "image/gif"] as const;
+const TEXT_MIME_TYPE = "text/plain";
+
+type PasteReadKind = "image" | "text";
+
+export interface Osc5522Packet {
+	metadata: Map<string, string>;
+	payload: string;
+}
+
+interface PasteListingState {
+	phase: "listing";
+	mimes: string[];
+	pw?: string;
+	loc?: string;
+}
+
+interface PasteReadState {
+	phase: "reading";
+	kind: PasteReadKind;
+	mimeType: string;
+	chunks: string[];
+}
+
+type PasteState = PasteListingState | PasteReadState;
+
+export interface EnhancedPasteHandlers {
+	write(data: string): void;
+	pasteText(text: string): void;
+	pasteImage(image: ImageContent): void | Promise<void>;
+	showStatus(message: string): void;
+}
+
+export function isOsc5522Packet(data: string): boolean {
+	return data.startsWith(OSC5522_PREFIX) && (data.endsWith(OSC_TERMINATOR_ST) || data.endsWith(OSC_TERMINATOR_BEL));
+}
+
+function decodeBase64Utf8(value: string): string | undefined {
+	try {
+		return Buffer.from(value, "base64").toString("utf8");
+	} catch {
+		return undefined;
+	}
+}
+
+function parseMetadata(raw: string): Map<string, string> {
+	const metadata = new Map<string, string>();
+	for (const part of raw.split(":")) {
+		const eq = part.indexOf("=");
+		if (eq <= 0) continue;
+		metadata.set(part.slice(0, eq), part.slice(eq + 1));
+	}
+	return metadata;
+}
+
+export function parseOsc5522Packet(data: string): Osc5522Packet | undefined {
+	if (!isOsc5522Packet(data)) return undefined;
+	const bodyEnd = data.endsWith(OSC_TERMINATOR_BEL) ? data.length - 1 : data.length - OSC_TERMINATOR_ST.length;
+	const body = data.slice(OSC5522_PREFIX.length, bodyEnd);
+	const separator = body.indexOf(";");
+	const metadataRaw = separator === -1 ? body : body.slice(0, separator);
+	const payload = separator === -1 ? "" : body.slice(separator + 1);
+	return { metadata: parseMetadata(metadataRaw), payload };
+}
+
+function choosePasteMime(mimes: readonly string[]): { kind: PasteReadKind; mimeType: string } | undefined {
+	for (const mimeType of IMAGE_MIME_PRIORITY) {
+		if (mimes.includes(mimeType)) return { kind: "image", mimeType };
+	}
+	return mimes.includes(TEXT_MIME_TYPE) ? { kind: "text", mimeType: TEXT_MIME_TYPE } : undefined;
+}
+
+export class EnhancedPasteController {
+	#state: PasteState | undefined;
+	#handlers: EnhancedPasteHandlers;
+
+	constructor(handlers: EnhancedPasteHandlers) {
+		this.#handlers = handlers;
+	}
+
+	enable(): void {
+		this.#handlers.write("\x1b[?5522h");
+	}
+
+	disable(): void {
+		this.#handlers.write("\x1b[?5522l");
+		this.#state = undefined;
+	}
+
+	handleInput(data: string): boolean {
+		const packet = parseOsc5522Packet(data);
+		if (!packet) return false;
+		void this.#handlePacket(packet);
+		return true;
+	}
+
+	async #handlePacket(packet: Osc5522Packet): Promise<void> {
+		const type = packet.metadata.get("type");
+		if (type !== "read") return;
+
+		const status = packet.metadata.get("status");
+		if (status === "OK") {
+			this.#handleOk(packet);
+			return;
+		}
+		if (status === "DATA") {
+			this.#handleData(packet);
+			return;
+		}
+		if (status === "DONE") {
+			await this.#handleDone();
+			return;
+		}
+		if (status) {
+			this.#state = undefined;
+			this.#handlers.showStatus(`Enhanced paste failed: ${status}`);
+		}
+	}
+
+	#handleOk(packet: Osc5522Packet): void {
+		if (this.#state?.phase === "reading") return;
+		const loc = packet.metadata.get("loc");
+		this.#state = {
+			phase: "listing",
+			mimes: [],
+			pw: packet.metadata.get("pw"),
+			loc: loc === "primary" ? loc : undefined,
+		};
+	}
+
+	#handleData(packet: Osc5522Packet): void {
+		const state = this.#state;
+		if (!state) return;
+		const encodedMime = packet.metadata.get("mime");
+		if (!encodedMime) return;
+		const mimeType = decodeBase64Utf8(encodedMime);
+		if (!mimeType) return;
+
+		if (state.phase === "listing") {
+			state.mimes.push(mimeType);
+			return;
+		}
+
+		if (state.mimeType === mimeType && packet.payload) {
+			state.chunks.push(packet.payload);
+		}
+	}
+
+	async #handleDone(): Promise<void> {
+		const state = this.#state;
+		if (!state) return;
+		if (state.phase === "listing") {
+			this.#finishListing(state);
+			return;
+		}
+		this.#state = undefined;
+		const bytes = Buffer.concat(state.chunks.map(chunk => Buffer.from(chunk, "base64")));
+		if (bytes.byteLength === 0) {
+			this.#handlers.showStatus("Clipboard paste was empty");
+			return;
+		}
+		if (state.kind === "text") {
+			this.#handlers.pasteText(bytes.toString("utf8"));
+			return;
+		}
+		await this.#handlers.pasteImage({
+			type: "image",
+			data: bytes.toString("base64"),
+			mimeType: state.mimeType,
+		});
+	}
+
+	#finishListing(state: PasteListingState): void {
+		const selected = choosePasteMime(state.mimes);
+		if (!selected) {
+			this.#state = undefined;
+			this.#handlers.showStatus("Clipboard paste has no supported text or image data");
+			return;
+		}
+
+		this.#state = {
+			phase: "reading",
+			kind: selected.kind,
+			mimeType: selected.mimeType,
+			chunks: [],
+		};
+
+		const metadata = [`type=read`, `mime=${Buffer.from(selected.mimeType, "utf8").toString("base64")}`];
+		if (state.loc) metadata.push(`loc=${state.loc}`);
+		if (state.pw) {
+			metadata.push(`pw=${state.pw}`, `name=${PASTE_EVENT_NAME_BASE64}`);
+		}
+		this.#handlers.write(`${OSC5522_PREFIX}${metadata.join(":")}${OSC_TERMINATOR_ST}`);
+	}
+}
diff --git a/packages/coding-agent/src/utils/title-generator.ts b/packages/coding-agent/src/utils/title-generator.ts
index 84f4eab1e..4761c5e60 100644
--- a/packages/coding-agent/src/utils/title-generator.ts
+++ b/packages/coding-agent/src/utils/title-generator.ts
@@ -6,6 +6,7 @@ import * as path from "node:path";
 import { type Api, type AssistantMessage, completeSimple, type Model, type Tool } from "@oh-my-pi/pi-ai";
 import { logger, prompt } from "@oh-my-pi/pi-utils";
 import type { ModelRegistry } from "../config/model-registry";
+
 import { resolveRoleSelection } from "../config/model-resolver";
 import type { Settings } from "../config/settings";
 import titleSystemPrompt from "../prompts/system/title-system.md" with { type: "text" };
@@ -238,7 +239,7 @@ export async function generateTitleOnline(
 				tools: [setTitleTool],
 			},
 			{
-				apiKey,
+				apiKey: registry.resolver(model.provider, { sessionId, baseUrl: model.baseUrl }),
 				maxTokens,
 				disableReasoning: true,
 				toolChoice: { type: "tool", name: SET_TITLE_TOOL_NAME },
diff --git a/packages/coding-agent/src/web/scrapers/github.ts b/packages/coding-agent/src/web/scrapers/github.ts
index 50ef8b50c..b7b7990c2 100644
--- a/packages/coding-agent/src/web/scrapers/github.ts
+++ b/packages/coding-agent/src/web/scrapers/github.ts
@@ -1,14 +1,28 @@
 import { $env, ptree } from "@oh-my-pi/pi-utils";
 import type { RenderResult, SpecialHandler } from "./types";
-import { buildResult, loadPage } from "./types";
+import { buildResult, formatMediaDuration, loadPage } from "./types";
 
 interface GitHubUrl {
-	type: "blob" | "tree" | "repo" | "issue" | "issues" | "pull" | "pulls" | "discussion" | "discussions" | "other";
+	type:
+		| "blob"
+		| "tree"
+		| "repo"
+		| "issue"
+		| "issues"
+		| "pull"
+		| "pulls"
+		| "discussion"
+		| "discussions"
+		| "actions-run"
+		| "actions-job"
+		| "other";
 	owner: string;
 	repo: string;
 	ref?: string;
 	path?: string;
 	number?: number;
+	runId?: number;
+	jobId?: number;
 }
 
 interface GitHubIssueComment {
@@ -20,7 +34,7 @@ interface GitHubIssueComment {
 /**
  * Parse GitHub URL into components
  */
-function parseGitHubUrl(url: string): GitHubUrl | null {
+export function parseGitHubUrl(url: string): GitHubUrl | null {
 	try {
 		const parsed = new URL(url);
 		if (parsed.hostname !== "github.com") return null;
@@ -54,6 +68,20 @@ function parseGitHubUrl(url: string): GitHubUrl | null {
 				return { type: "pulls", owner, repo };
 			case "pulls":
 				return { type: "pulls", owner, repo };
+			case "actions": {
+				// /actions/runs/{runId}                      → run summary + jobs
+				// /actions/runs/{runId}/job/{jobId}          → single job (web URL uses singular "job")
+				// /actions/runs/{runId}/jobs/{jobId}         → single job (API-style plural)
+				if (subParts[0] === "runs" && /^\d+$/.test(subParts[1] ?? "")) {
+					const runId = parseInt(subParts[1], 10);
+					const seg = subParts[2];
+					if ((seg === "job" || seg === "jobs") && /^\d+$/.test(subParts[3] ?? "")) {
+						return { type: "actions-job", owner, repo, runId, jobId: parseInt(subParts[3], 10) };
+					}
+					return { type: "actions-run", owner, repo, runId };
+				}
+				return { type: "other", owner, repo };
+			}
 			case "discussions":
 				if (subParts.length > 0 && /^\d+$/.test(subParts[0])) {
 					return { type: "discussion", owner, repo, number: parseInt(subParts[0], 10) };
@@ -371,6 +399,212 @@ async function renderGitHubRepo(
 	return { content: md, ok: true };
 }
 
+interface GitHubActionsStep {
+	name: string;
+	status: string;
+	conclusion: string | null;
+	number: number;
+	started_at: string | null;
+	completed_at: string | null;
+}
+
+interface GitHubActionsJob {
+	id: number;
+	run_id: number;
+	name: string;
+	status: string;
+	conclusion: string | null;
+	started_at: string | null;
+	completed_at: string | null;
+	html_url: string | null;
+	steps?: GitHubActionsStep[];
+	runner_name?: string | null;
+	labels?: string[];
+	workflow_name?: string | null;
+	head_branch?: string | null;
+	head_sha?: string;
+}
+
+interface GitHubActionsRun {
+	id: number;
+	name?: string | null;
+	display_title?: string;
+	run_number: number;
+	run_attempt?: number;
+	event: string;
+	status: string;
+	conclusion: string | null;
+	head_branch?: string | null;
+	head_sha?: string;
+	html_url: string;
+	created_at: string;
+	updated_at: string;
+	run_started_at?: string;
+	actor?: { login: string };
+	triggering_actor?: { login: string };
+}
+
+/** Combine status + conclusion into a single label, e.g. `completed (failure)`. */
+function statusLabel(status: string, conclusion: string | null | undefined): string {
+	return conclusion ? `${status} (${conclusion})` : status;
+}
+
+/** Wall-clock duration between two ISO timestamps, formatted HH:MM:SS / MM:SS. Empty when unknown. */
+function actionDuration(start?: string | null, end?: string | null): string {
+	if (!start || !end) return "";
+	const ms = Date.parse(end) - Date.parse(start);
+	if (!Number.isFinite(ms) || ms < 0) return "";
+	return formatMediaDuration(Math.round(ms / 1000));
+}
+
+/** Escape `|` so step/job names can't break a markdown table row. */
+function escapeCell(text: string): string {
+	return text.replaceAll("|", "\\|");
+}
+
+/**
+ * Strip the per-line ISO-8601 timestamp prefix GitHub prepends to every job log line.
+ * Cuts ~28 bytes/line of noise while preserving the message text. Also drops the leading
+ * UTF-8 BOM GitHub puts at the start of the log file (otherwise the first line's timestamp
+ * survives because `^` no longer sits before a digit).
+ */
+export function stripActionsLogTimestamps(logs: string): string {
+	return logs.replace(/^\uFEFF/, "").replace(/^\d{4}-\d{2}-\d{2}T[\d:.]+Z /gm, "");
+}
+
+/** Render a job's steps as a markdown table. Empty string when there are no steps. */
+function renderActionsSteps(steps?: GitHubActionsStep[]): string {
+	if (!steps || steps.length === 0) return "";
+	let md = "| # | Step | Status | Conclusion | Duration |\n";
+	md += "|---|------|--------|------------|----------|\n";
+	for (const step of steps) {
+		const dur = actionDuration(step.started_at, step.completed_at) || "-";
+		md += `| ${step.number} | ${escapeCell(step.name)} | ${step.status} | ${step.conclusion ?? "-"} | ${dur} |\n`;
+	}
+	return `${md}\n`;
+}
+
+/** Run-level metadata lines shared by the run and job renderers. */
+function renderActionsRunMeta(run: GitHubActionsRun): string {
+	let md = `**Workflow:** ${run.name ?? "(unknown)"}\n`;
+	md += `**Run:** #${run.run_number}`;
+	if (run.run_attempt && run.run_attempt > 1) md += ` (attempt ${run.run_attempt})`;
+	md += ` · ${statusLabel(run.status, run.conclusion)}\n`;
+	if (run.head_branch) {
+		md += `**Branch:** ${run.head_branch}${run.head_sha ? ` @ ${run.head_sha.slice(0, 7)}` : ""}\n`;
+	}
+	const actor = run.triggering_actor?.login ?? run.actor?.login;
+	md += `**Event:** ${run.event}${actor ? ` · by @${actor}` : ""}\n`;
+	const started = run.run_started_at ?? run.created_at;
+	const dur = actionDuration(started, run.updated_at);
+	md += `Started: ${started}${dur ? ` · Duration: ${dur}` : ""}\n`;
+	md += `URL: ${run.html_url}\n`;
+	return md;
+}
+
+/** Fetch a job's plain-text logs. Returns null when unavailable (no token / expired / private). */
+async function fetchGitHubJobLogs(
+	owner: string,
+	repo: string,
+	jobId: number,
+	timeout: number,
+	signal?: AbortSignal,
+): Promise<string | null> {
+	const headers: Record<string, string> = {
+		Accept: "application/vnd.github+json",
+		"X-GitHub-Api-Version": "2022-11-28",
+	};
+	const token = $env.GITHUB_TOKEN || $env.GH_TOKEN;
+	if (token) headers.Authorization = `Bearer ${token}`;
+
+	// 302 → signed log URL on a different origin; fetch strips Authorization on the cross-origin hop.
+	const result = await loadPage(`https://api.github.com/repos/${owner}/${repo}/actions/jobs/${jobId}/logs`, {
+		timeout,
+		headers,
+		signal,
+	});
+	return result.ok && result.content ? result.content : null;
+}
+
+/**
+ * Render a workflow run: run metadata plus a per-job breakdown. Steps are listed for any job that
+ * did not succeed (the debugging-relevant ones); successful jobs collapse to a single line.
+ */
+async function renderGitHubActionsRun(
+	gh: GitHubUrl,
+	timeout: number,
+	signal?: AbortSignal,
+): Promise<{ content: string; ok: boolean }> {
+	const runResult = await fetchGitHubApi(`/repos/${gh.owner}/${gh.repo}/actions/runs/${gh.runId}`, timeout, signal);
+	if (!runResult.ok || !runResult.data) return { content: "", ok: false };
+
+	const run = runResult.data as GitHubActionsRun;
+	let md = `# ${run.display_title || run.name || `Run #${run.run_number}`}\n\n`;
+	md += renderActionsRunMeta(run);
+	md += `\n---\n\n`;
+
+	const jobsResult = await fetchGitHubApi(
+		`/repos/${gh.owner}/${gh.repo}/actions/runs/${gh.runId}/jobs?per_page=100`,
+		timeout,
+		signal,
+	);
+	if (jobsResult.ok && jobsResult.data) {
+		const jobs = (jobsResult.data as { jobs?: GitHubActionsJob[] }).jobs ?? [];
+		md += `## Jobs (${jobs.length})\n\n`;
+		for (const job of jobs) {
+			const dur = actionDuration(job.started_at, job.completed_at);
+			md += `### ${escapeCell(job.name)} — ${statusLabel(job.status, job.conclusion)}${dur ? ` (${dur})` : ""}\n\n`;
+			if (job.conclusion !== "success") {
+				md += renderActionsSteps(job.steps);
+			}
+		}
+	}
+
+	return { content: md, ok: true };
+}
+
+/**
+ * Render a single workflow job: run context, step table, and the full job logs.
+ */
+async function renderGitHubActionsJob(
+	gh: GitHubUrl,
+	timeout: number,
+	signal?: AbortSignal,
+): Promise<{ content: string; ok: boolean }> {
+	const jobResult = await fetchGitHubApi(`/repos/${gh.owner}/${gh.repo}/actions/jobs/${gh.jobId}`, timeout, signal);
+	if (!jobResult.ok || !jobResult.data) return { content: "", ok: false };
+
+	const job = jobResult.data as GitHubActionsJob;
+
+	// Best-effort run context for nicer headers; the job render stands on its own without it.
+	const runResult = await fetchGitHubApi(`/repos/${gh.owner}/${gh.repo}/actions/runs/${job.run_id}`, timeout, signal);
+	const run = runResult.ok && runResult.data ? (runResult.data as GitHubActionsRun) : null;
+
+	let md = `# ${escapeCell(job.name)}\n\n`;
+	if (run) {
+		md += renderActionsRunMeta(run);
+	} else if (job.workflow_name) {
+		md += `**Workflow:** ${job.workflow_name}\n`;
+		if (job.head_branch) md += `**Branch:** ${job.head_branch}\n`;
+	}
+	const dur = actionDuration(job.started_at, job.completed_at);
+	md += `**Job:** ${escapeCell(job.name)} · ${statusLabel(job.status, job.conclusion)}${dur ? ` · ${dur}` : ""}\n`;
+	if (job.runner_name) md += `**Runner:** ${job.runner_name}\n`;
+	if (job.html_url) md += `URL: ${job.html_url}\n`;
+	md += `\n---\n\n`;
+
+	const steps = renderActionsSteps(job.steps);
+	if (steps) md += `## Steps\n\n${steps}`;
+
+	const logs = await fetchGitHubJobLogs(gh.owner, gh.repo, job.id, timeout, signal);
+	md += `## Logs\n\n`;
+	md += logs
+		? stripActionsLogTimestamps(logs)
+		: "*Logs unavailable — requires a GITHUB_TOKEN/GH_TOKEN with read access, or the run's logs have expired.*\n";
+
+	return { content: md, ok: true };
+}
+
 /**
  * Handle GitHub URLs specially
  */
@@ -445,6 +679,24 @@ export const handleGitHub: SpecialHandler = async (
 			}
 			break;
 		}
+
+		case "actions-run": {
+			notes.push(`Fetched via GitHub API`);
+			const result = await renderGitHubActionsRun(gh, timeout, signal);
+			if (result.ok) {
+				return buildResult(result.content, { url, method: "github-actions-run", fetchedAt, notes });
+			}
+			break;
+		}
+
+		case "actions-job": {
+			notes.push(`Fetched via GitHub API`);
+			const result = await renderGitHubActionsJob(gh, timeout, signal);
+			if (result.ok) {
+				return buildResult(result.content, { url, method: "github-actions-job", fetchedAt, notes });
+			}
+			break;
+		}
 	}
 
 	// Fall back to null (let normal rendering handle it)
diff --git a/packages/coding-agent/src/web/scrapers/youtube.ts b/packages/coding-agent/src/web/scrapers/youtube.ts
index 1ecd4090f..6dec1276b 100644
--- a/packages/coding-agent/src/web/scrapers/youtube.ts
+++ b/packages/coding-agent/src/web/scrapers/youtube.ts
@@ -113,8 +113,9 @@ export const handleYouTube: SpecialHandler = async (
 	const notes: string[] = [];
 	const videoUrl = `https://www.youtube.com/watch?v=${yt.videoId}`;
 
-	// Prefer Parallel extract when credentials are available
-	if (settings.get("providers.parallelFetch") && findParallelApiKey(storage)) {
+	// Prefer Parallel extract when it sits in the reader chain and creds exist
+	const fetchPreference = settings.get("providers.fetch");
+	if ((fetchPreference === "auto" || fetchPreference === "parallel") && findParallelApiKey(storage)) {
 		try {
 			const parallelResult = await extractWithParallel(
 				[videoUrl],
diff --git a/packages/coding-agent/src/web/search/providers/anthropic.ts b/packages/coding-agent/src/web/search/providers/anthropic.ts
index 07a2df740..5a62bb5d8 100644
--- a/packages/coding-agent/src/web/search/providers/anthropic.ts
+++ b/packages/coding-agent/src/web/search/providers/anthropic.ts
@@ -7,12 +7,14 @@
 import {
 	type AnthropicAuthConfig,
 	type AnthropicSystemBlock,
+	type ApiKey,
 	type AuthStorage,
 	buildAnthropicAuthConfig,
 	buildAnthropicSearchHeaders,
 	buildAnthropicSystemBlocks,
 	buildAnthropicUrl,
 	stripClaudeToolPrefix,
+	withAuth,
 } from "@oh-my-pi/pi-ai";
 import { $env } from "@oh-my-pi/pi-utils";
 import type {
@@ -247,18 +249,13 @@ export async function searchAnthropic(
 ): Promise<SearchResponse> {
 	const searchApiKey = $env.ANTHROPIC_SEARCH_API_KEY;
 	const searchBaseUrl = $env.ANTHROPIC_SEARCH_BASE_URL;
-	let auth: AnthropicAuthConfig | undefined;
+	const keyOrResolver: ApiKey | undefined = searchApiKey
+		? searchApiKey
+		: "authStorage" in params
+			? params.authStorage.resolver("anthropic", { sessionId: params.sessionId })
+			: undefined;
 
-	if (searchApiKey) {
-		auth = buildAnthropicAuthConfig(searchApiKey, searchBaseUrl);
-	} else if ("authStorage" in params) {
-		const apiKey = await params.authStorage.getApiKey("anthropic", params.sessionId, {
-			signal: params.signal,
-		});
-		if (apiKey) auth = buildAnthropicAuthConfig(apiKey, searchBaseUrl);
-	}
-
-	if (!auth) {
+	if (!keyOrResolver) {
 		throw new Error(
 			"No Anthropic credentials found. Set ANTHROPIC_SEARCH_API_KEY or ANTHROPIC_API_KEY, or configure Anthropic OAuth.",
 		);
@@ -267,14 +264,23 @@ export async function searchAnthropic(
 	const model = getModel();
 	const systemPrompt = "authStorage" in params ? params.systemPrompt : params.system_prompt;
 	const maxTokens = "authStorage" in params ? params.maxOutputTokens : params.max_tokens;
-	const response = await callSearch(
-		auth,
-		model,
-		params.query,
-		systemPrompt,
-		maxTokens,
-		params.temperature,
-		params.signal,
+	const response = await withAuth(
+		keyOrResolver,
+		key =>
+			callSearch(
+				buildAnthropicAuthConfig(key, searchBaseUrl),
+				model,
+				params.query,
+				systemPrompt,
+				maxTokens,
+				params.temperature,
+				params.signal,
+			),
+		{
+			signal: params.signal,
+			missingKeyMessage:
+				"No Anthropic credentials found. Set ANTHROPIC_SEARCH_API_KEY or ANTHROPIC_API_KEY, or configure Anthropic OAuth.",
+		},
 	);
 
 	const result = parseResponse(response);
diff --git a/packages/coding-agent/src/web/search/providers/exa.ts b/packages/coding-agent/src/web/search/providers/exa.ts
index bbca2a5b7..7c0642cbd 100644
--- a/packages/coding-agent/src/web/search/providers/exa.ts
+++ b/packages/coding-agent/src/web/search/providers/exa.ts
@@ -6,7 +6,7 @@
  * Requests per-result summaries via `contents.summary` and synthesizes
  * them into a combined `answer` string on the SearchResponse.
  */
-import { type AuthStorage, getEnvApiKey } from "@oh-my-pi/pi-ai";
+import { type ApiKey, type AuthStorage, getEnvApiKey, withAuth } from "@oh-my-pi/pi-ai";
 import { settings } from "../../../config/settings";
 import { callExaTool, findApiKey, isSearchResponse } from "../../../exa/mcp-client";
 
@@ -228,11 +228,19 @@ async function callExaMcpSearch(params: ExaSearchParams): Promise<ExaSearchRespo
 
 /** Execute Exa web search */
 export async function searchExa(params: ExaSearchParams): Promise<SearchResponse> {
+	// AuthStorage-backed key takes precedence (existing behavior); probe it once
+	// so the env-key and keyless-MCP fallbacks below stay intact, then drive the
+	// authStorage path through the central force-refresh/rotate retry policy.
 	const storedKey = params.authStorage
 		? await params.authStorage.getApiKey("exa", params.sessionId, { signal: params.signal })
 		: undefined;
-	const apiKey = storedKey ?? getEnvApiKey("exa");
-	const response = apiKey ? await callExaSearch(apiKey, params) : await callExaMcpSearch(params);
+	const keyOrResolver: ApiKey | undefined =
+		storedKey && params.authStorage
+			? params.authStorage.resolver("exa", { sessionId: params.sessionId })
+			: getEnvApiKey("exa");
+	const response = keyOrResolver
+		? await withAuth(keyOrResolver, key => callExaSearch(key, params), { signal: params.signal })
+		: await callExaMcpSearch(params);
 
 	// Convert to unified SearchResponse
 	const sources: SearchSource[] = [];
diff --git a/packages/coding-agent/src/web/search/providers/kimi.ts b/packages/coding-agent/src/web/search/providers/kimi.ts
index 2bc8d53b1..08401dbe6 100644
--- a/packages/coding-agent/src/web/search/providers/kimi.ts
+++ b/packages/coding-agent/src/web/search/providers/kimi.ts
@@ -4,7 +4,7 @@
  * Uses Moonshot Kimi Code search API to retrieve web results.
  * Endpoint: POST https://api.kimi.com/coding/v1/search
  */
-import type { AuthStorage } from "@oh-my-pi/pi-ai";
+import { type ApiKey, type AuthStorage, withAuth } from "@oh-my-pi/pi-ai";
 import { $env } from "@oh-my-pi/pi-utils";
 
 import type { SearchResponse, SearchSource } from "../../../web/search/types";
@@ -54,20 +54,26 @@ function resolveBaseUrl(): string {
 	return asTrimmed($env.MOONSHOT_SEARCH_BASE_URL) ?? asTrimmed($env.KIMI_SEARCH_BASE_URL) ?? KIMI_SEARCH_URL;
 }
 
-/** Find Kimi search credentials from environment or AuthStorage. */
-async function findApiKey(
+/**
+ * Resolve the Kimi search credential. Highest precedence is the static env key;
+ * otherwise an AuthStorage-backed resolver for whichever stored provider id
+ * holds a key (`moonshot` first, then `kimi-code`), so a stale token triggers
+ * the central force-refresh / sibling-rotate retry. Returns `undefined` when
+ * neither is configured.
+ */
+async function resolveKey(
 	authStorage: AuthStorage,
 	sessionId: string | undefined,
 	signal: AbortSignal | undefined,
-): Promise<string | null> {
+): Promise<ApiKey | undefined> {
 	const envKey = asTrimmed($env.MOONSHOT_SEARCH_API_KEY) ?? asTrimmed($env.KIMI_SEARCH_API_KEY);
 	if (envKey) return envKey;
 
-	return (
-		(await authStorage.getApiKey("moonshot", sessionId, { signal })) ??
-		(await authStorage.getApiKey("kimi-code", sessionId, { signal })) ??
-		null
-	);
+	for (const provider of ["moonshot", "kimi-code"] as const) {
+		const stored = await authStorage.getApiKey(provider, sessionId, { signal });
+		if (stored) return authStorage.resolver(provider, { sessionId });
+	}
+	return undefined;
 }
 
 async function callKimiSearch(
@@ -108,20 +114,25 @@ async function callKimiSearch(
 
 /** Execute Kimi web search. */
 export async function searchKimi(params: KimiSearchParams): Promise<SearchResponse> {
-	const apiKey = await findApiKey(params.authStorage, params.sessionId, params.signal);
-	if (!apiKey) {
+	const keyOrResolver = await resolveKey(params.authStorage, params.sessionId, params.signal);
+	if (!keyOrResolver) {
 		throw new Error(
 			"Kimi search credentials not found. Set MOONSHOT_SEARCH_API_KEY, KIMI_SEARCH_API_KEY, MOONSHOT_API_KEY, or login with 'omp /login moonshot'.",
 		);
 	}
 
 	const limit = clampNumResults(params.num_results, DEFAULT_NUM_RESULTS, MAX_NUM_RESULTS);
-	const { response, requestId } = await callKimiSearch(apiKey, {
-		query: params.query,
-		limit,
-		includeContent: params.include_content ?? false,
-		signal: params.signal,
-	});
+	const { response, requestId } = await withAuth(
+		keyOrResolver,
+		key =>
+			callKimiSearch(key, {
+				query: params.query,
+				limit,
+				includeContent: params.include_content ?? false,
+				signal: params.signal,
+			}),
+		{ signal: params.signal },
+	);
 	const sources: SearchSource[] = [];
 
 	for (const result of response.search_results ?? []) {
diff --git a/packages/coding-agent/src/web/search/providers/parallel.ts b/packages/coding-agent/src/web/search/providers/parallel.ts
index 19f1b8164..876ef4a8c 100644
--- a/packages/coding-agent/src/web/search/providers/parallel.ts
+++ b/packages/coding-agent/src/web/search/providers/parallel.ts
@@ -1,4 +1,4 @@
-import { type AuthStorage, getEnvApiKey } from "@oh-my-pi/pi-ai";
+import { type ApiKey, type AuthStorage, getEnvApiKey, withAuth } from "@oh-my-pi/pi-ai";
 import type { SearchResponse } from "../../../web/search/types";
 import { SearchProviderError } from "../../../web/search/types";
 import { ParallelApiError, type ParallelSearchResult, type ParallelSearchSource } from "../../parallel";
@@ -123,30 +123,41 @@ async function searchWithAuthStorage(
 		);
 	}
 
-	const response = await fetch(PARALLEL_SEARCH_URL, {
-		method: "POST",
-		headers: {
-			Accept: "application/json",
-			"Content-Type": "application/json",
-			"x-api-key": apiKey,
-			"parallel-beta": PARALLEL_BETA_HEADER,
-		},
-		body: JSON.stringify({
-			objective,
-			search_queries: queries,
-			mode: "fast",
-			excerpts: {
-				max_chars_per_result: 10_000,
-			},
-		}),
-		signal: withHardTimeout(params.signal),
-	});
-	if (!response.ok) {
-		throw parseParallelErrorResponse(response.status, await response.text());
-	}
+	// Drive the (already-present) credential through the central force-refresh /
+	// sibling-rotate retry policy. The `ParallelApiError` thrown below carries a
+	// `statusCode`, which `withAuth`'s default classifier reads to detect a
+	// retryable 401 / usage-limit.
+	const keyOrResolver: ApiKey = authStorage.resolver("parallel", { sessionId });
+	return withAuth(
+		keyOrResolver,
+		async key => {
+			const response = await fetch(PARALLEL_SEARCH_URL, {
+				method: "POST",
+				headers: {
+					Accept: "application/json",
+					"Content-Type": "application/json",
+					"x-api-key": key,
+					"parallel-beta": PARALLEL_BETA_HEADER,
+				},
+				body: JSON.stringify({
+					objective,
+					search_queries: queries,
+					mode: "fast",
+					excerpts: {
+						max_chars_per_result: 10_000,
+					},
+				}),
+				signal: withHardTimeout(params.signal),
+			});
+			if (!response.ok) {
+				throw parseParallelErrorResponse(response.status, await response.text());
+			}
 
-	const payload: unknown = await response.json();
-	return parseSearchPayload(payload);
+			const payload: unknown = await response.json();
+			return parseSearchPayload(payload);
+		},
+		{ signal: params.signal },
+	);
 }
 
 export async function searchParallel(
diff --git a/packages/coding-agent/src/web/search/providers/perplexity.ts b/packages/coding-agent/src/web/search/providers/perplexity.ts
index 2bc2ac837..70284cdfb 100644
--- a/packages/coding-agent/src/web/search/providers/perplexity.ts
+++ b/packages/coding-agent/src/web/search/providers/perplexity.ts
@@ -1,10 +1,11 @@
 /**
  * Perplexity Web Search Provider
  *
- * Supports three auth modes:
+ * Supports four auth modes:
  * - Cookies (`PERPLEXITY_COOKIES`) via `www.perplexity.ai/rest/sse/perplexity_ask`
  * - OAuth/session bearer via `AuthStorage` and `www.perplexity.ai/rest/sse/perplexity_ask`
  * - API key (`PERPLEXITY_API_KEY`) via `api.perplexity.ai/chat/completions`
+ * - Anonymous via `www.perplexity.ai/rest/sse/perplexity_ask`
  */
 
 import { type AuthStorage, getEnvApiKey } from "@oh-my-pi/pi-ai";
@@ -32,6 +33,8 @@ const DEFAULT_NUM_SEARCH_RESULTS = 20;
 const OAUTH_EXPIRY_BUFFER_MS = 5 * 60 * 1000;
 const OAUTH_API_VERSION = "2.18";
 const OAUTH_USER_AGENT = "Perplexity/641 CFNetwork/1568 Darwin/25.2.0";
+const ANONYMOUS_USER_AGENT =
+	"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/149.0.0.0 Safari/537.36";
 
 type PerplexityAuth =
 	| {
@@ -45,6 +48,9 @@ type PerplexityAuth =
 	| {
 			type: "cookies";
 			cookies: string;
+	  }
+	| {
+			type: "anonymous";
 	  };
 
 interface PerplexityOAuthStreamMarkdownBlock {
@@ -149,6 +155,112 @@ function mergeOAuthEventSnapshot(
 
 	return merged;
 }
+
+function asRecord(value: unknown): Record<string, unknown> | null {
+	if (typeof value !== "object" || value === null || Array.isArray(value)) return null;
+	return value as Record<string, unknown>;
+}
+
+function parseJson(text: string): unknown | null {
+	try {
+		return JSON.parse(text);
+	} catch {
+		return null;
+	}
+}
+
+function textFromChunks(value: unknown): string | null {
+	if (!Array.isArray(value) || value.length === 0) return null;
+	let text = "";
+	for (const chunk of value) {
+		if (typeof chunk !== "string") return null;
+		text += chunk;
+	}
+	return text.length > 0 ? text : null;
+}
+
+function textFromStructuredAnswer(value: unknown): string | null {
+	if (!Array.isArray(value)) return null;
+	for (const item of value) {
+		const record = asRecord(item);
+		if (!record) continue;
+		const text = record.text;
+		if (typeof text === "string" && text.length > 0) return text;
+		const chunks = textFromChunks(record.chunks);
+		if (chunks) return chunks;
+	}
+	return null;
+}
+
+function answerFromTextPayload(payload: Record<string, unknown>): string | null {
+	const structured = textFromStructuredAnswer(payload.structured_answer);
+	if (structured) return structured;
+	const chunks = textFromChunks(payload.chunks);
+	if (chunks) return chunks;
+	const answer = payload.answer;
+	return typeof answer === "string" && answer.length > 0 ? answer : null;
+}
+
+function parseOAuthTextPayload(text: string): Record<string, unknown> | null {
+	const parsed = parseJson(text);
+	const direct = asRecord(parsed);
+	if (direct) return direct;
+	if (!Array.isArray(parsed)) return null;
+
+	for (const item of parsed) {
+		const step = asRecord(item);
+		const content = asRecord(step?.content);
+		const answer = content?.answer;
+		if (typeof answer !== "string" || answer.length === 0) continue;
+		const payload = asRecord(parseJson(answer));
+		if (payload) return payload;
+	}
+	return null;
+}
+
+function parseOAuthTextAnswer(text: string): string {
+	const payload = parseOAuthTextPayload(text);
+	if (payload) {
+		const answer = answerFromTextPayload(payload);
+		if (answer) return answer;
+	}
+
+	const parsed = parseJson(text);
+	if (!Array.isArray(parsed)) return text;
+	for (const item of parsed) {
+		const step = asRecord(item);
+		const content = asRecord(step?.content);
+		const answer = content?.answer;
+		if (typeof answer === "string" && answer.length > 0) return answer;
+	}
+	return text;
+}
+
+function sourcesFromTextPayload(text: string | undefined): SearchSource[] {
+	if (!text) return [];
+	const payload = parseOAuthTextPayload(text);
+	const webResults = payload?.web_results;
+	if (!Array.isArray(webResults) || webResults.length === 0) return [];
+
+	const sources: SearchSource[] = [];
+	for (const value of webResults) {
+		const result = asRecord(value);
+		if (!result) continue;
+		const url = result.url;
+		if (typeof url !== "string" || url.length === 0) continue;
+		const name = result.name ?? result.title;
+		const snippet = result.snippet;
+		const timestamp = result.timestamp;
+		sources.push({
+			title: typeof name === "string" && name.length > 0 ? name : url,
+			url,
+			snippet: typeof snippet === "string" ? snippet : undefined,
+			publishedDate: typeof timestamp === "string" ? timestamp : undefined,
+			ageSeconds: dateToAgeSeconds(typeof timestamp === "string" ? timestamp : undefined),
+		});
+	}
+	return sources;
+}
 export interface PerplexitySearchParams {
 	signal?: AbortSignal;
 	query: string;
@@ -216,7 +328,7 @@ async function findPerplexityAuth(
 	authStorage: AuthStorage,
 	sessionId: string | undefined,
 	signal: AbortSignal | undefined,
-): Promise<PerplexityAuth | null> {
+): Promise<PerplexityAuth> {
 	// 1. PERPLEXITY_COOKIES env var
 	const cookies = $env.PERPLEXITY_COOKIES?.trim();
 	if (cookies) {
@@ -235,7 +347,9 @@ async function findPerplexityAuth(
 	if (apiKey) {
 		return { type: "api_key", token: apiKey };
 	}
-	return null;
+
+	// 4. The consumer ask endpoint currently accepts unauthenticated browser-style requests.
+	return { type: "anonymous" };
 }
 
 /** Call Perplexity API-key endpoint. */
@@ -284,7 +398,7 @@ function buildOAuthSources(event: PerplexityOAuthStreamEvent): SearchSource[] {
 			}));
 	}
 
-	return (event.sources_list ?? [])
+	const sources = (event.sources_list ?? [])
 		.filter(source => typeof source.url === "string" && source.url.length > 0)
 		.map(source => ({
 			title: source.title ?? source.url ?? "",
@@ -293,11 +407,13 @@ function buildOAuthSources(event: PerplexityOAuthStreamEvent): SearchSource[] {
 			publishedDate: source.date,
 			ageSeconds: dateToAgeSeconds(source.date),
 		}));
+	if (sources.length > 0) return sources;
+	return sourcesFromTextPayload(event.text);
 }
 
 function buildOAuthAnswer(event: PerplexityOAuthStreamEvent): string {
 	if (!event.blocks?.length) {
-		return typeof event.text === "string" ? event.text : "";
+		return typeof event.text === "string" ? parseOAuthTextAnswer(event.text) : "";
 	}
 
 	const markdownBlock = event.blocks.find(
@@ -324,51 +440,77 @@ function buildOAuthAnswer(event: PerplexityOAuthStreamEvent): string {
 		}
 	}
 	if (typeof event.text === "string" && event.text.length > 0) {
-		return event.text;
+		return parseOAuthTextAnswer(event.text);
 	}
 	return "";
 }
 
-async function callPerplexityOAuth(
-	auth: { type: "oauth"; token: string } | { type: "cookies"; cookies: string },
+async function callPerplexityAsk(
+	auth: { type: "oauth"; token: string } | { type: "cookies"; cookies: string } | { type: "anonymous" },
 	params: PerplexitySearchParams,
 ): Promise<{ answer: string; sources: SearchSource[]; model?: string; requestId?: string }> {
 	const requestId = crypto.randomUUID();
-	const effectiveQuery = params.system_prompt ? `${params.system_prompt}\n\n${params.query}` : params.query;
+	// The consumer `perplexity_ask` endpoint is itself a research assistant and
+	// has no system-message slot. Prepending the API-style system prompt to the
+	// query makes the model read it as a meta-instruction and refuse with
+	// "I don't have access to web-search tools in this turn", so ask-endpoint
+	// searches send the bare query. (The API-key path still uses system_prompt
+	// as a proper `system` message.)
+	const effectiveQuery = params.query;
+
+	const headers: Record<string, string> = {
+		"Content-Type": "application/json",
+		Accept: "text/event-stream",
+		Origin: "https://www.perplexity.ai",
+		Referer: "https://www.perplexity.ai/",
+		"User-Agent": auth.type === "anonymous" ? ANONYMOUS_USER_AGENT : OAUTH_USER_AGENT,
+		"X-Request-ID": requestId,
+	};
+	if (auth.type === "oauth") {
+		// The ask endpoint authenticates via the next-auth session cookie, NOT a
+		// bearer header — a bearer (even a garbage one) is ignored and the request
+		// silently falls back to the anonymous free `turbo` model regardless of
+		// `model_preference`. The stored OAuth token IS the Perplexity session JWT
+		// (the native app injects the same value as this cookie), so sending it as
+		// the cookie is what unlocks the account's Pro model selection.
+		headers.Cookie = `__Secure-next-auth.session-token=${auth.token}`;
+	} else if (auth.type === "cookies") {
+		headers.Cookie = auth.cookies;
+	}
+	if (auth.type !== "anonymous") {
+		headers["X-App-ApiClient"] = "default";
+		headers["X-App-ApiVersion"] = OAUTH_API_VERSION;
+		headers["X-Perplexity-Request-Reason"] = "submit";
+	}
+
+	const requestParams: Record<string, unknown> = {
+		query_str: effectiveQuery,
+		search_focus: "internet",
+		mode: "copilot",
+		model_preference: "experimental",
+		sources: ["web"],
+		attachments: [],
+		frontend_uuid: crypto.randomUUID(),
+		frontend_context_uuid: crypto.randomUUID(),
+		version: OAUTH_API_VERSION,
+		language: "en-US",
+		timezone: Intl.DateTimeFormat().resolvedOptions().timeZone,
+		search_recency_filter: params.search_recency_filter ?? null,
+		is_incognito: true,
+		use_schematized_api: true,
+		skip_search_enabled: true,
+	};
+	if (auth.type === "anonymous") {
+		requestParams.send_back_text_in_streaming_api = true;
+		requestParams.source = "default";
+	}
 
 	const response = await fetch(PERPLEXITY_OAUTH_ASK_URL, {
 		method: "POST",
-		headers: {
-			...(auth.type === "cookies" ? { Cookie: auth.cookies } : { Authorization: `Bearer ${auth.token}` }),
-			"Content-Type": "application/json",
-			Accept: "text/event-stream",
-			Origin: "https://www.perplexity.ai",
-			Referer: "https://www.perplexity.ai/",
-			"User-Agent": OAUTH_USER_AGENT,
-			"X-App-ApiClient": "default",
-			"X-App-ApiVersion": OAUTH_API_VERSION,
-			"X-Perplexity-Request-Reason": "submit",
-			"X-Request-ID": requestId,
-		},
+		headers,
 		body: JSON.stringify({
 			query_str: effectiveQuery,
-			params: {
-				query_str: effectiveQuery,
-				search_focus: "internet",
-				mode: "copilot",
-				model_preference: "pplx_pro_upgraded",
-				sources: ["web"],
-				attachments: [],
-				frontend_uuid: crypto.randomUUID(),
-				frontend_context_uuid: crypto.randomUUID(),
-				version: OAUTH_API_VERSION,
-				language: "en-US",
-				timezone: Intl.DateTimeFormat().resolvedOptions().timeZone,
-				search_recency_filter: params.search_recency_filter ?? null,
-				is_incognito: true,
-				use_schematized_api: true,
-				skip_search_enabled: true,
-			},
+			params: requestParams,
 		}),
 		signal: withHardTimeout(params.signal),
 	});
@@ -379,13 +521,13 @@ async function callPerplexityOAuth(
 		if (classified) throw classified;
 		throw new SearchProviderError(
 			"perplexity",
-			`Perplexity OAuth API error (${response.status}): ${errorText}`,
+			`Perplexity ask API error (${response.status}): ${errorText}`,
 			response.status,
 		);
 	}
 
 	if (!response.body) {
-		throw new SearchProviderError("perplexity", "Perplexity OAuth API returned no response body", 500);
+		throw new SearchProviderError("perplexity", "Perplexity ask API returned no response body", 500);
 	}
 
 	let answer = "";
@@ -397,7 +539,7 @@ async function callPerplexityOAuth(
 	for await (const event of readSseJson<PerplexityOAuthStreamEvent>(response.body, params.signal)) {
 		if (event.error_code) {
 			const message = event.error_message ?? event.error_code;
-			throw new SearchProviderError("perplexity", `Perplexity OAuth stream error: ${message}`, 400);
+			throw new SearchProviderError("perplexity", `Perplexity ask stream error: ${message}`, 400);
 		}
 
 		mergedEvent = mergeOAuthEventSnapshot(mergedEvent, event);
@@ -500,20 +642,17 @@ function applySourceLimit(result: SearchResponse, limit?: number): SearchRespons
 /** Execute Perplexity web search */
 export async function searchPerplexity(params: PerplexitySearchParams): Promise<SearchResponse> {
 	const auth = await findPerplexityAuth(params.authStorage, params.sessionId, params.signal);
-	if (!auth) {
-		throw new Error("Perplexity auth not found. Set PERPLEXITY_COOKIES, PERPLEXITY_API_KEY, or login via OAuth.");
-	}
 
-	if (auth.type === "oauth" || auth.type === "cookies") {
-		const oauthResult = await callPerplexityOAuth(auth, params);
+	if (auth.type !== "api_key") {
+		const askResult = await callPerplexityAsk(auth, params);
 		return applySourceLimit(
 			{
 				provider: "perplexity",
-				answer: oauthResult.answer || undefined,
-				sources: oauthResult.sources,
-				model: oauthResult.model,
-				requestId: oauthResult.requestId,
-				authMode: "oauth",
+				answer: askResult.answer || undefined,
+				sources: askResult.sources,
+				model: askResult.model,
+				requestId: askResult.requestId,
+				authMode: auth.type === "anonymous" ? "anonymous" : "oauth",
 			},
 			params.num_results,
 		);
@@ -562,6 +701,15 @@ export class PerplexityProvider extends SearchProvider {
 		return !!$env.PERPLEXITY_COOKIES?.trim() || authStorage.hasAuth("perplexity") || !!findApiKey();
 	}
 
+	/**
+	 * Perplexity accepts anonymous browser-style ask requests, but keep auto
+	 * provider selection credential-gated so a configured provider keeps priority
+	 * over the anonymous fallback.
+	 */
+	isExplicitlyAvailable(_authStorage: AuthStorage): boolean {
+		return true;
+	}
+
 	search(params: SearchParams): Promise<SearchResponse> {
 		return searchPerplexity({
 			signal: params.signal,
diff --git a/packages/coding-agent/src/web/search/providers/synthetic.ts b/packages/coding-agent/src/web/search/providers/synthetic.ts
index c531a9a65..12d4a6ca1 100644
--- a/packages/coding-agent/src/web/search/providers/synthetic.ts
+++ b/packages/coding-agent/src/web/search/providers/synthetic.ts
@@ -5,7 +5,7 @@
  * Endpoint: POST https://api.synthetic.new/v2/search
  */
 
-import { type AuthStorage, getEnvApiKey } from "@oh-my-pi/pi-ai";
+import { type ApiKey, type AuthStorage, getEnvApiKey, withAuth } from "@oh-my-pi/pi-ai";
 import type { SearchResponse, SearchSource } from "../../../web/search/types";
 import { SearchProviderError } from "../../../web/search/types";
 import type { SearchParams } from "./base";
@@ -66,12 +66,14 @@ async function callSyntheticSearch(
 
 /** Execute Synthetic web search. */
 export async function searchSynthetic(params: SearchParams): Promise<SearchResponse> {
-	const apiKey = await findApiKey(params.authStorage, params.sessionId, params.signal);
-	if (!apiKey) {
-		throw new Error("Synthetic credentials not found. Set SYNTHETIC_API_KEY or login with 'omp /login synthetic'.");
-	}
+	const keyOrResolver: ApiKey = params.authStorage.resolver("synthetic", {
+		sessionId: params.sessionId,
+	});
 
-	const data = await callSyntheticSearch(apiKey, params.query, params.signal);
+	const data = await withAuth(keyOrResolver, key => callSyntheticSearch(key, params.query, params.signal), {
+		signal: params.signal,
+		missingKeyMessage: "Synthetic credentials not found. Set SYNTHETIC_API_KEY or login with 'omp /login synthetic'.",
+	});
 	const sources: SearchSource[] = [];
 
 	for (const result of data.results ?? []) {
diff --git a/packages/coding-agent/src/web/search/providers/tavily.ts b/packages/coding-agent/src/web/search/providers/tavily.ts
index a32a27856..9b5e0a9f3 100644
--- a/packages/coding-agent/src/web/search/providers/tavily.ts
+++ b/packages/coding-agent/src/web/search/providers/tavily.ts
@@ -4,7 +4,7 @@
  * Uses Tavily's agent-focused search API to return structured results with an
  * optional synthesized answer.
  */
-import { type AuthStorage, getEnvApiKey } from "@oh-my-pi/pi-ai";
+import { type ApiKey, type AuthStorage, getEnvApiKey, withAuth } from "@oh-my-pi/pi-ai";
 import type { SearchResponse, SearchSource } from "../../../web/search/types";
 import { SearchProviderError } from "../../../web/search/types";
 import { clampNumResults, dateToAgeSeconds } from "../utils";
@@ -127,15 +127,16 @@ export async function searchTavily(params: SearchParams): Promise<SearchResponse
 		recency: params.recency,
 		signal: params.signal,
 	};
-	const apiKey = await findApiKey(params.authStorage, params.sessionId, params.signal);
-	if (!apiKey) {
-		throw new Error(
-			'Tavily credentials not found. Set TAVILY_API_KEY or configure an API key for provider "tavily".',
-		);
-	}
+	const keyOrResolver: ApiKey = params.authStorage.resolver("tavily", {
+		sessionId: params.sessionId,
+	});
 
 	const numResults = clampNumResults(tavilyParams.num_results, DEFAULT_NUM_RESULTS, MAX_NUM_RESULTS);
-	const response = await callTavilySearch(apiKey, tavilyParams);
+	const response = await withAuth(keyOrResolver, key => callTavilySearch(key, tavilyParams), {
+		signal: params.signal,
+		missingKeyMessage:
+			'Tavily credentials not found. Set TAVILY_API_KEY or configure an API key for provider "tavily".',
+	});
 	const sources: SearchSource[] = [];
 
 	for (const result of response.results ?? []) {
diff --git a/packages/coding-agent/src/web/search/providers/zai.ts b/packages/coding-agent/src/web/search/providers/zai.ts
index 4795038b8..2f39ff682 100644
--- a/packages/coding-agent/src/web/search/providers/zai.ts
+++ b/packages/coding-agent/src/web/search/providers/zai.ts
@@ -4,7 +4,7 @@
  * Calls Z.AI's remote MCP server (`webSearchPrime`) and adapts results into
  * the unified SearchResponse shape used by the web search tool.
  */
-import { type AuthStorage, getEnvApiKey } from "@oh-my-pi/pi-ai";
+import { type ApiKey, type AuthStorage, getEnvApiKey, withAuth } from "@oh-my-pi/pi-ai";
 import { asRecord, asString } from "../../../web/scrapers/utils";
 import type { SearchResponse, SearchSource } from "../../../web/search/types";
 import { SearchProviderError } from "../../../web/search/types";
@@ -278,12 +278,14 @@ function toSources(results: ZaiSearchResult[]): SearchSource[] {
 
 /** Execute Z.AI web search via remote MCP endpoint. */
 export async function searchZai(params: ZaiSearchParams): Promise<SearchResponse> {
-	const apiKey = await findApiKey(params.authStorage, params.sessionId, params.signal);
-	if (!apiKey) {
-		throw new Error("Z.AI credentials not found. Set ZAI_API_KEY or login with 'omp /login zai'.");
-	}
+	const keyOrResolver: ApiKey = params.authStorage.resolver("zai", {
+		sessionId: params.sessionId,
+	});
 
-	const rawResult = await callZaiSearch(apiKey, params);
+	const rawResult = await withAuth(keyOrResolver, key => callZaiSearch(key, params), {
+		signal: params.signal,
+		missingKeyMessage: "Z.AI credentials not found. Set ZAI_API_KEY or login with 'omp /login zai'.",
+	});
 	const payload = parseSearchPayload(rawResult);
 	let sources = toSources(payload.results);
 
diff --git a/packages/coding-agent/src/web/search/render.ts b/packages/coding-agent/src/web/search/render.ts
index eef69b30d..2e802f606 100644
--- a/packages/coding-agent/src/web/search/render.ts
+++ b/packages/coding-agent/src/web/search/render.ts
@@ -15,23 +15,16 @@ import {
 	formatMoreItems,
 	formatStatusIcon,
 	getDomain,
-	getPreviewLines,
 	PREVIEW_LIMITS,
-	TRUNCATE_LENGTHS,
+	replaceTabs,
 	truncateToWidth,
 } from "../../tools/render-utils";
-import { renderStatusLine, renderTreeList } from "../../tui";
-import { CachedOutputBlock } from "../../tui/output-block";
+import { renderStatusLine, renderTreeList, urlHyperlink } from "../../tui";
+import { CachedOutputBlock, markFramedBlockComponent } from "../../tui/output-block";
 import { getSearchProviderLabel } from "./provider";
 import type { SearchResponse } from "./types";
 
-const MAX_COLLAPSED_ANSWER_LINES = PREVIEW_LIMITS.COLLAPSED_LINES;
-const MAX_SNIPPET_LINES = 2;
-const MAX_SNIPPET_LINE_LEN = TRUNCATE_LENGTHS.LINE;
 const MAX_COLLAPSED_ITEMS = PREVIEW_LIMITS.COLLAPSED_ITEMS;
-const MAX_QUERY_PREVIEW = 2;
-const MAX_QUERY_LEN = 90;
-const MAX_REQUEST_ID_LEN = 36;
 
 function renderFallbackText(contentText: string, expanded: boolean, theme: Theme): Component {
 	const lines = contentText.split("\n").filter(line => line.trim());
@@ -66,6 +59,21 @@ export interface SearchRenderDetails {
 	error?: string;
 }
 
+/** Render a web search failure as a framed error panel, matching the success layout. */
+function renderSearchErrorPanel(message: string, providerLabel: string | undefined, theme: Theme): Component {
+	const header = renderStatusLine({ icon: "error", title: "Web Search", description: providerLabel }, theme);
+	const body = theme.fg("error", `Error: ${replaceTabs(message)}`);
+	const outputBlock = new CachedOutputBlock();
+	return markFramedBlockComponent({
+		render(width: number): string[] {
+			return outputBlock.render({ header, state: "error", sections: [{ lines: [body] }], width }, theme);
+		},
+		invalidate() {
+			outputBlock.invalidate();
+		},
+	});
+}
+
 /** Render web search result with tree-based layout */
 export function renderSearchResult(
 	result: { content: Array<{ type: string; text?: string }>; details?: SearchRenderDetails },
@@ -78,9 +86,12 @@ export function renderSearchResult(
 ): Component {
 	const details = result.details;
 
-	// Handle error case
+	// Handle error case as a framed panel, matching the success layout.
 	if (details?.error) {
-		return new Text(theme.fg("error", `Error: ${details.error}`), 0, 0);
+		const errorProvider = details.response?.provider;
+		const errorProviderLabel =
+			errorProvider && errorProvider !== "none" ? getSearchProviderLabel(errorProvider) : undefined;
+		return renderSearchErrorPanel(details.error, errorProviderLabel, theme);
 	}
 
 	const rawText = result.content?.find(block => block.type === "text")?.text?.trim() ?? "";
@@ -91,8 +102,6 @@ export function renderSearchResult(
 
 	const sources = Array.isArray(response.sources) ? response.sources : [];
 	const sourceCount = sources.length;
-	const citations = Array.isArray(response.citations) ? response.citations : [];
-	const citationCount = citations.length;
 	const searchQueries = Array.isArray(response.searchQueries)
 		? response.searchQueries.filter(item => typeof item === "string")
 		: [];
@@ -118,16 +127,11 @@ export function renderSearchResult(
 		theme,
 	);
 
-	const metaLines: string[] = [];
-	metaLines.push(`${theme.fg("muted", "Provider:")} ${theme.fg("text", providerLabel)}`);
-	if (response.authMode)
-		metaLines.push(
-			`${theme.fg("muted", "Auth:")} ${theme.fg("text", response.authMode === "oauth" ? "OAuth" : response.authMode === "api_key" ? "API key" : response.authMode)}`,
-		);
-	if (response.model) metaLines.push(`${theme.fg("muted", "Model:")} ${theme.fg("text", response.model)}`);
-	metaLines.push(`${theme.fg("muted", "Sources:")} ${theme.fg("text", String(sourceCount))}`);
-	if (citationCount > 0)
-		metaLines.push(`${theme.fg("muted", "Citations:")} ${theme.fg("text", String(citationCount))}`);
+	const authShort =
+		response.authMode === "oauth" ? "OAuth" : response.authMode === "api_key" ? "API" : response.authMode;
+	let providerInfo = response.model ? `${response.model} @ ${providerLabel}` : providerLabel;
+	if (authShort) providerInfo += ` (${authShort})`;
+	const metaLines: string[] = [`${theme.fg("muted", "Provider:")} ${theme.fg("text", providerInfo)}`];
 	if (response.usage) {
 		const usageParts: string[] = [];
 		if (response.usage.inputTokens !== undefined) usageParts.push(`in ${response.usage.inputTokens}`);
@@ -137,22 +141,11 @@ export function renderSearchResult(
 		if (usageParts.length > 0)
 			metaLines.push(`${theme.fg("muted", "Usage:")} ${theme.fg("text", usageParts.join(theme.sep.dot))}`);
 	}
-	if (response.requestId) {
-		metaLines.push(
-			`${theme.fg("muted", "Request:")} ${theme.fg("text", truncateToWidth(response.requestId, MAX_REQUEST_ID_LEN))}`,
-		);
-	}
-	if (searchQueries.length > 0) {
-		const queriesPreview = searchQueries.slice(0, MAX_QUERY_PREVIEW);
-		const queryList = queriesPreview.map(q => truncateToWidth(q, MAX_QUERY_LEN));
-		const suffix = searchQueries.length > queriesPreview.length ? "…" : "";
-		metaLines.push(`${theme.fg("muted", "Queries:")} ${theme.fg("text", queryList.join("; "))}${suffix}`);
-	}
 
 	const answerMarkdown = contentText ? new Markdown(contentText, 0, 0, getMarkdownTheme()) : undefined;
 	const outputBlock = new CachedOutputBlock();
 
-	return {
+	return markFramedBlockComponent({
 		render(width: number): string[] {
 			// Read mutable state at render time
 			const { expanded } = options;
@@ -163,15 +156,15 @@ export function renderSearchResult(
 			let answerLines: string[];
 			if (renderedAnswer.length === 0) {
 				answerLines = [theme.fg("muted", "No answer text returned")];
-			} else if (expanded) {
-				answerLines = renderedAnswer;
-			} else {
-				const collapsedCap = args?.maxAnswerLines ?? MAX_COLLAPSED_ANSWER_LINES;
-				answerLines = renderedAnswer.slice(0, collapsedCap);
+			} else if (args?.maxAnswerLines !== undefined && !expanded) {
+				// CLI compact mode (`omp q`) caps the answer; the TUI passes no cap and shows it in full.
+				answerLines = renderedAnswer.slice(0, args.maxAnswerLines);
 				const remaining = renderedAnswer.length - answerLines.length;
 				if (remaining > 0) {
 					answerLines.push(theme.fg("muted", formatMoreItems(remaining, "line")));
 				}
+			} else {
+				answerLines = renderedAnswer;
 			}
 
 			const sourceTree = renderTreeList(
@@ -187,30 +180,22 @@ export function renderSearchResult(
 								: typeof src.url === "string" && src.url.trim()
 									? src.url
 									: "Untitled";
-						const title = truncateToWidth(titleText, MAX_SNIPPET_LINE_LEN);
 						const url = typeof src.url === "string" ? src.url : "";
 						const domain = url ? getDomain(url) : "";
 						const age =
 							formatAge(src.ageSeconds) || (typeof src.publishedDate === "string" ? src.publishedDate : "");
 						const metaParts: string[] = [];
 						if (domain) metaParts.push(theme.fg("dim", `(${domain})`));
-						if (typeof src.author === "string" && src.author.trim())
-							metaParts.push(theme.fg("muted", truncateToWidth(src.author.trim(), 40)));
 						if (age) metaParts.push(theme.fg("muted", age));
 						const metaSep = theme.fg("dim", theme.sep.dot);
 						const metaSuffix = metaParts.length > 0 ? ` ${metaParts.join(metaSep)}` : "";
-						const srcLines: string[] = [
-							truncateToWidth(`${theme.fg("accent", title)}${metaSuffix}`, MAX_SNIPPET_LINE_LEN),
-						];
-						const snippetText = typeof src.snippet === "string" ? src.snippet : "";
-						if (snippetText.trim()) {
-							const snippetLines = getPreviewLines(snippetText, MAX_SNIPPET_LINES, MAX_SNIPPET_LINE_LEN);
-							for (const snippetLine of snippetLines) {
-								srcLines.push(theme.fg("muted", `${theme.format.dash} ${snippetLine}`));
-							}
-						}
-						if (url) srcLines.push(theme.fg("mdLinkUrl", truncateToWidth(url, MAX_SNIPPET_LINE_LEN)));
-						return srcLines;
+						// One line per source: the title links to its URL, followed by domain · age.
+						// Reserve room for the box borders, the tree branch, and the meta suffix.
+						const lineBudget = Math.max(24, width - 6);
+						const titleBudget = Math.max(12, lineBudget - Bun.stringWidth(metaSuffix));
+						const title = theme.fg("accent", truncateToWidth(titleText, titleBudget));
+						const linkedTitle = url ? urlHyperlink(url, title) : title;
+						return [`${linkedTitle}${metaSuffix}`];
 					},
 				},
 				theme,
@@ -246,7 +231,7 @@ export function renderSearchResult(
 		invalidate() {
 			outputBlock.invalidate();
 		},
-	};
+	});
 }
 
 /** Render web search call (query preview) */
diff --git a/packages/coding-agent/src/web/search/types.ts b/packages/coding-agent/src/web/search/types.ts
index e37e6d961..334971033 100644
--- a/packages/coding-agent/src/web/search/types.ts
+++ b/packages/coding-agent/src/web/search/types.ts
@@ -33,7 +33,11 @@ export const SEARCH_PROVIDER_OPTIONS = [
 		description: "Automatically uses the first configured web-search provider",
 	},
 	{ value: "tavily", label: "Tavily", description: "Requires TAVILY_API_KEY" },
-	{ value: "perplexity", label: "Perplexity", description: "Requires PERPLEXITY_COOKIES or PERPLEXITY_API_KEY" },
+	{
+		value: "perplexity",
+		label: "Perplexity",
+		description: "Uses auth when configured; explicit selection falls back to anonymous search",
+	},
 	{ value: "brave", label: "Brave", description: "Requires BRAVE_API_KEY" },
 	{ value: "jina", label: "Jina", description: "Requires JINA_API_KEY" },
 	{ value: "kimi", label: "Kimi", description: "Requires MOONSHOT_SEARCH_API_KEY or MOONSHOT_API_KEY" },
diff --git a/packages/coding-agent/test/acp-agent.test.ts b/packages/coding-agent/test/acp-agent.test.ts
index 4a22fff8b..597c3c643 100644
--- a/packages/coding-agent/test/acp-agent.test.ts
+++ b/packages/coding-agent/test/acp-agent.test.ts
@@ -615,7 +615,7 @@ describe("ACP agent", () => {
 		await Bun.sleep(0);
 	});
 
-	it("plan-approval standing handler renames the plan and exits plan mode on apply", async () => {
+	it("plan-approval standing handler approves the agent-named plan and exits plan mode on apply", async () => {
 		const harness = await createHarness();
 		Settings.instance.set("plan.enabled", true);
 
@@ -625,7 +625,9 @@ describe("ACP agent", () => {
 
 		const artifactsDir = session.sessionManager.getArtifactsDir();
 		expect(artifactsDir).not.toBeNull();
-		const planPath = path.join(artifactsDir!, "local", "PLAN.md");
+		// The agent writes to its chosen `local://<slug>-plan.md` and resolves with
+		// the matching slug — the file is never renamed.
+		const planPath = path.join(artifactsDir!, "local", "words-counter-plan.md");
 		await Bun.write(planPath, "# Words Counter\n\nFile contents.");
 
 		const updatesBefore = harness.updates.length;
@@ -636,21 +638,20 @@ describe("ACP agent", () => {
 			extra: { title: "words-counter" },
 		})) as {
 			content: Array<{ type: string; text: string }>;
-			details: { sourceToolName: string; sourceResultDetails: { finalPlanFilePath: string; title: string } };
+			details: { sourceToolName: string; sourceResultDetails: { planFilePath: string; title: string } };
 		};
 
 		// Plan-approval payload is shaped for `event-controller` / ACP renderers.
 		expect(result.details.sourceToolName).toBe("plan_approval");
 		expect(result.details.sourceResultDetails.title).toBe("words-counter");
-		expect(result.details.sourceResultDetails.finalPlanFilePath).toBe("local://words-counter.md");
+		expect(result.details.sourceResultDetails.planFilePath).toBe("local://words-counter-plan.md");
 		expect(result.content[0]?.text).toMatch(/Plan approved/);
-		// Plan file is renamed and the source path no longer exists.
-		expect(await Bun.file(path.join(artifactsDir!, "local", "words-counter.md")).exists()).toBe(true);
-		expect(await Bun.file(planPath).exists()).toBe(false);
+		// Plan file keeps its agent-chosen name — no rename.
+		expect(await Bun.file(planPath).exists()).toBe(true);
 		// Mode + handler are cleared; the agent regains write tools next turn.
 		expect(session.planModeState).toBeUndefined();
 		expect(session.standingResolveHandler).toBeUndefined();
-		expect(session.planReferencePath).toBe("local://words-counter.md");
+		expect(session.planReferencePath).toBe("local://words-counter-plan.md");
 		const approvalUpdates = harness.updates.slice(updatesBefore);
 		// Mode-change notifications reached the client so Zed's UI and config
 		// selector both reflect the approval-driven exit.
diff --git a/packages/coding-agent/test/acp-builtins.test.ts b/packages/coding-agent/test/acp-builtins.test.ts
index 40c3f197a..48f2112ed 100644
--- a/packages/coding-agent/test/acp-builtins.test.ts
+++ b/packages/coding-agent/test/acp-builtins.test.ts
@@ -492,20 +492,6 @@ describe("wave 3 commands", () => {
 		expect(output[0]).toContain("Usage: /memory");
 	});
 
-	it("/memory view: outputs memory payload (or empty message)", async () => {
-		const { output, runtime } = createRuntime();
-		const result = await executeAcpBuiltinSlashCommand("/memory view", runtime);
-		expect(result).toEqual({ consumed: true });
-		expect(output.length).toBeGreaterThan(0);
-	});
-
-	it("/memory (no args): defaults to view", async () => {
-		const { output, runtime } = createRuntime();
-		const result = await executeAcpBuiltinSlashCommand("/memory", runtime);
-		expect(result).toEqual({ consumed: true });
-		expect(output.length).toBeGreaterThan(0);
-	});
-
 	// /todo start fuzzy match
 	it("/todo start: finds pending task by substring and starts it", async () => {
 		const { output, session, runtime } = createRuntime();
@@ -667,37 +653,6 @@ describe("wave 4 commands", () => {
 	});
 
 	// /plugins
-	it("/plugins list: outputs without throwing when registries are empty", async () => {
-		const { MarketplaceManager } = await import("../src/extensibility/plugins/marketplace");
-		const { PluginManager } = await import("../src/extensibility/plugins");
-		const listInstalledSpy = spyOn(MarketplaceManager.prototype, "listInstalledPlugins").mockResolvedValue([]);
-		const npmListSpy = spyOn(PluginManager.prototype, "list").mockResolvedValue([]);
-		try {
-			const { output, runtime } = createRuntime();
-			const result = await executeAcpBuiltinSlashCommand("/plugins list", runtime);
-			expect(result).toEqual({ consumed: true });
-			expect(output.length).toBeGreaterThan(0);
-		} finally {
-			listInstalledSpy.mockRestore();
-			npmListSpy.mockRestore();
-		}
-	});
-
-	it("/plugins (no args): defaults to list", async () => {
-		const { MarketplaceManager } = await import("../src/extensibility/plugins/marketplace");
-		const { PluginManager } = await import("../src/extensibility/plugins");
-		const listInstalledSpy = spyOn(MarketplaceManager.prototype, "listInstalledPlugins").mockResolvedValue([]);
-		const npmListSpy = spyOn(PluginManager.prototype, "list").mockResolvedValue([]);
-		try {
-			const { output, runtime } = createRuntime();
-			const result = await executeAcpBuiltinSlashCommand("/plugins", runtime);
-			expect(result).toEqual({ consumed: true });
-			expect(output.length).toBeGreaterThan(0);
-		} finally {
-			listInstalledSpy.mockRestore();
-			npmListSpy.mockRestore();
-		}
-	});
 
 	// /todo start with in_progress status in fuzzy list
 	it("/todo start: resolves ambiguous matches by preferring active tasks", async () => {
diff --git a/packages/coding-agent/test/acp-stdout-hygiene.test.ts b/packages/coding-agent/test/acp-stdout-hygiene.test.ts
index c9e1cc74e..b3345701d 100644
--- a/packages/coding-agent/test/acp-stdout-hygiene.test.ts
+++ b/packages/coding-agent/test/acp-stdout-hygiene.test.ts
@@ -164,7 +164,6 @@ describe("ACP stdout hygiene", () => {
 		proc.stdin.flush();
 
 		const firstLine = await readFirstFrame(proc.stdout);
-		expect(firstLine.length).toBeGreaterThan(0);
 		expect(firstLine[0]).toBe("{");
 
 		const message = JSON.parse(firstLine) as {
diff --git a/packages/coding-agent/test/agent-session-compaction.test.ts b/packages/coding-agent/test/agent-session-compaction.test.ts
index 071139525..4b62a22ea 100644
--- a/packages/coding-agent/test/agent-session-compaction.test.ts
+++ b/packages/coding-agent/test/agent-session-compaction.test.ts
@@ -115,31 +115,6 @@ describe.skipIf(!e2eApiKey("ANTHROPIC_API_KEY"))("AgentSession compaction e2e",
 		expect(firstMsg.role).toBe("compactionSummary");
 	}, 120000);
 
-	it("should maintain valid session state after compaction", async () => {
-		await createSession();
-
-		// Build up history
-		await session.prompt("What is the capital of France? One word answer.");
-		await session.agent.waitForIdle();
-
-		await session.prompt("What is the capital of Germany? One word answer.");
-		await session.agent.waitForIdle();
-
-		// Compact
-		await session.compact();
-
-		// Session should still be usable
-		await session.prompt("What is the capital of Italy? One word answer.");
-		await session.agent.waitForIdle();
-
-		// Should have messages after compaction
-		expect(session.messages.length).toBeGreaterThan(0);
-
-		// The agent should have responded
-		const assistantMessages = session.messages.filter(m => m.role === "assistant");
-		expect(assistantMessages.length).toBeGreaterThan(0);
-	}, 180000);
-
 	it("should persist compaction to session file", async () => {
 		await createSession();
 
@@ -206,9 +181,5 @@ describe.skipIf(!e2eApiKey("ANTHROPIC_API_KEY"))("AgentSession compaction e2e",
 		);
 		// Manual compaction doesn't emit auto_compaction events
 		expect(autoCompactionEvents.length).toBe(0);
-
-		// Regular events should have been emitted
-		const messageEndEvents = events.filter(e => e.type === "message_end");
-		expect(messageEndEvents.length).toBeGreaterThan(0);
 	}, 120000);
 });
diff --git a/packages/coding-agent/test/agent-session-concurrent.test.ts b/packages/coding-agent/test/agent-session-concurrent.test.ts
index 34710c476..9160918af 100644
--- a/packages/coding-agent/test/agent-session-concurrent.test.ts
+++ b/packages/coding-agent/test/agent-session-concurrent.test.ts
@@ -6,6 +6,7 @@ import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
+import { scheduler } from "node:timers/promises";
 import { Agent, AgentBusyError, type AgentTool } from "@oh-my-pi/pi-agent-core";
 import { type AssistantMessage, getBundledModel, type Message, type ToolCall } from "@oh-my-pi/pi-ai";
 import { createMockModel } from "@oh-my-pi/pi-ai/providers/mock";
@@ -25,6 +26,19 @@ import { createAssistantMessage } from "./helpers/agent-session-setup";
 
 // Mock stream that mimics AssistantMessageEventStream
 
+// AgentSession schedules its TTSR retry and context-promotion continuations
+// through `scheduler.wait(delayMs, { signal })` (node:timers/promises), with
+// blind 50ms/100ms "settle" delays. Tests that drive a continuation to
+// completion would otherwise pay that wall-clock time on every run. This spy
+// collapses the blind delay to a single macrotask hop (`scheduler.wait(0)`)
+// while preserving the real abort-signal semantics, so the continuation still
+// fires only after the aborted/overflowed turn has been recorded. Each test
+// that opts in must run inside a block whose afterEach restores mocks.
+const originalSchedulerWait = scheduler.wait.bind(scheduler);
+function collapseSchedulerSettleDelays(): void {
+	vi.spyOn(scheduler, "wait").mockImplementation((_delayMs, options) => originalSchedulerWait(0, options));
+}
+
 describe("AgentSession concurrent prompt guard", () => {
 	let session: AgentSession;
 	let tempDir: string;
@@ -101,7 +115,7 @@ describe("AgentSession concurrent prompt guard", () => {
 		const deadline = Date.now() + timeoutMs;
 		while (Date.now() < deadline) {
 			if (predicate()) return;
-			await Bun.sleep(10);
+			await Bun.sleep(1);
 		}
 
 		throw new Error("Timed out waiting for condition");
@@ -131,7 +145,7 @@ describe("AgentSession concurrent prompt guard", () => {
 		await waitFor(() => session.isStreaming);
 
 		// steer should work while streaming
-		expect(() => session.steer("Steering message")).not.toThrow();
+		await session.steer("Steer while streaming");
 		expect(session.queuedMessageCount).toBe(1);
 
 		// Cleanup
@@ -139,6 +153,79 @@ describe("AgentSession concurrent prompt guard", () => {
 		await firstPrompt.catch(() => {});
 	});
 
+	it("interrupts active work and immediately sends queued steering messages", async () => {
+		const model = getBundledModel("anthropic", "claude-sonnet-4-5")!;
+		const callMessages: Message[][] = [];
+
+		const agent = new Agent({
+			getApiKey: () => "test-key",
+			initialState: {
+				model,
+				systemPrompt: ["Test"],
+				tools: [],
+			},
+			convertToLlm,
+			streamFn: (_model, context, options) => {
+				const callIndex = callMessages.length;
+				callMessages.push([...context.messages]);
+				const stream = new AssistantMessageEventStream();
+				queueMicrotask(() => {
+					stream.push({ type: "start", partial: createAssistantMessage("") });
+					if (callIndex > 0) {
+						stream.push({ type: "done", reason: "stop", message: createAssistantMessage("Handled steer") });
+					}
+				});
+				options?.signal?.addEventListener(
+					"abort",
+					() => {
+						stream.push({
+							type: "error",
+							reason: "aborted",
+							error: createAssistantMessage("Interrupted"),
+						});
+					},
+					{ once: true },
+				);
+				return stream;
+			},
+		});
+
+		const sessionManager = SessionManager.inMemory();
+		const settings = Settings.isolated();
+		const authStorage = await AuthStorage.create(path.join(tempDir, "testauth-interrupt-flush.db"));
+		authStorages.push(authStorage);
+		const modelRegistry = new ModelRegistry(authStorage, path.join(tempDir, "models-interrupt-flush.yml"));
+		authStorage.setRuntimeApiKey("anthropic", "test-key");
+
+		session = new AgentSession({
+			agent,
+			sessionManager,
+			settings,
+			modelRegistry,
+		});
+
+		const firstPrompt = session.prompt("First message").catch(() => {});
+		await waitFor(() => session.isStreaming && callMessages.length === 1);
+
+		await session.steer("Send this now");
+		expect(session.getQueuedMessages().steering).toEqual(["Send this now"]);
+
+		await session.interruptAndFlushQueuedMessages({ reason: "Interrupted by user" });
+		await firstPrompt;
+
+		expect(callMessages).toHaveLength(2);
+		expect(
+			callMessages[1]?.some(message => {
+				if (typeof message.content === "string") {
+					return message.content.includes("Send this now");
+				}
+
+				return message.content.some(content => content.type === "text" && content.text.includes("Send this now"));
+			}),
+		).toBe(true);
+		expect(session.getQueuedMessages()).toEqual({ steering: [], followUp: [] });
+	});
+
 	it("should allow followUp() while streaming", async () => {
 		await createSession();
 
@@ -147,7 +234,7 @@ describe("AgentSession concurrent prompt guard", () => {
 		await waitFor(() => session.isStreaming);
 
 		// followUp should work while streaming
-		expect(() => session.followUp("Follow-up message")).not.toThrow();
+		await session.followUp("Follow-up while streaming");
 		expect(session.queuedMessageCount).toBe(1);
 
 		// Cleanup
@@ -595,13 +682,14 @@ describe("AgentSession TTSR resume gate", () => {
 		if (tempDir && fs.existsSync(tempDir)) {
 			fs.rmSync(tempDir, { recursive: true });
 		}
+		vi.restoreAllMocks();
 	});
 
 	async function waitFor(predicate: () => boolean, timeoutMs = 500): Promise<void> {
 		const deadline = Date.now() + timeoutMs;
 		while (Date.now() < deadline) {
 			if (predicate()) return;
-			await Bun.sleep(10);
+			await Bun.sleep(1);
 		}
 
 		throw new Error("Timed out waiting for condition");
@@ -674,6 +762,7 @@ describe("AgentSession TTSR resume gate", () => {
 	}
 
 	it("prompt() blocks until TTSR interrupt continuation completes", async () => {
+		collapseSchedulerSettleDelays();
 		const model = getBundledModel("anthropic", "claude-sonnet-4-5")!;
 		let streamCallCount = 0;
 		let continuationCompleted = false;
@@ -733,7 +822,114 @@ describe("AgentSession TTSR resume gate", () => {
 		expect(session.isStreaming).toBe(false);
 	});
 
+	it("labels aborted tool placeholders with the TTSR rule reason", async () => {
+		collapseSchedulerSettleDelays();
+		const model = getBundledModel("anthropic", "claude-sonnet-4-5")!;
+		let streamCallCount = 0;
+
+		const ttsrManager = new TtsrManager({
+			enabled: true,
+			contextMode: "discard",
+			interruptMode: "always",
+			repeatMode: "once",
+			repeatGap: 10,
+		});
+		ttsrManager.addRule(testRule);
+
+		const toolCallContent: ToolCall = {
+			type: "toolCall",
+			id: "call_ttsr_abort_reason",
+			name: "mock_edit",
+			arguments: { snippet: "let val = result.unwrap(" },
+		};
+
+		const makeToolCallMsg = (stopReason: "toolUse" | "aborted" = "toolUse"): AssistantMessage => ({
+			role: "assistant",
+			content: [toolCallContent],
+			api: "anthropic-messages",
+			provider: "anthropic",
+			model: "mock",
+			usage: {
+				input: 0,
+				output: 0,
+				cacheRead: 0,
+				cacheWrite: 0,
+				totalTokens: 0,
+				cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+			},
+			stopReason,
+			timestamp: Date.now(),
+		});
+
+		const agent = new Agent({
+			getApiKey: () => "test-key",
+			initialState: { model, systemPrompt: ["Test"], tools: [] },
+			streamFn: (_model, _context, options) => {
+				streamCallCount++;
+				const stream = new AssistantMessageEventStream();
+				const signal = options?.signal;
+				if (streamCallCount === 1) {
+					queueMicrotask(() => {
+						const partial = makeToolCallMsg();
+						if (signal) {
+							signal.addEventListener(
+								"abort",
+								() => {
+									stream.push({
+										type: "error",
+										reason: "aborted",
+										error: makeToolCallMsg("aborted"),
+									});
+								},
+								{ once: true },
+							);
+						}
+						stream.push({ type: "start", partial });
+						stream.push({ type: "toolcall_start", contentIndex: 0, partial });
+						stream.push({
+							type: "toolcall_delta",
+							contentIndex: 0,
+							delta: 'let val = result.unwrap("oops")',
+							partial,
+						});
+					});
+				} else {
+					pushContinuationStream(stream, () => {});
+				}
+				return stream;
+			},
+		});
+
+		const sessionManager = SessionManager.inMemory();
+		const settings = Settings.isolated();
+		const authStorage = await AuthStorage.create(path.join(tempDir, "testauth-abort-reason.db"));
+		authStorages.push(authStorage);
+		const modelRegistry = new ModelRegistry(authStorage, path.join(tempDir, "models.yml"));
+		authStorage.setRuntimeApiKey("anthropic", "test-key");
+		session = new AgentSession({ agent, sessionManager, settings, modelRegistry, ttsrManager });
+
+		await session.prompt("Write some Rust code");
+
+		const toolResult = sessionManager
+			.getEntries()
+			.find(
+				entry =>
+					entry.type === "message" &&
+					entry.message.role === "toolResult" &&
+					entry.message.toolCallId === toolCallContent.id,
+			);
+		expect(toolResult?.type).toBe("message");
+		const text =
+			toolResult?.type === "message" && toolResult.message.role === "toolResult"
+				? (toolResult.message.content.find((part): part is { type: "text"; text: string } => part.type === "text")
+						?.text ?? "")
+				: "";
+		expect(text).toContain("Tool execution was aborted: TTSR matched rule: no-unwrap");
+		expect(text).not.toContain("Request was aborted");
+	});
+
 	it("relativizes the rule file path in the TTSR interrupt injection (no absolute leak)", async () => {
+		collapseSchedulerSettleDelays();
 		const model = getBundledModel("anthropic", "claude-sonnet-4-5")!;
 		let streamCallCount = 0;
 
@@ -946,6 +1142,7 @@ describe("AgentSession TTSR resume gate", () => {
 	});
 
 	it("prompt() waits for TTSR continuation with tool calls to finish", async () => {
+		collapseSchedulerSettleDelays();
 		const model = getBundledModel("anthropic", "claude-sonnet-4-5")!;
 		let streamCallCount = 0;
 		let toolExecutionFinished = false;
@@ -1306,6 +1503,7 @@ describe("AgentSession TTSR resume gate", () => {
 	});
 
 	it("prompt() waits for context-promotion continuation to finish", async () => {
+		collapseSchedulerSettleDelays();
 		const authStorage = await AuthStorage.create(path.join(tempDir, "testauth-promo.db"));
 		authStorages.push(authStorage);
 		authStorage.setRuntimeApiKey("openai-codex", "test-key");
diff --git a/packages/coding-agent/test/agent-session-context-promotion.test.ts b/packages/coding-agent/test/agent-session-context-promotion.test.ts
index dcbe8af92..06347689b 100644
--- a/packages/coding-agent/test/agent-session-context-promotion.test.ts
+++ b/packages/coding-agent/test/agent-session-context-promotion.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
+import { afterAll, afterEach, beforeAll, describe, expect, it, vi } from "bun:test";
 import * as path from "node:path";
 import { Agent } from "@oh-my-pi/pi-agent-core";
 import type { AssistantMessage, Model, ProviderSessionState } from "@oh-my-pi/pi-ai";
@@ -15,19 +15,26 @@ describe("AgentSession context promotion", () => {
 	let modelRegistry: ModelRegistry;
 	let authStorage: AuthStorage;
 
-	beforeEach(async () => {
+	beforeAll(async () => {
+		// ModelRegistry eagerly loads the immutable bundled model catalog in its
+		// constructor (~100ms). The catalog and auth fixture never change between
+		// tests here (tests only read models and add benign extra runtime keys),
+		// so build them once instead of paying ~950ms across the 9 cases.
 		tempDir = TempDir.createSync("@pi-context-promotion-");
 		authStorage = await AuthStorage.create(path.join(tempDir.path(), "testauth.db"));
 		authStorage.setRuntimeApiKey("openai-codex", "test-key");
 		modelRegistry = new ModelRegistry(authStorage);
 	});
 
+	afterAll(() => {
+		authStorage.close();
+		tempDir.removeSync();
+	});
+
 	afterEach(async () => {
 		if (session) {
 			await session.dispose();
 		}
-		authStorage.close();
-		tempDir.removeSync();
 	});
 
 	function createOverflowMessage(
@@ -113,6 +120,17 @@ describe("AgentSession context promotion", () => {
 		throw new Error("Timed out waiting for condition");
 	}
 
+	// Deterministically drain the fire-and-forget `agent_end` handler that
+	// `emitExternalEvent` dispatches. The handler's terminal maintenance work
+	// (`#checkCompaction`) is microtask-based on the no-promotion paths, so a
+	// single macrotask turn fully flushes it; `waitForIdle` then settles any
+	// tracked continuation. Used by the negative tests, which assert that *no*
+	// promotion happened and therefore need the handler to have actually run.
+	async function settle(): Promise<void> {
+		await new Promise(resolve => setTimeout(resolve, 0));
+		await session.waitForIdle();
+	}
+
 	it("promotes to a larger-context model on overflow and clears codex websocket session state", async () => {
 		const sparkModel = modelRegistry.find("openai-codex", "gpt-5.3-codex-spark");
 		const codexModel = modelRegistry.find("openai-codex", "gpt-5.5");
@@ -390,7 +408,7 @@ describe("AgentSession context promotion", () => {
 		session.agent.emitExternalEvent({ type: "message_end", message: overflowMessage });
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [overflowMessage] });
 
-		await Bun.sleep(30);
+		await settle();
 
 		expect(session.model?.provider).toBe(sparkModel.provider);
 		expect(session.model?.id).toBe(sparkModel.id);
@@ -472,7 +490,7 @@ describe("AgentSession context promotion", () => {
 		session.agent.emitExternalEvent({ type: "message_end", message: staleIncomplete });
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [staleIncomplete] });
 
-		await Bun.sleep(30);
+		await settle();
 
 		expect(session.model?.provider).toBe(codexModel.provider);
 		expect(session.model?.id).toBe(codexModel.id);
diff --git a/packages/coding-agent/test/agent-session-fresh.test.ts b/packages/coding-agent/test/agent-session-fresh.test.ts
new file mode 100644
index 000000000..aa294d743
--- /dev/null
+++ b/packages/coding-agent/test/agent-session-fresh.test.ts
@@ -0,0 +1,101 @@
+import { afterEach, describe, expect, it } from "bun:test";
+import * as path from "node:path";
+import { Agent, AppendOnlyContextManager } from "@oh-my-pi/pi-agent-core";
+import type { ProviderSessionState } from "@oh-my-pi/pi-ai";
+import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
+import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import { AgentSession } from "@oh-my-pi/pi-coding-agent/session/agent-session";
+import { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
+import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
+import { TempDir } from "@oh-my-pi/pi-utils";
+
+interface FreshHarness {
+	agent: Agent;
+	session: AgentSession;
+	sessionManager: SessionManager;
+}
+
+const cleanup: Array<() => Promise<void>> = [];
+
+afterEach(async () => {
+	while (cleanup.length > 0) {
+		const run = cleanup.pop();
+		if (run) await run();
+	}
+});
+
+async function createFreshHarness(): Promise<FreshHarness> {
+	const tempDir = TempDir.createSync("@pi-agent-session-fresh-");
+	const authStorage = await AuthStorage.create(path.join(tempDir.path(), "auth.db"));
+	const modelRegistry = new ModelRegistry(authStorage, path.join(tempDir.path(), "models.yml"));
+	const sessionManager = SessionManager.create(tempDir.path(), path.join(tempDir.path(), "sessions"));
+	const agent = new Agent({
+		initialState: {
+			systemPrompt: ["Test"],
+			tools: [],
+			messages: [],
+		},
+	});
+	const session = new AgentSession({
+		agent,
+		sessionManager,
+		settings: Settings.isolated(),
+		modelRegistry,
+	});
+	cleanup.push(async () => {
+		await session.dispose();
+		authStorage.close();
+		tempDir.removeSync();
+	});
+	return { agent, session, sessionManager };
+}
+
+describe("AgentSession fresh provider state", () => {
+	it("rotates only the provider-facing session id and prunes cached stream state", async () => {
+		const { agent, session, sessionManager } = await createFreshHarness();
+		const persistedSessionId = sessionManager.getSessionId();
+		const persistedSessionFile = sessionManager.getSessionFile();
+		const persistedHeaderId = sessionManager.getHeader()?.id;
+		let closeCount = 0;
+		const providerState: ProviderSessionState = {
+			close() {
+				closeCount += 1;
+			},
+		};
+		session.providerSessionState.set("websocket", providerState);
+
+		const appendOnlyContext = new AppendOnlyContextManager();
+		agent.setAppendOnlyContext(appendOnlyContext);
+		appendOnlyContext.syncMessages([{ role: "user", content: "cached context" }]);
+		appendOnlyContext.build({ systemPrompt: ["Test"], messages: [], tools: [] }, { intentTracing: false });
+
+		const result = session.freshSession();
+
+		expect(result).toBeDefined();
+		if (!result) return;
+		expect(result.previousSessionId).toBe(persistedSessionId);
+		expect(result.sessionId).not.toBe(persistedSessionId);
+		expect(result.closedProviderSessions).toBe(1);
+		expect(agent.sessionId).toBe(result.sessionId);
+		expect(session.sessionId).toBe(result.sessionId);
+		expect(sessionManager.getSessionId()).toBe(persistedSessionId);
+		expect(sessionManager.getHeader()?.id).toBe(persistedHeaderId);
+		expect(sessionManager.getSessionFile()).toBe(persistedSessionFile);
+		expect(closeCount).toBe(1);
+		expect(session.providerSessionState.size).toBe(0);
+		expect(appendOnlyContext.log.length).toBe(0);
+		expect(appendOnlyContext.prefix.built).toBe(false);
+	});
+
+	it("drops the transient provider id when a real new session starts", async () => {
+		const { session, sessionManager } = await createFreshHarness();
+		const freshResult = session.freshSession();
+		expect(freshResult).toBeDefined();
+		if (!freshResult) return;
+
+		await session.newSession();
+
+		expect(session.sessionId).toBe(sessionManager.getSessionId());
+		expect(session.sessionId).not.toBe(freshResult.sessionId);
+	});
+});
diff --git a/packages/coding-agent/test/agent-session-handoff.test.ts b/packages/coding-agent/test/agent-session-handoff.test.ts
index e90506b33..3ecb612aa 100644
--- a/packages/coding-agent/test/agent-session-handoff.test.ts
+++ b/packages/coding-agent/test/agent-session-handoff.test.ts
@@ -1,8 +1,8 @@
-import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
 import * as path from "node:path";
 import { Agent } from "@oh-my-pi/pi-agent-core";
 import * as compactionModule from "@oh-my-pi/pi-agent-core/compaction";
-import type { AssistantMessage, ToolCall } from "@oh-my-pi/pi-ai";
+import type { AssistantMessage, Model, ToolCall } from "@oh-my-pi/pi-ai";
 import { getBundledModel } from "@oh-my-pi/pi-ai/models";
 import { createMockModel } from "@oh-my-pi/pi-ai/providers/mock";
 import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
@@ -14,26 +14,67 @@ import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manage
 import { TempDir } from "@oh-my-pi/pi-utils";
 
 describe("AgentSession handoff", () => {
+	// Immutable across the whole file: the model registry's synchronous bundled-model
+	// load dominates per-test setup (~100ms each), and the auth store + bundled model
+	// never change. Build them once. Per-test mutable state (session, session file,
+	// emitted events) is rebuilt in beforeEach.
+	let sharedDir: TempDir;
+	let authStorage: AuthStorage;
+	let modelRegistry: ModelRegistry;
+	let model: Model;
+
 	let tempDir: TempDir;
 	let session: AgentSession;
 	let sessionManager: SessionManager;
-	let authStorage: AuthStorage;
-	let modelRegistry: ModelRegistry;
 	let events: AgentSessionEvent[];
 
+	/** Poll `predicate` until it holds (returns as soon as the state is reached) or the
+	 *  deadline elapses. Replaces blind settle sleeps for tests with a positive signal. */
+	async function waitFor(predicate: () => boolean, timeoutMs = 1_000): Promise<void> {
+		const deadline = Date.now() + timeoutMs;
+		while (!predicate()) {
+			if (Date.now() >= deadline) {
+				throw new Error("Timed out waiting for condition");
+			}
+			await Bun.sleep(1);
+		}
+	}
+
+	/** Drain post-turn maintenance deterministically for negative tests (those proving
+	 *  maintenance did NOT run, where there is no positive signal to poll on). Post-turn
+	 *  work is scheduled fire-and-forget: a single event-loop turn lets the handler run to
+	 *  its decision and register any compaction pass as a tracked post-prompt task, then
+	 *  `waitForIdle()` drains that task to completion. */
+	async function drainMaintenance(): Promise<void> {
+		await Bun.sleep(0);
+		await session.waitForIdle();
+	}
+
+	beforeAll(async () => {
+		sharedDir = TempDir.createSync("@pi-handoff-shared-");
+		authStorage = await AuthStorage.create(path.join(sharedDir.path(), "testauth.db"));
+		authStorage.setRuntimeApiKey("anthropic", "test-key");
+		modelRegistry = new ModelRegistry(authStorage);
+
+		const bundled = getBundledModel("anthropic", "claude-sonnet-4-5");
+		if (!bundled) {
+			throw new Error("Expected built-in anthropic model to exist");
+		}
+		model = bundled;
+	});
+
+	afterAll(async () => {
+		authStorage.close();
+		try {
+			await sharedDir.remove();
+		} catch {}
+	});
+
 	beforeEach(async () => {
 		tempDir = TempDir.createSync("@pi-handoff-");
-		authStorage = await AuthStorage.create(path.join(tempDir.path(), "testauth.db"));
-		authStorage.setRuntimeApiKey("anthropic", "test-key");
-		modelRegistry = new ModelRegistry(authStorage);
 		sessionManager = SessionManager.create(tempDir.path(), tempDir.path());
 		events = [];
 
-		const model = getBundledModel("anthropic", "claude-sonnet-4-5");
-		if (!model) {
-			throw new Error("Expected built-in anthropic model to exist");
-		}
-
 		const agent = new Agent({
 			initialState: {
 				model,
@@ -85,7 +126,6 @@ describe("AgentSession handoff", () => {
 		if (session) {
 			await session.dispose();
 		}
-		authStorage.close();
 		try {
 			await tempDir.remove();
 		} catch {}
@@ -97,7 +137,7 @@ describe("AgentSession handoff", () => {
 		const generateHandoffSpy = vi.spyOn(compactionModule, "generateHandoff").mockResolvedValue(handoffText);
 
 		const result = await session.handoff();
-		await Bun.sleep(20);
+		await drainMaintenance();
 
 		expect(generateHandoffSpy).toHaveBeenCalledTimes(1);
 		expect(result?.document).toBe(handoffText);
@@ -123,7 +163,11 @@ describe("AgentSession handoff", () => {
 		});
 
 		await session.prompt("pending prompt ".repeat(120));
-		await Bun.sleep(20);
+		await waitFor(
+			() =>
+				compactSpy.mock.calls.length === 1 &&
+				events.some(event => event.type === "auto_compaction_end" && event.aborted === false),
+		);
 
 		expect(compactSpy).toHaveBeenCalledTimes(1);
 		expect(promptSpy).toHaveBeenCalledTimes(1);
@@ -177,7 +221,7 @@ describe("AgentSession handoff", () => {
 			isError: false,
 		});
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [assistantMessage] });
-		await Bun.sleep(20);
+		await drainMaintenance();
 
 		expect(handoffSpy).not.toHaveBeenCalled();
 		expect(events.filter(event => event.type === "auto_compaction_start")).toHaveLength(0);
@@ -259,7 +303,7 @@ describe("AgentSession handoff", () => {
 		const handoffSpy = vi.spyOn(session, "handoff");
 		session.agent.emitExternalEvent({ type: "message_end", message: assistantMessage });
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [assistantMessage] });
-		await Bun.sleep(20);
+		await drainMaintenance();
 
 		expect(handoffSpy).not.toHaveBeenCalled();
 		expect(events.filter(event => event.type === "auto_compaction_start")).toHaveLength(0);
@@ -307,7 +351,7 @@ describe("AgentSession handoff", () => {
 
 		session.agent.emitExternalEvent({ type: "message_end", message: overflowAssistant });
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [overflowAssistant] });
-		await Bun.sleep(20);
+		await waitFor(() => events.filter(event => event.type === "auto_compaction_end").length === 1);
 
 		expect(handoffSpy).not.toHaveBeenCalled();
 		const startEvents = events.filter(event => event.type === "auto_compaction_start");
@@ -352,7 +396,11 @@ describe("AgentSession handoff", () => {
 
 		session.agent.emitExternalEvent({ type: "message_end", message: assistantMessage });
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [assistantMessage] });
-		await Bun.sleep(20);
+		await waitFor(
+			() =>
+				handoffSpy.mock.calls.length === 1 &&
+				events.filter(event => event.type === "auto_compaction_end").length === 1,
+		);
 
 		expect(handoffSpy).toHaveBeenCalledTimes(1);
 		expect(handoffSpy).toHaveBeenCalledWith(expect.stringContaining("Threshold-triggered maintenance"), {
@@ -500,7 +548,8 @@ describe("AgentSession handoff", () => {
 
 		session.agent.emitExternalEvent({ type: "message_end", message: assistantMessage });
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [assistantMessage] });
-		await Bun.sleep(20);
+		await waitFor(() => handoffSpy.mock.calls.length === 1);
+		await session.waitForIdle();
 
 		expect(handoffSpy).toHaveBeenCalledTimes(1);
 		// The bug surfaced as agent.continue() racing the deferred handoff. With the fix,
@@ -554,7 +603,7 @@ describe("AgentSession handoff", () => {
 		session.agent.emitExternalEvent({ type: "message_end", message: assistantMessage });
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [assistantMessage] });
 		// Let the deferred handoff post-prompt task enter the generateHandoff await.
-		await Bun.sleep(20);
+		await waitFor(() => session.isGeneratingHandoff);
 		expect(generateHandoffSpy).toHaveBeenCalledTimes(1);
 		expect(session.isGeneratingHandoff).toBe(true);
 
@@ -601,7 +650,7 @@ describe("AgentSession handoff", () => {
 
 		session.agent.emitExternalEvent({ type: "message_end", message: assistantMessage });
 		session.agent.emitExternalEvent({ type: "agent_end", messages: [assistantMessage] });
-		await Bun.sleep(20);
+		await waitFor(() => events.filter(event => event.type === "auto_compaction_end").length === 1);
 
 		expect(handoffSpy).toHaveBeenCalledTimes(1);
 		const endEvents = events.filter(event => event.type === "auto_compaction_end");
diff --git a/packages/coding-agent/test/agent-session-model-persistence.test.ts b/packages/coding-agent/test/agent-session-model-persistence.test.ts
index 7b741027a..76f089355 100644
--- a/packages/coding-agent/test/agent-session-model-persistence.test.ts
+++ b/packages/coding-agent/test/agent-session-model-persistence.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, beforeEach, describe, expect, it } from "bun:test";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it } from "bun:test";
 import * as path from "node:path";
 import { Agent } from "@oh-my-pi/pi-agent-core";
 import { type Api, Effort, getBundledModel, type Model } from "@oh-my-pi/pi-ai";
@@ -18,7 +18,24 @@ describe("AgentSession model persistence", () => {
 	let tempDir: TempDir;
 	let session: AgentSession | undefined;
 	let sessionSettings: Settings;
-	const authStorages: AuthStorage[] = [];
+	// Auth storage (SQLite DB) and the model registry are immutable across these tests:
+	// every test sets the same anthropic runtime key and only ever reads the bundled model
+	// list. Building them once avoids ~12 SQLite opens + registry constructions.
+	let sharedDir: TempDir;
+	let sharedAuthStorage: AuthStorage;
+	let sharedModelRegistry: ModelRegistry;
+
+	beforeAll(async () => {
+		sharedDir = TempDir.createSync("@pi-model-persistence-shared-");
+		sharedAuthStorage = await AuthStorage.create(path.join(sharedDir.path(), "auth.db"));
+		sharedAuthStorage.setRuntimeApiKey("anthropic", "test-key");
+		sharedModelRegistry = new ModelRegistry(sharedAuthStorage, path.join(sharedDir.path(), "models.yml"));
+	});
+
+	afterAll(() => {
+		sharedAuthStorage.close();
+		sharedDir.removeSync();
+	});
 
 	beforeEach(() => {
 		tempDir = TempDir.createSync("@pi-model-persistence-");
@@ -29,9 +46,6 @@ describe("AgentSession model persistence", () => {
 			await session.dispose();
 			session = undefined;
 		}
-		for (const authStorage of authStorages.splice(0)) {
-			authStorage.close();
-		}
 		tempDir.removeSync();
 	});
 
@@ -84,13 +98,7 @@ describe("AgentSession model persistence", () => {
 		modelRoles?: Record<string, string>;
 		persist?: boolean;
 	}): Promise<{ modelRegistry: ModelRegistry; settings: Settings; session: AgentSession }> {
-		const authStorage = await AuthStorage.create(path.join(tempDir.path(), `testauth-${authStorages.length}.db`));
-		authStorages.push(authStorage);
-		authStorage.setRuntimeApiKey("anthropic", "test-key");
-		const modelRegistry = new ModelRegistry(
-			authStorage,
-			path.join(tempDir.path(), `models-${authStorages.length}.yml`),
-		);
+		const modelRegistry = sharedModelRegistry;
 		const model =
 			options?.initialModel ??
 			options?.selectInitialModel?.(modelRegistry.getAvailable()) ??
@@ -118,7 +126,7 @@ describe("AgentSession model persistence", () => {
 		session = new AgentSession({
 			agent,
 			sessionManager: options?.persist
-				? SessionManager.create(tempDir.path(), path.join(tempDir.path(), `active-${authStorages.length}`))
+				? SessionManager.create(tempDir.path(), path.join(tempDir.path(), "active"))
 				: SessionManager.inMemory(),
 			settings: sessionSettings,
 			modelRegistry,
@@ -131,19 +139,12 @@ describe("AgentSession model persistence", () => {
 		targetSessionFile: string,
 		settings: Settings = Settings.isolated(),
 	): Promise<CreateAgentSessionResult> {
-		const authStorage = await AuthStorage.create(path.join(tempDir.path(), `testauth-${authStorages.length}.db`));
-		authStorages.push(authStorage);
-		authStorage.setRuntimeApiKey("anthropic", "test-key");
-		const modelRegistry = new ModelRegistry(
-			authStorage,
-			path.join(tempDir.path(), `models-${authStorages.length}.yml`),
-		);
 		const sessionManager = await SessionManager.open(targetSessionFile, path.join(tempDir.path(), "startup"));
 		const result = await createAgentSession({
 			cwd: tempDir.path(),
 			agentDir: tempDir.path(),
-			authStorage,
-			modelRegistry,
+			authStorage: sharedAuthStorage,
+			modelRegistry: sharedModelRegistry,
 			sessionManager,
 			settings,
 			disableExtensionDiscovery: true,
diff --git a/packages/coding-agent/test/agent-session-new-session-todos.test.ts b/packages/coding-agent/test/agent-session-new-session-todos.test.ts
deleted file mode 100644
index 7ff94c906..000000000
--- a/packages/coding-agent/test/agent-session-new-session-todos.test.ts
+++ /dev/null
@@ -1,126 +0,0 @@
-import { afterEach, beforeEach, describe, expect, it } from "bun:test";
-import * as fs from "node:fs";
-import * as os from "node:os";
-import * as path from "node:path";
-import { Agent } from "@oh-my-pi/pi-agent-core";
-import { getBundledModel } from "@oh-my-pi/pi-ai";
-import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
-import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
-import { AgentSession } from "@oh-my-pi/pi-coding-agent/session/agent-session";
-import { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
-import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
-import type { ToolSession } from "@oh-my-pi/pi-coding-agent/tools";
-import { TodoTool } from "@oh-my-pi/pi-coding-agent/tools";
-import { Snowflake } from "@oh-my-pi/pi-utils";
-
-/**
- * Regression test: /new (AgentSession.newSession) must fully switch to a new session file
- * before the call resolves.
- *
- * If it doesn't, UI code that reloads todos immediately after /new will read the old
- * session artifact dir and keep showing stale todos.
- */
-describe("AgentSession newSession clears todo artifacts", () => {
-	let tempDir: string;
-	let session: AgentSession;
-	let sessionManager: SessionManager;
-	let authStorage: AuthStorage | undefined;
-
-	beforeEach(async () => {
-		tempDir = path.join(os.tmpdir(), `pi-new-session-todos-test-${Snowflake.next()}`);
-		fs.mkdirSync(tempDir, { recursive: true });
-
-		sessionManager = SessionManager.create(tempDir, tempDir);
-		const settings = Settings.isolated();
-		authStorage = await AuthStorage.create(path.join(tempDir, "testauth.db"));
-		const modelRegistry = new ModelRegistry(authStorage, path.join(tempDir, "models.yml"));
-
-		const model = getBundledModel("anthropic", "claude-sonnet-4-5");
-		if (!model) {
-			throw new Error("Test model not found in registry");
-		}
-
-		const toolSession: ToolSession = {
-			cwd: tempDir,
-			hasUI: false,
-			getSessionFile: () => sessionManager.getSessionFile() ?? null,
-			getSessionSpawns: () => "*",
-			settings,
-		};
-
-		const agent = new Agent({
-			getApiKey: () => "test",
-			initialState: {
-				model,
-				systemPrompt: ["test"],
-				tools: [new TodoTool(toolSession)],
-			},
-		});
-
-		session = new AgentSession({
-			agent,
-			sessionManager,
-			settings,
-			modelRegistry,
-		});
-
-		// Must subscribe to enable session persistence hooks
-		session.subscribe(() => {});
-	});
-
-	afterEach(async () => {
-		if (session) {
-			await session.dispose();
-		}
-		authStorage?.close();
-		authStorage = undefined;
-		if (tempDir && fs.existsSync(tempDir)) {
-			fs.rmSync(tempDir, { recursive: true });
-		}
-	});
-
-	it("should not carry over todo state to the new session branch", async () => {
-		const oldSessionFile = session.sessionFile;
-		expect(oldSessionFile).toBeDefined();
-
-		session.setTodoPhases([
-			{
-				name: "Tasks",
-				tasks: [{ content: "do the thing", status: "pending" }],
-			},
-		]);
-		expect(session.getTodoPhases()).toHaveLength(1);
-		expect(session.getTodoPhases()[0]?.tasks).toHaveLength(1);
-		await session.newSession();
-
-		const newSessionFile = session.sessionFile;
-		expect(newSessionFile).toBeDefined();
-		expect(newSessionFile).not.toBe(oldSessionFile);
-
-		expect(session.getTodoPhases()).toHaveLength(0);
-	});
-
-	it("should clear stale todo cache when branching from the first user message", async () => {
-		sessionManager.appendMessage({
-			role: "user",
-			content: "start task",
-			timestamp: Date.now(),
-		});
-
-		const branchCandidates = session.getUserMessagesForBranching();
-		expect(branchCandidates).toHaveLength(1);
-
-		session.setTodoPhases([
-			{
-				name: "Execution",
-				tasks: [{ content: "stale from old branch", status: "in_progress" }],
-			},
-		]);
-		expect(session.getTodoPhases()).toHaveLength(1);
-
-		const result = await session.branch(branchCandidates[0].entryId);
-		expect(result.cancelled).toBe(false);
-		expect(result.selectedText).toBe("start task");
-		expect(session.getTodoPhases()).toHaveLength(0);
-	});
-});
diff --git a/packages/coding-agent/test/agent-session-openai-responses-replay.test.ts b/packages/coding-agent/test/agent-session-openai-responses-replay.test.ts
index 7c5505e33..fd5452369 100644
--- a/packages/coding-agent/test/agent-session-openai-responses-replay.test.ts
+++ b/packages/coding-agent/test/agent-session-openai-responses-replay.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, describe, expect, it, vi } from "bun:test";
+import { afterAll, afterEach, beforeAll, describe, expect, it, vi } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
@@ -12,8 +12,11 @@ import type {
 	Usage,
 } from "@oh-my-pi/pi-ai/types";
 import { createOpenAIResponsesHistoryPayload } from "@oh-my-pi/pi-ai/utils";
+import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
+import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import { createAgentSession } from "@oh-my-pi/pi-coding-agent/sdk";
 import type { AgentSession } from "@oh-my-pi/pi-coding-agent/session/agent-session";
-import type { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
+import { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
 import {
 	type SessionEntry,
 	SessionManager,
@@ -209,20 +212,21 @@ async function createPersistedSession(
 	return { sessionFile, treeTargetId: result?.treeTargetId };
 }
 
+// ModelRegistry construction loads the bundled model catalog plus the on-disk
+// cache (~100ms) and dominated this file's runtime when rebuilt once per test.
+// The registry and its pinned AuthStorage are immutable across these tests (none
+// mutate the catalog or stored credentials), so a single instance is shared via
+// createAgentSession's `modelRegistry`/`authStorage` seam. The registry pins
+// itself to its AuthStorage, so both MUST be the same shared instances.
+let sharedModelRegistry: ModelRegistry;
+let sharedRegistryDir: string;
+
 async function createSessionHarness(
 	tempDir: string,
 	sessionManager: SessionManager,
 	options: { provider?: Parameters<typeof getBundledModel>[0]; modelId?: string } = {},
-): Promise<{ session: AgentSession; authStorage: AuthStorage }> {
+): Promise<{ session: AgentSession }> {
 	const { provider = "openai", modelId = "gpt-5-mini" } = options;
-	const [{ createAgentSession }, { Settings }, { AuthStorage }] = await Promise.all([
-		import("@oh-my-pi/pi-coding-agent/sdk"),
-		import("@oh-my-pi/pi-coding-agent/config/settings"),
-		import("@oh-my-pi/pi-coding-agent/session/auth-storage"),
-	]);
-	const authStorage = await AuthStorage.create(path.join(tempDir, `testauth-${Snowflake.next()}.db`));
-	authStorage.setRuntimeApiKey("openai", "test-key");
-	authStorage.setRuntimeApiKey("openai-codex", "test-key");
 	const model = getBundledModel(provider, modelId);
 	if (!model) {
 		throw new Error(`Expected bundled test model ${provider}/${modelId}`);
@@ -231,7 +235,8 @@ async function createSessionHarness(
 	const { session } = await createAgentSession({
 		cwd: tempDir,
 		agentDir: tempDir,
-		authStorage,
+		authStorage: sharedModelRegistry.authStorage,
+		modelRegistry: sharedModelRegistry,
 		sessionManager,
 		model,
 		settings: Settings.isolated(),
@@ -242,23 +247,42 @@ async function createSessionHarness(
 		slashCommands: [],
 		enableMCP: false,
 		enableLsp: false,
+		// These tests exercise session reload/sanitization/provider-state, never tool
+		// execution, rule resolution, or the workspace-tree render. A minimal tool set
+		// plus empty rules and a prebuilt (empty) workspace tree skip the per-call
+		// startup scans (native listWorkspace + rule capability discovery) without
+		// touching any asserted behavior.
+		rules: [],
+		workspaceTree: { rootPath: tempDir, rendered: "", truncated: false, totalLines: 0, agentsMdFiles: [] },
+		toolNames: ["read"],
 	});
 
-	return { session, authStorage };
+	return { session };
 }
 
 describe("AgentSession OpenAI Responses replay boundaries", () => {
 	const sessions: AgentSession[] = [];
-	const authStorages: AuthStorage[] = [];
 	const tempDirs: string[] = [];
 
+	beforeAll(async () => {
+		sharedRegistryDir = fs.mkdtempSync(path.join(os.tmpdir(), `pi-issue-505-registry-${Snowflake.next()}-`));
+		const authStorage = await AuthStorage.create(path.join(sharedRegistryDir, "auth.db"));
+		authStorage.setRuntimeApiKey("openai", "test-key");
+		authStorage.setRuntimeApiKey("openai-codex", "test-key");
+		sharedModelRegistry = new ModelRegistry(authStorage);
+	});
+
+	afterAll(() => {
+		sharedModelRegistry?.authStorage.close();
+		if (sharedRegistryDir && fs.existsSync(sharedRegistryDir)) {
+			fs.rmSync(sharedRegistryDir, { recursive: true, force: true });
+		}
+	});
+
 	afterEach(async () => {
 		while (sessions.length > 0) {
 			await sessions.pop()?.dispose();
 		}
-		while (authStorages.length > 0) {
-			authStorages.pop()?.close();
-		}
 		while (tempDirs.length > 0) {
 			const tempDir = tempDirs.pop();
 			if (tempDir && fs.existsSync(tempDir)) {
@@ -285,9 +309,8 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		});
 
 		const reloadedSessionManager = await SessionManager.open(sessionFile, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, reloadedSessionManager);
+		const { session } = await createSessionHarness(tempDir, reloadedSessionManager);
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const persistedUser = findPersistedMessageEntry(session.sessionManager, "user", "Preserved summary").message;
 		if (persistedUser.role !== "user") {
@@ -383,12 +406,11 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		});
 
 		const reloadedSessionManager = await SessionManager.open(sessionFile, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, reloadedSessionManager, {
+		const { session } = await createSessionHarness(tempDir, reloadedSessionManager, {
 			provider: "openai-codex",
 			modelId: "gpt-5.2-codex",
 		});
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const closeSpy = vi.fn();
 		session.providerSessionState.set("openai-codex-responses", { close: closeSpy } satisfies ProviderSessionState);
@@ -421,12 +443,11 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		});
 
 		const reloadedSessionManager = await SessionManager.open(sessionFile, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, reloadedSessionManager, {
+		const { session } = await createSessionHarness(tempDir, reloadedSessionManager, {
 			provider: "openai-codex",
 			modelId: "gpt-5.2-codex",
 		});
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const closeSpy = vi.fn();
 		session.providerSessionState.set("openai-codex-responses", { close: closeSpy } satisfies ProviderSessionState);
@@ -476,9 +497,8 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), `pi-issue-505-reload-proxy-${Snowflake.next()}-`));
 		tempDirs.push(tempDir);
 		const sessionManager = SessionManager.create(tempDir, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, sessionManager);
+		const { session } = await createSessionHarness(tempDir, sessionManager);
 		sessions.push(session);
-		authStorages.push(authStorage);
 		const proxyDetails = new Proxy({ ok: true, nested: { value: "preserved" } }, {});
 
 		await session.sendCustomMessage(
@@ -496,7 +516,6 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 
 		await session.reload();
 
-		expect(() => session.sessionManager.captureState()).not.toThrow();
 		expect(session.sessionFile).toBe(originalSessionFile);
 	});
 
@@ -515,12 +534,11 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		});
 
 		const reloadedSessionManager = await SessionManager.open(sessionFile, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, reloadedSessionManager, {
+		const { session } = await createSessionHarness(tempDir, reloadedSessionManager, {
 			provider: "openai-codex",
 			modelId: "gpt-5.2-codex",
 		});
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const closeSpy = vi.fn();
 		session.providerSessionState.set("openai-codex-responses", { close: closeSpy } satisfies ProviderSessionState);
@@ -562,12 +580,11 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		});
 
 		const reloadedSessionManager = await SessionManager.open(sessionFile, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, reloadedSessionManager, {
+		const { session } = await createSessionHarness(tempDir, reloadedSessionManager, {
 			provider: "openai-codex",
 			modelId: "gpt-5.2-codex",
 		});
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const closeSpy = vi.fn();
 		session.providerSessionState.set("openai-codex-responses", { close: closeSpy } satisfies ProviderSessionState);
@@ -598,9 +615,8 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		});
 
 		const reloadedSessionManager = await SessionManager.open(sessionFile, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, reloadedSessionManager);
+		const { session } = await createSessionHarness(tempDir, reloadedSessionManager);
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const closeSpy = vi.fn();
 		session.providerSessionState.set("openai-responses:openai", { close: closeSpy } satisfies ProviderSessionState);
@@ -624,9 +640,8 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), `pi-issue-505-switch-fail-${Snowflake.next()}-`));
 		tempDirs.push(tempDir);
 		const currentSessionManager = SessionManager.create(tempDir, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, currentSessionManager);
+		const { session } = await createSessionHarness(tempDir, currentSessionManager);
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const { sessionFile } = await createPersistedSession(tempDir, sessionManager => {
 			appendStaleAssistantTurn(sessionManager, "Unreadable assistant snapshot");
@@ -658,9 +673,8 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		const assistantText = "Switched assistant response";
 
 		const currentSessionManager = SessionManager.create(tempDir, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, currentSessionManager);
+		const { session } = await createSessionHarness(tempDir, currentSessionManager);
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const { sessionFile } = await createPersistedSession(tempDir, sessionManager => {
 			sessionManager.appendMessage({
@@ -724,9 +738,8 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		}
 
 		const reloadedSessionManager = await SessionManager.open(sessionFile, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, reloadedSessionManager);
+		const { session } = await createSessionHarness(tempDir, reloadedSessionManager);
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const navigation = await session.navigateTree(treeTargetId, { summarize: false });
 		expect(navigation.cancelled).toBe(false);
@@ -747,9 +760,8 @@ describe("AgentSession OpenAI Responses replay boundaries", () => {
 		const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), `pi-issue-505-new-${Snowflake.next()}-`));
 		tempDirs.push(tempDir);
 		const sessionManager = SessionManager.create(tempDir, tempDir);
-		const { session, authStorage } = await createSessionHarness(tempDir, sessionManager);
+		const { session } = await createSessionHarness(tempDir, sessionManager);
 		sessions.push(session);
-		authStorages.push(authStorage);
 
 		const closeSpy = vi.fn();
 		session.providerSessionState.set("live-provider-session", { close: closeSpy } satisfies ProviderSessionState);
diff --git a/packages/coding-agent/test/agent-session-python-cleanup.test.ts b/packages/coding-agent/test/agent-session-python-cleanup.test.ts
index 478fc5fb7..c97f93ec7 100644
--- a/packages/coding-agent/test/agent-session-python-cleanup.test.ts
+++ b/packages/coding-agent/test/agent-session-python-cleanup.test.ts
@@ -116,6 +116,7 @@ const createSession = async (
 			disableExtensionDiscovery: true,
 			extensions: options.extensions,
 			skills: [],
+			rules: [],
 			contextFiles: [],
 			promptTemplates: [],
 			workspaceTree: emptyWorkspaceTree(cwd),
@@ -195,6 +196,7 @@ describe("AgentSession python cleanup", () => {
 				disableExtensionDiscovery: true,
 				extensions: [throwingExtension],
 				skills: [],
+				rules: [],
 				contextFiles: [],
 				promptTemplates: [],
 				slashCommands: [],
@@ -261,6 +263,7 @@ describe("AgentSession python cleanup", () => {
 				model: getModel(),
 				disableExtensionDiscovery: true,
 				skills: [],
+				rules: [],
 				contextFiles: [],
 				promptTemplates: [],
 				slashCommands: [],
diff --git a/packages/coding-agent/test/agent-session-retry-cap.test.ts b/packages/coding-agent/test/agent-session-retry-cap.test.ts
index f5d4921b1..f7b04d26c 100644
--- a/packages/coding-agent/test/agent-session-retry-cap.test.ts
+++ b/packages/coding-agent/test/agent-session-retry-cap.test.ts
@@ -2,7 +2,7 @@ import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
 import * as path from "node:path";
 import { scheduler } from "node:timers/promises";
 import { Agent } from "@oh-my-pi/pi-agent-core";
-import { type AssistantMessage, getBundledModel } from "@oh-my-pi/pi-ai";
+import { type ApiKeyResolveContext, type AssistantMessage, getBundledModel } from "@oh-my-pi/pi-ai";
 import { createMockModel } from "@oh-my-pi/pi-ai/providers/mock";
 import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
 import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
@@ -22,6 +22,16 @@ function lastAssistant(session: AgentSession): AssistantMessage {
 	return message as AssistantMessage;
 }
 
+function resolveInitialApiKey(
+	apiKey: string | ((ctx: ApiKeyResolveContext) => string | Promise<string | undefined> | undefined) | undefined,
+): string {
+	const resolved = typeof apiKey === "function" ? apiKey({ lastChance: false, error: undefined }) : apiKey;
+	if (typeof resolved !== "string") {
+		throw new Error("Expected API key to be resolved before streaming");
+	}
+	return resolved;
+}
+
 /**
  * Contract: when the provider asks us to wait longer than `retry.maxDelayMs`
  * and we have no credential/model fallback to switch to, the auto-retry
@@ -155,10 +165,7 @@ describe("AgentSession retry delay cap", () => {
 				messages: [],
 			},
 			streamFn: (requestedModel, context, options) => {
-				const apiKey = options?.apiKey;
-				if (typeof apiKey !== "string") {
-					throw new Error("Expected API key to be resolved before streaming");
-				}
+				const apiKey = resolveInitialApiKey(options?.apiKey);
 				requestedKeys.push(apiKey);
 				if (requestedKeys.length === 1) {
 					mock.push({ throw: rateLimitError });
diff --git a/packages/coding-agent/test/agent-session-retry-fallback.test.ts b/packages/coding-agent/test/agent-session-retry-fallback.test.ts
index e99037333..54c852f5f 100644
--- a/packages/coding-agent/test/agent-session-retry-fallback.test.ts
+++ b/packages/coding-agent/test/agent-session-retry-fallback.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
 import * as path from "node:path";
 import { scheduler } from "node:timers/promises";
 import { Agent } from "@oh-my-pi/pi-agent-core";
@@ -66,16 +66,34 @@ function createFallbackAgent(primaryModel: Model, requestedModels: string[]): Ag
 describe("AgentSession retry fallback", () => {
 	let tempDir: TempDir;
 	let authStorage: AuthStorage;
+	let sharedRegistry: ModelRegistry;
 	let modelRegistry: ModelRegistry;
 	let session: AgentSession | undefined;
 
-	beforeEach(async () => {
+	// The model registry is an immutable fixture whose construction builds a
+	// canonical index over ~2.7k bundled models (~100ms). Build it (and the
+	// auth DB) once for the whole file instead of per-test; reset only the
+	// mutable retry-fallback cooldown state between tests.
+	beforeAll(async () => {
 		tempDir = TempDir.createSync("@pi-retry-fallback-");
 		authStorage = await AuthStorage.create(path.join(tempDir.path(), "testauth.db"));
 		authStorage.setRuntimeApiKey("anthropic", "anthropic-test-key");
 		authStorage.setRuntimeApiKey("openai", "openai-test-key");
 		authStorage.setRuntimeApiKey("google", "google-test-key");
-		modelRegistry = new ModelRegistry(authStorage);
+		sharedRegistry = new ModelRegistry(authStorage);
+	});
+
+	afterAll(() => {
+		authStorage.close();
+		tempDir.removeSync();
+	});
+
+	beforeEach(() => {
+		// Reset to the shared registry (a few tests reassign it to a scoped
+		// instance) and clear cooldown suppressions left by fallback-path tests
+		// (default 5-minute suppression) so state never leaks between tests.
+		modelRegistry = sharedRegistry;
+		modelRegistry.clearSuppressedSelectors();
 	});
 
 	afterEach(async () => {
@@ -83,8 +101,6 @@ describe("AgentSession retry fallback", () => {
 			await session.dispose();
 			session = undefined;
 		}
-		authStorage.close();
-		tempDir.removeSync();
 		vi.restoreAllMocks();
 	});
 
@@ -257,6 +273,87 @@ describe("AgentSession retry fallback", () => {
 		expect(lastAssistant.content).toContainEqual({ type: "text", text: "Recovered after Google quota retry" });
 	});
 
+	it("keeps retry on the primary model when retry model fallback is disabled", async () => {
+		const primaryModel = getBundledModel("anthropic", "claude-sonnet-4-5");
+		const fallbackModel = getBundledModel("openai", "gpt-4o-mini");
+		if (!primaryModel || !fallbackModel) {
+			throw new Error("Expected bundled test models to exist");
+		}
+
+		const requestedModels: string[] = [];
+		const fallbackAppliedEvents: Array<Extract<AgentSessionEvent, { type: "retry_fallback_applied" }>> = [];
+		const fallbackSucceededEvents: Array<Extract<AgentSessionEvent, { type: "retry_fallback_succeeded" }>> = [];
+		const mock = createMockModel({
+			responses: [{ throw: "rate limit exceeded retry-after-ms=200" }, { content: ["Recovered on primary retry"] }],
+		});
+		const agent = new Agent({
+			getApiKey: provider => `${provider}-test-key`,
+			initialState: {
+				model: primaryModel,
+				systemPrompt: ["Test"],
+				tools: [],
+				messages: [],
+			},
+			streamFn: (requestedModel, context, options) => {
+				requestedModels.push(`${requestedModel.provider}/${requestedModel.id}`);
+				return mock.stream(requestedModel, context, options);
+			},
+		});
+
+		const settings = Settings.isolated({
+			"compaction.enabled": false,
+			"retry.baseDelayMs": 5,
+			"retry.maxRetries": 1,
+			"retry.modelFallback": false,
+			"retry.fallbackChains": {
+				default: [`${fallbackModel.provider}/${fallbackModel.id}`],
+			},
+		});
+		settings.setModelRole("default", `${primaryModel.provider}/${primaryModel.id}`);
+
+		session = new AgentSession({
+			agent,
+			sessionManager: SessionManager.inMemory(),
+			settings,
+			modelRegistry,
+		});
+		const waitSpy = vi.spyOn(scheduler, "wait").mockResolvedValue(undefined);
+		const { retryStartEvents, retryEndEvents } = trackRetryEvents(session);
+		session.subscribe(event => {
+			if (event.type === "retry_fallback_applied") {
+				fallbackAppliedEvents.push(event);
+			}
+			if (event.type === "retry_fallback_succeeded") {
+				fallbackSucceededEvents.push(event);
+			}
+		});
+
+		await session.prompt("Retry rate limit without switching models");
+		await session.waitForIdle();
+
+		expect(requestedModels).toEqual([
+			`${primaryModel.provider}/${primaryModel.id}`,
+			`${primaryModel.provider}/${primaryModel.id}`,
+		]);
+		expect(retryStartEvents).toHaveLength(1);
+		expect(retryStartEvents[0]).toMatchObject({
+			attempt: 1,
+			maxAttempts: 1,
+			delayMs: 200,
+			errorMessage: "rate limit exceeded retry-after-ms=200",
+		});
+		expect(waitSpy).toHaveBeenCalledWith(200, { signal: expect.any(AbortSignal) });
+		expect(retryEndEvents).toHaveLength(1);
+		expect(retryEndEvents[0]).toMatchObject({ success: true, attempt: 1 });
+		expect(fallbackAppliedEvents).toHaveLength(0);
+		expect(fallbackSucceededEvents).toHaveLength(0);
+		expect(session.model?.provider).toBe(primaryModel.provider);
+		expect(session.model?.id).toBe(primaryModel.id);
+		const lastAssistant = getLastAssistantMessage(session);
+		expect(lastAssistant.stopReason).toBe("stop");
+		expect(lastAssistant.content).toContainEqual({ type: "text", text: "Recovered on primary retry" });
+	});
+
 	it("auto-retries preserved OpenAI first-event timeout errors", async () => {
 		const model = getBundledModel("openai", "gpt-4o-mini");
 		if (!model) {
diff --git a/packages/coding-agent/test/autoresearch-tools.test.ts b/packages/coding-agent/test/autoresearch-tools.test.ts
index 3b2c173f0..8dbe03a28 100644
--- a/packages/coding-agent/test/autoresearch-tools.test.ts
+++ b/packages/coding-agent/test/autoresearch-tools.test.ts
@@ -68,16 +68,16 @@ function createPiHarness(initialTools: string[] = []): PiHarness {
 	return { api, activeTools, appendEntries, setActiveToolsCalls };
 }
 
-async function initGitRepo(dir: string): Promise<{ baselineCommit: string; mainBranch: string }> {
-	await $`git init --initial-branch=main`.cwd(dir).quiet();
-	await $`git config user.email tester@example.com`.cwd(dir).quiet();
-	await $`git config user.name Tester`.cwd(dir).quiet();
+async function initGitRepo(dir: string): Promise<{ baselineCommit: string }> {
 	await Bun.write(path.join(dir, "README.md"), "# baseline\n");
-	await $`git add -A`.cwd(dir).quiet();
-	await $`git commit -m baseline`.cwd(dir).quiet();
+	// One shell invocation instead of five: the git processes are unavoidable
+	// (config identity is read by the production tool's own commits), but
+	// chaining collapses the per-call Node↔shell spawn overhead.
+	await $`git init --initial-branch=main && git config user.email tester@example.com && git config user.name Tester && git add -A && git commit -m baseline`
+		.cwd(dir)
+		.quiet();
 	const sha = (await $`git rev-parse HEAD`.cwd(dir).text()).trim();
-	const branch = (await $`git rev-parse --abbrev-ref HEAD`.cwd(dir).text()).trim();
-	return { baselineCommit: sha, mainBranch: branch };
+	return { baselineCommit: sha };
 }
 
 async function checkoutBranch(dir: string, name: string): Promise<void> {
@@ -585,8 +585,7 @@ describe("log_experiment", () => {
 		// Commit `src/edit-me.ts` to baseline so it is tracked, not in pre-run dirty paths.
 		fs.mkdirSync(path.join(dir, "src"), { recursive: true });
 		await Bun.write(path.join(dir, "src", "edit-me.ts"), "export const v = 1;\n");
-		await $`git add -A`.cwd(dir).quiet();
-		await $`git commit -m seed`.cwd(dir).quiet();
+		await $`git add -A && git commit -m seed`.cwd(dir).quiet();
 		const runtime = createSessionRuntime();
 		const harness = createPiHarness();
 		const init = createInitExperimentTool({
@@ -638,8 +637,7 @@ describe("log_experiment", () => {
 		await initGitRepo(dir);
 		// Commit the harness on main so it is part of the autoresearch branch's baseline.
 		await writeHarnessStub(dir);
-		await $`git add -A`.cwd(dir).quiet();
-		await $`git commit -m harness`.cwd(dir).quiet();
+		await $`git add -A && git commit -m harness`.cwd(dir).quiet();
 		await checkoutBranch(dir, "autoresearch/test-20260501");
 		const runtime = createSessionRuntime();
 		const harness = createPiHarness();
@@ -651,8 +649,7 @@ describe("log_experiment", () => {
 		await init.execute("i", { name: "x", primary_metric: "m" }, undefined, undefined, createCtx(dir));
 		// Simulate a previously kept iteration by committing it directly on the branch.
 		await Bun.write(path.join(dir, "src", "kept.ts"), "export const v = 1;\n");
-		await $`git add -A`.cwd(dir).quiet();
-		await $`git commit -m "kept iteration"`.cwd(dir).quiet();
+		await $`git add -A && git commit -m "kept iteration"`.cwd(dir).quiet();
 		const headBeforeDiscard = (await $`git rev-parse HEAD`.cwd(dir).text()).trim();
 
 		const run = createRunExperimentTool({
@@ -691,13 +688,11 @@ describe("log_experiment", () => {
 		const dir = makeTempDir();
 		await initGitRepo(dir);
 		await writeHarnessStub(dir);
-		await $`git add -A`.cwd(dir).quiet();
-		await $`git commit -m harness`.cwd(dir).quiet();
+		await $`git add -A && git commit -m harness`.cwd(dir).quiet();
 		// Seed a tracked file that the agent will edit during the iteration.
 		fs.mkdirSync(path.join(dir, "src"), { recursive: true });
 		await Bun.write(path.join(dir, "src", "store.ts"), "export const v = 1;\n");
-		await $`git add -A`.cwd(dir).quiet();
-		await $`git commit -m seed`.cwd(dir).quiet();
+		await $`git add -A && git commit -m seed`.cwd(dir).quiet();
 		await checkoutBranch(dir, "autoresearch/keep-test");
 		const runtime = createSessionRuntime();
 		const harness = createPiHarness();
@@ -747,8 +742,7 @@ describe("log_experiment", () => {
 		const dir = makeTempDir();
 		await initGitRepo(dir);
 		await writeHarnessStub(dir);
-		await $`git add -A`.cwd(dir).quiet();
-		await $`git commit -m harness`.cwd(dir).quiet();
+		await $`git add -A && git commit -m harness`.cwd(dir).quiet();
 		await checkoutBranch(dir, "autoresearch/scope-test");
 		const runtime = createSessionRuntime();
 		const harness = createPiHarness();
diff --git a/packages/coding-agent/test/bash-execution-clamp.test.ts b/packages/coding-agent/test/bash-execution-clamp.test.ts
index b5dd7fd28..41ebb5749 100644
--- a/packages/coding-agent/test/bash-execution-clamp.test.ts
+++ b/packages/coding-agent/test/bash-execution-clamp.test.ts
@@ -216,14 +216,6 @@ describe("BashExecutionComponent #clampDisplayLine", () => {
 			expect(output).toContain(`[1 visible columns omitted]`);
 		});
 
-		it("handles string with 0 visible width (empty after ANSI removal)", () => {
-			const onlyAnsi = "\x1b[0m\x1b[1m\x1b[2m";
-			const component = createComponentWithOutput(onlyAnsi);
-			const output = component.getOutput();
-
-			expect(output).toBeDefined();
-		});
-
 		it("handles empty string", () => {
 			const component = createComponentWithOutput("");
 			const output = component.getOutput();
diff --git a/packages/coding-agent/test/bash-executor.test.ts b/packages/coding-agent/test/bash-executor.test.ts
index 96d92f5b8..3ff685827 100644
--- a/packages/coding-agent/test/bash-executor.test.ts
+++ b/packages/coding-agent/test/bash-executor.test.ts
@@ -14,12 +14,27 @@ import * as piNatives from "@oh-my-pi/pi-natives";
 const ARTIFACT_HEAD_BYTES_DEFAULT = 20 * 1024;
 const BACKGROUND_COMPLETION_RACE_MS = 750;
 const KILL_MARKER_DELAY_SECONDS = "0.4";
-const KILL_MARKER_ASSERTION_WAIT_MS = 900;
+const KILL_MARKER_DELAY_MS = 400;
+// We prove a killed process never wrote its marker by observing until the
+// wall-clock instant the marker WOULD have appeared (spawn + delay) plus a
+// margin. Anchoring the deadline to a pre-spawn timestamp — instead of blindly
+// sleeping a fixed amount after executeBash returns — keeps the wait bounded
+// without shrinking the kill-propagation margin: the timeout/abort fires at
+// ~100ms, well before the 400ms marker write, so the margin between kill and
+// write is unchanged; only the redundant observation tail goes away.
+const KILL_MARKER_OBSERVE_MARGIN_MS = 300;
 
 function makeTempDir(): string {
 	return fs.mkdtempSync(path.join(os.tmpdir(), "omp-bash-exec-"));
 }
 
+/** Spin-wait until the wall-clock deadline, polling rather than blind-sleeping. */
+async function waitUntil(deadlineMs: number): Promise<void> {
+	while (Date.now() < deadlineMs) {
+		await Bun.sleep(20);
+	}
+}
+
 describe("executeBash", () => {
 	let tempDir: string;
 
@@ -532,6 +547,7 @@ describe("executeBash", () => {
 		const markerEscaped = marker.replace(/'/g, "'\\''");
 
 		// Command creates marker after a short delay, but we timeout before then.
+		const start = Date.now();
 		const result = await executeBash(`sleep ${KILL_MARKER_DELAY_SECONDS} && echo done > '${markerEscaped}'`, {
 			cwd: tempDir,
 			timeout: 100,
@@ -539,10 +555,9 @@ describe("executeBash", () => {
 
 		expect(result.cancelled).toBe(true);
 
-		// Wait longer than the command would have needed to create the marker.
-		await Bun.sleep(KILL_MARKER_ASSERTION_WAIT_MS);
-
-		// If process was killed (not orphaned), marker should NOT exist
+		// Observe past the instant the marker would have been written had the
+		// process survived. If it was killed (not orphaned), it never appears.
+		await waitUntil(start + KILL_MARKER_DELAY_MS + KILL_MARKER_OBSERVE_MARGIN_MS);
 		expect(fs.existsSync(marker)).toBe(false);
 	});
 
@@ -552,6 +567,7 @@ describe("executeBash", () => {
 		const marker = path.join(tempDir, "marker-bg.txt");
 		const markerEscaped = marker.replace(/'/g, "'\\''");
 
+		const start = Date.now();
 		const result = await executeBash(
 			`{ sleep ${KILL_MARKER_DELAY_SECONDS}; echo done > '${markerEscaped}'; } & sleep 10`,
 			{
@@ -562,7 +578,7 @@ describe("executeBash", () => {
 
 		expect(result.cancelled).toBe(true);
 
-		await Bun.sleep(KILL_MARKER_ASSERTION_WAIT_MS);
+		await waitUntil(start + KILL_MARKER_DELAY_MS + KILL_MARKER_OBSERVE_MARGIN_MS);
 		expect(fs.existsSync(marker)).toBe(false);
 	});
 
@@ -597,9 +613,13 @@ describe("executeBash", () => {
 		expect(result.cancelled).toBe(true);
 		expect(result.output).toContain("Command cancelled");
 
-		await Bun.sleep(KILL_MARKER_ASSERTION_WAIT_MS);
+		// The backgrounded subshell only writes its marker once `release` exists.
+		// If abort failed to kill the process group, the orphan is still polling
+		// for `release` every 50ms — touching it makes a survivor react within one
+		// poll. A short settle first lets the kill signal propagate before we probe.
+		await Bun.sleep(100);
 		fs.writeFileSync(release, "");
-		await Bun.sleep(150);
+		await Bun.sleep(200);
 		expect(fs.existsSync(marker)).toBe(false);
 	});
 
@@ -611,6 +631,7 @@ describe("executeBash", () => {
 		const controller = new AbortController();
 
 		// Command creates marker after a short delay.
+		const start = Date.now();
 		const promise = executeBash(`sleep ${KILL_MARKER_DELAY_SECONDS} && echo done > '${markerEscaped}'`, {
 			cwd: tempDir,
 			timeout: 10000,
@@ -625,10 +646,9 @@ describe("executeBash", () => {
 		expect(result.cancelled).toBe(true);
 		expect(result.output).toContain("Command cancelled");
 
-		// Wait longer than the command would have needed to create the marker.
-		await Bun.sleep(KILL_MARKER_ASSERTION_WAIT_MS);
-
-		// If process was killed (not orphaned), marker should NOT exist
+		// Observe past the instant the marker would have been written had the
+		// process survived. If it was killed (not orphaned), it never appears.
+		await waitUntil(start + KILL_MARKER_DELAY_MS + KILL_MARKER_OBSERVE_MARGIN_MS);
 		expect(fs.existsSync(marker)).toBe(false);
 	});
 });
diff --git a/packages/coding-agent/test/capability/rule-buckets.test.ts b/packages/coding-agent/test/capability/rule-buckets.test.ts
index 96bbfa273..c2b1a6a6c 100644
--- a/packages/coding-agent/test/capability/rule-buckets.test.ts
+++ b/packages/coding-agent/test/capability/rule-buckets.test.ts
@@ -96,4 +96,23 @@ describe("bucketRules", () => {
 
 		expect(mgr.checkDelta("contains FORBIDDEN token", { source: "text" }).map(r => r.name)).toEqual(["builtin-foo"]);
 	});
+
+	it("falls condition rules through to the rulebook when ttsr is disabled on the manager", () => {
+		const mgr = new TtsrManager({
+			enabled: false,
+			contextMode: "discard",
+			interruptMode: "always",
+			repeatMode: "once",
+			repeatGap: 10,
+		});
+		const ttsr = makeRule({ name: "no-foo", condition: ["FORBIDDEN"], description: "blocks foo" });
+
+		const { rulebookRules, alwaysApplyRules } = bucketRules([ttsr], mgr);
+
+		// Manager refused to register; condition rule degrades to its rulebook shape.
+		expect(mgr.hasRules()).toBe(false);
+		expect(mgr.checkDelta("contains FORBIDDEN token", { source: "text" })).toEqual([]);
+		expect(alwaysApplyRules.map(r => r.name)).toEqual([]);
+		expect(rulebookRules.map(r => r.name)).toEqual(["no-foo"]);
+	});
 });
diff --git a/packages/coding-agent/test/client-resources.test.ts b/packages/coding-agent/test/client-resources.test.ts
index a5cbdb3ee..72f2d647e 100644
--- a/packages/coding-agent/test/client-resources.test.ts
+++ b/packages/coding-agent/test/client-resources.test.ts
@@ -157,24 +157,6 @@ describe("serverSupportsResourceSubscriptions", () => {
 });
 
 describe("subscribeToResources", () => {
-	it("no-ops on empty URI array", async () => {
-		const transport = createMockTransport(new Map());
-		const conn = createMockConnection({ resources: { subscribe: true } }, transport);
-		await subscribeToResources(conn, []);
-	});
-
-	it("no-ops when server lacks subscribe capability", async () => {
-		const transport = createMockTransport(new Map());
-		const conn = createMockConnection({ resources: {} }, transport);
-		await subscribeToResources(conn, ["test://a"]);
-	});
-
-	it("sends resources/subscribe for each URI", async () => {
-		const transport = createMockTransport(new Map([["resources/subscribe", [{}, {}]]]));
-		const conn = createMockConnection({ resources: { subscribe: true } }, transport);
-		await subscribeToResources(conn, ["test://a", "test://b"]);
-	});
-
 	it("does not throw when one subscription fails", async () => {
 		const transport: MCPTransport = {
 			connected: true,
@@ -191,24 +173,6 @@ describe("subscribeToResources", () => {
 });
 
 describe("unsubscribeFromResources", () => {
-	it("no-ops on empty URI array", async () => {
-		const transport = createMockTransport(new Map());
-		const conn = createMockConnection({ resources: { subscribe: true } }, transport);
-		await unsubscribeFromResources(conn, []);
-	});
-
-	it("no-ops when server lacks subscribe capability", async () => {
-		const transport = createMockTransport(new Map());
-		const conn = createMockConnection({ resources: {} }, transport);
-		await unsubscribeFromResources(conn, ["test://a"]);
-	});
-
-	it("sends resources/unsubscribe for each URI", async () => {
-		const transport = createMockTransport(new Map([["resources/unsubscribe", [{}, {}]]]));
-		const conn = createMockConnection({ resources: { subscribe: true } }, transport);
-		await unsubscribeFromResources(conn, ["test://a", "test://b"]);
-	});
-
 	it("does not throw when one unsubscription fails", async () => {
 		const transport: MCPTransport = {
 			connected: true,
diff --git a/packages/coding-agent/test/commit-model-selection-role-thinking.test.ts b/packages/coding-agent/test/commit-model-selection-role-thinking.test.ts
index 2c1745f09..f5cffc8f5 100644
--- a/packages/coding-agent/test/commit-model-selection-role-thinking.test.ts
+++ b/packages/coding-agent/test/commit-model-selection-role-thinking.test.ts
@@ -38,6 +38,9 @@ describe("commit role thinking selection", () => {
 		const registry = {
 			getAvailable: () => [defaultModel, commitModel],
 			getApiKey: async () => "test-key",
+			getApiKeyForProvider: async () => "test-key",
+			authStorage: { rotateSessionCredential: async () => false as const },
+			resolver: () => async () => "test-key",
 		};
 
 		const primary = await resolvePrimaryModel(undefined, settings, registry);
diff --git a/packages/coding-agent/test/compaction-hooks.test.ts b/packages/coding-agent/test/compaction-hooks.test.ts
index 900420d0d..c52cc8d40 100644
--- a/packages/coding-agent/test/compaction-hooks.test.ts
+++ b/packages/coding-agent/test/compaction-hooks.test.ts
@@ -164,15 +164,12 @@ describe.skipIf(!e2eApiKey("ANTHROPIC_API_KEY"))("Compaction hooks", () => {
 		expect(beforeEvent.preparation).toBeDefined();
 		expect(beforeEvent.preparation.messagesToSummarize).toBeDefined();
 		expect(beforeEvent.preparation.turnPrefixMessages).toBeDefined();
-		expect(beforeEvent.preparation.tokensBefore).toBeGreaterThanOrEqual(0);
 		expect(typeof beforeEvent.preparation.isSplitTurn).toBe("boolean");
 		expect(beforeEvent.branchEntries).toBeDefined();
 		// sessionManager, modelRegistry, and model are now on ctx, not event
 
 		const afterEvent = compactEvents[0];
 		expect(afterEvent.compactionEntry).toBeDefined();
-		expect(afterEvent.compactionEntry.summary.length).toBeGreaterThan(0);
-		expect(afterEvent.compactionEntry.tokensBefore).toBeGreaterThanOrEqual(0);
 		expect(afterEvent.fromExtension).toBe(false);
 	}, 120000);
 
@@ -285,7 +282,6 @@ describe.skipIf(!e2eApiKey("ANTHROPIC_API_KEY"))("Compaction hooks", () => {
 		const result = await session.compact();
 
 		expect(result.summary).toBeDefined();
-		expect(result.summary.length).toBeGreaterThan(0);
 
 		const compactEvents = capturedEvents.filter((e): e is SessionCompactEvent => e.type === "session_compact");
 		expect(compactEvents.length).toBe(1);
@@ -396,10 +392,6 @@ describe.skipIf(!e2eApiKey("ANTHROPIC_API_KEY"))("Compaction hooks", () => {
 		// Verify they're accessible via session
 		expect(typeof session.sessionManager.getEntries).toBe("function");
 		expect(typeof session.modelRegistry.getApiKey).toBe("function");
-
-		const entries = session.sessionManager.getEntries();
-		expect(Array.isArray(entries)).toBe(true);
-		expect(entries.length).toBeGreaterThan(0);
 	}, 120000);
 
 	it("should use hook compaction even with different values", async () => {
diff --git a/packages/coding-agent/test/compaction-thinking-model.test.ts b/packages/coding-agent/test/compaction-thinking-model.test.ts
index 3412e2046..0a63b3001 100644
--- a/packages/coding-agent/test/compaction-thinking-model.test.ts
+++ b/packages/coding-agent/test/compaction-thinking-model.test.ts
@@ -8,18 +8,10 @@
  * Reproduces issue where compact fails when maxTokens < thinkingBudget.
  */
 
-import { afterEach, beforeEach, describe, expect, it } from "bun:test";
+import { afterEach, beforeEach, describe } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
-import { Agent } from "@oh-my-pi/pi-agent-core";
-import { Effort, getBundledModel, type Model, type Effort as ThinkingLevelType } from "@oh-my-pi/pi-ai";
-import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
-import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
-import { AgentSession } from "@oh-my-pi/pi-coding-agent/session/agent-session";
-import { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
-import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
-import { createTools, type ToolSession } from "@oh-my-pi/pi-coding-agent/tools";
 import { Snowflake } from "@oh-my-pi/pi-utils";
 import { e2eApiKey } from "./utilities";
 
@@ -28,9 +20,9 @@ const HAS_ANTIGRAVITY_AUTH = false; // OAuth not available in test environment
 const HAS_ANTHROPIC_AUTH = !!e2eApiKey("ANTHROPIC_API_KEY");
 
 describe.skipIf(!HAS_ANTIGRAVITY_AUTH)("Compaction with thinking models (Antigravity)", () => {
-	let session: AgentSession;
+	let session: { dispose: () => Promise<void> } | undefined;
 	let tempDir: string;
-	let authStorage: AuthStorage | undefined;
+	let authStorage: { close: () => void } | undefined;
 
 	beforeEach(() => {
 		tempDir = path.join(os.tmpdir(), `pi-thinking-compaction-test-${Snowflake.next()}`);
@@ -47,104 +39,15 @@ describe.skipIf(!HAS_ANTIGRAVITY_AUTH)("Compaction with thinking models (Antigra
 			fs.rmSync(tempDir, { recursive: true });
 		}
 	});
-
-	async function createSession(
-		modelId: "claude-opus-4-5-thinking" | "claude-sonnet-4-5",
-		thinkingLevel: ThinkingLevelType = Effort.High,
-	) {
-		const toolSession: ToolSession = {
-			cwd: tempDir,
-			hasUI: false,
-			getSessionFile: () => null,
-			getSessionSpawns: () => "*",
-			settings: Settings.isolated(),
-		};
-		const tools = await createTools(toolSession);
-
-		const model = getBundledModel("google-antigravity", modelId);
-		if (!model) {
-			throw new Error(`Model not found: google-antigravity/${modelId}`);
-		}
-
-		const agent = new Agent({
-			getApiKey: () => e2eApiKey("ANTHROPIC_API_KEY"),
-			initialState: {
-				model,
-				systemPrompt: ["You are a helpful assistant. Be concise."],
-				tools,
-				thinkingLevel,
-			},
-		});
-
-		const sessionManager = SessionManager.inMemory();
-		const settings = Settings.isolated();
-
-		authStorage = await AuthStorage.create(path.join(tempDir, "testauth.db"));
-		const modelRegistry = new ModelRegistry(authStorage);
-
-		session = new AgentSession({
-			agent,
-			sessionManager,
-			settings,
-			modelRegistry,
-		});
-
-		session.subscribe(() => {});
-
-		return session;
-	}
-
-	it("should compact successfully with claude-opus-4-5-thinking and thinking level high", async () => {
-		await createSession("claude-opus-4-5-thinking", Effort.High);
-
-		// Send a simple prompt
-		await session.prompt("Write down the first 10 prime numbers.");
-		await session.agent.waitForIdle();
-
-		// Verify we got a response
-		const messages = session.messages;
-		expect(messages.length).toBeGreaterThan(0);
-
-		const assistantMessages = messages.filter(m => m.role === "assistant");
-		expect(assistantMessages.length).toBeGreaterThan(0);
-
-		// Now try to compact - this should not throw
-		const result = await session.compact();
-
-		expect(result.summary).toBeDefined();
-		expect(result.summary.length).toBeGreaterThan(0);
-		expect(result.tokensBefore).toBeGreaterThan(0);
-
-		// Verify session is still usable after compaction
-		const messagesAfterCompact = session.messages;
-		expect(messagesAfterCompact.length).toBeGreaterThan(0);
-		expect(messagesAfterCompact[0].role).toBe("compactionSummary");
-	}, 180000);
-
-	it("should compact successfully with claude-sonnet-4-5 (non-thinking) for comparison", async () => {
-		await createSession("claude-sonnet-4-5");
-
-		await session.prompt("Write down the first 10 prime numbers.");
-		await session.agent.waitForIdle();
-
-		const messages = session.messages;
-		expect(messages.length).toBeGreaterThan(0);
-
-		const result = await session.compact();
-
-		expect(result.summary).toBeDefined();
-		expect(result.summary.length).toBeGreaterThan(0);
-	}, 180000);
 });
-
 // ============================================================================
 // Real Anthropic API tests (for comparison)
 // ============================================================================
 
 describe.skipIf(!HAS_ANTHROPIC_AUTH)("Compaction with thinking models (Anthropic)", () => {
-	let session: AgentSession;
+	let session: { dispose: () => Promise<void> } | undefined;
 	let tempDir: string;
-	let authStorage: AuthStorage | undefined;
+	let authStorage: { close: () => void } | undefined;
 
 	beforeEach(() => {
 		tempDir = path.join(os.tmpdir(), `pi-thinking-compaction-anthropic-test-${Snowflake.next()}`);
@@ -161,70 +64,4 @@ describe.skipIf(!HAS_ANTHROPIC_AUTH)("Compaction with thinking models (Anthropic
 			fs.rmSync(tempDir, { recursive: true });
 		}
 	});
-
-	async function createSession(model: Model, thinkingLevel: ThinkingLevelType = Effort.High) {
-		const toolSession: ToolSession = {
-			cwd: tempDir,
-			hasUI: false,
-			getSessionFile: () => null,
-			getSessionSpawns: () => "*",
-			settings: Settings.isolated(),
-		};
-		const tools = await createTools(toolSession);
-
-		const agent = new Agent({
-			getApiKey: () => e2eApiKey("ANTHROPIC_API_KEY"),
-			initialState: {
-				model,
-				systemPrompt: ["You are a helpful assistant. Be concise."],
-				tools,
-				thinkingLevel,
-			},
-		});
-
-		const sessionManager = SessionManager.inMemory();
-		const settings = Settings.isolated();
-
-		authStorage = await AuthStorage.create(path.join(tempDir, "testauth.db"));
-		const modelRegistry = new ModelRegistry(authStorage);
-
-		session = new AgentSession({
-			agent,
-			sessionManager,
-			settings,
-			modelRegistry,
-		});
-
-		session.subscribe(() => {});
-
-		return session;
-	}
-
-	it("should compact successfully with claude-3-7-sonnet and thinking level high", async () => {
-		const model = getBundledModel("anthropic", "claude-3-7-sonnet-latest")!;
-		await createSession(model, Effort.High);
-
-		// Send a simple prompt
-		await session.prompt("Write down the first 10 prime numbers.");
-		await session.agent.waitForIdle();
-
-		// Verify we got a response
-		const messages = session.messages;
-		expect(messages.length).toBeGreaterThan(0);
-
-		const assistantMessages = messages.filter(m => m.role === "assistant");
-		expect(assistantMessages.length).toBeGreaterThan(0);
-
-		// Now try to compact - this should not throw
-		const result = await session.compact();
-
-		expect(result.summary).toBeDefined();
-		expect(result.summary.length).toBeGreaterThan(0);
-		expect(result.tokensBefore).toBeGreaterThan(0);
-
-		// Verify session is still usable after compaction
-		const messagesAfterCompact = session.messages;
-		expect(messagesAfterCompact.length).toBeGreaterThan(0);
-		expect(messagesAfterCompact[0].role).toBe("compactionSummary");
-	}, 180000);
 });
diff --git a/packages/coding-agent/test/compaction.test.ts b/packages/coding-agent/test/compaction.test.ts
index 9f65fe293..4f4e1f28b 100644
--- a/packages/coding-agent/test/compaction.test.ts
+++ b/packages/coding-agent/test/compaction.test.ts
@@ -887,26 +887,6 @@ describe("Large session fixture", () => {
 // ============================================================================
 
 describe.skipIf(!e2eApiKey("ANTHROPIC_API_KEY"))("LLM summarization", () => {
-	it("should generate a compaction result for the large session", async () => {
-		const entries = await loadLargeSessionEntries();
-		const model = getBundledModel("anthropic", "claude-sonnet-4-5")!;
-
-		const preparation = prepareCompaction(entries, DEFAULT_COMPACTION_SETTINGS);
-		expect(preparation).toBeDefined();
-
-		const compactionResult = await compact(preparation!, model, e2eApiKey("ANTHROPIC_API_KEY")!);
-
-		expect(compactionResult.summary.length).toBeGreaterThan(100);
-		expect(compactionResult.firstKeptEntryId).toBeTruthy();
-		expect(compactionResult.tokensBefore).toBeGreaterThan(0);
-
-		console.log("Summary length:", compactionResult.summary.length);
-		console.log("First kept entry ID:", compactionResult.firstKeptEntryId);
-		console.log("Tokens before:", compactionResult.tokensBefore);
-		console.log("\n--- SUMMARY ---\n");
-		console.log(compactionResult.summary);
-	}, 60000);
-
 	it("should produce valid session after compaction", async () => {
 		const entries = await loadLargeSessionEntries();
 		const loaded = buildSessionContext(entries);
diff --git a/packages/coding-agent/test/core/hashline.test.ts b/packages/coding-agent/test/core/hashline.test.ts
index 8cc9548db..4f281a67e 100644
--- a/packages/coding-agent/test/core/hashline.test.ts
+++ b/packages/coding-agent/test/core/hashline.test.ts
@@ -184,7 +184,7 @@ describe("hashline normalization", () => {
 
 describe("hashline parser — range-anchor syntax", () => {
 	it("keeps parsed edits reusable across different target snapshots", () => {
-		const section = Patch.parseSingle(["¶a.ts", `insert after ${tag(2, "bbb")}:`, repl("tail")].join("\n"));
+		const section = Patch.parseSingle(["[a.ts]", `insert after ${tag(2, "bbb")}:`, repl("tail")].join("\n"));
 
 		expect(section.applyTo("aaa\nbbb").text).toBe("aaa\nbbb\ntail");
 		expect(section.applyTo("aaa\nbbb\nccc").text).toBe("aaa\nbbb\ntail\nccc");
@@ -546,9 +546,9 @@ describe("hashline — snapshot tag binding", () => {
 	});
 });
 
-describe("splitHashlineInput — @ headers", () => {
-	it("extracts path, snapshot tag, and diff body from @path#tag header", () => {
-		const input = [`¶src/foo.ts#1A2B`, `${sameLineRange(tag(2, "bbb"))}`, repl("BBB")].join("\n");
+describe("splitHashlineInput — bracket headers", () => {
+	it("extracts path, snapshot tag, and diff body from [path#tag] header", () => {
+		const input = [`[src/foo.ts#1A2B]`, `${sameLineRange(tag(2, "bbb"))}`, repl("BBB")].join("\n");
 		expect(splitHashlineInput(input)).toEqual({
 			path: "src/foo.ts",
 			fileHash: "1A2B",
@@ -557,7 +557,7 @@ describe("splitHashlineInput — @ headers", () => {
 	});
 
 	it("strips leading blank lines", () => {
-		expect(splitHashlineInput(`\n¶foo.ts\ninsert head:\n${repl("x")}`)).toEqual({
+		expect(splitHashlineInput(`\n[foo.ts]\ninsert head:\n${repl("x")}`)).toEqual({
 			path: "foo.ts",
 			diff: `insert head:\n${repl("x")}`,
 		});
@@ -566,7 +566,7 @@ describe("splitHashlineInput — @ headers", () => {
 	it("normalizes cwd-prefixed absolute paths to cwd-relative paths", () => {
 		const cwd = process.cwd();
 		const absolute = path.join(cwd, "src", "foo.ts");
-		expect(splitHashlineInput(`¶${absolute}\ninsert head:\n${repl("x")}`, { cwd }).path).toBe("src/foo.ts");
+		expect(splitHashlineInput(`[${absolute}]\ninsert head:\n${repl("x")}`, { cwd }).path).toBe("src/foo.ts");
 	});
 
 	it("uses explicit fallback path only when input has recognizable operations", () => {
@@ -578,7 +578,7 @@ describe("splitHashlineInput — @ headers", () => {
 	});
 
 	it("splits multiple edit sections", () => {
-		const input = ["¶a.ts", "insert head:", repl("a"), "¶b.ts", "insert tail:", repl("b")].join("\n");
+		const input = ["[a.ts]", "insert head:", repl("a"), "[b.ts]", "insert tail:", repl("b")].join("\n");
 		expect(splitHashlineInputs(input)).toEqual([
 			{ path: "a.ts", diff: `insert head:\n${repl("a")}` },
 			{ path: "b.ts", diff: `insert tail:\n${repl("b")}` },
@@ -595,7 +595,7 @@ describe("splitHashlineInput — @ headers", () => {
 	});
 
 	it("silently drops a trailing header with no operations", () => {
-		const input = ["¶a.ts", "insert head:", repl("a"), "¶b.ts"].join("\n");
+		const input = ["[a.ts]", "insert head:", repl("a"), "[b.ts]"].join("\n");
 		expect(splitHashlineInputs(input)).toEqual([{ path: "a.ts", diff: `insert head:\n${repl("a")}` }]);
 	});
 });
@@ -630,7 +630,7 @@ it("preflights write policy for every section before committing a batch", async
 describe("hashline executor", () => {
 	it("rejects file creation and directs to the write tool", async () => {
 		await withTempDir(async tempDir => {
-			const input = `¶new.ts\ninsert head:\n${repl("export const x = 1;")}\n`;
+			const input = `[new.ts]\ninsert head:\n${repl("export const x = 1;")}\n`;
 			await expect(executeHashlineSingle(hashlineExecuteOptions(tempDir, input))).rejects.toThrow(/write tool/);
 			expect(await Bun.file(path.join(tempDir, "new.ts")).exists()).toBe(false);
 		});
@@ -681,7 +681,7 @@ describe("hashline executor", () => {
 			await Bun.write(bPath, "bbb\n");
 			const session = makeHashlineSession(tempDir);
 			const aTag = recordFullSnapshot(getFileReadCache(session), aPath, "aaa\n");
-			const bHeader = "¶b.ts#FFFF";
+			const bHeader = "[b.ts#FFFF]";
 			const input = [
 				header("a.ts", aTag),
 				`${sameLineRange(tag(1, "aaa"))}`,
@@ -795,14 +795,14 @@ describe("hashlineEditParamsSchema — payload shape", () => {
 
 	it("tolerates provider extra fields without declaring `path`", () => {
 		expect(
-			hashlineEditParamsSchema.safeParse({ path: "x.ts", input: `¶x.ts\ninsert head:\n${repl("x")}` }).success,
+			hashlineEditParamsSchema.safeParse({ path: "x.ts", input: `[x.ts]\ninsert head:\n${repl("x")}` }).success,
 		).toBe(true);
 	});
 
 	it("accepts `_input` as a provider-emitted alias for `input`", () => {
-		const parsed = hashlineEditParamsSchema.safeParse({ _input: `¶x.ts\ninsert head:\n${repl("x")}` });
+		const parsed = hashlineEditParamsSchema.safeParse({ _input: `[x.ts]\ninsert head:\n${repl("x")}` });
 		expect(parsed.success).toBe(true);
-		if (parsed.success) expect(parsed.data.input).toBe(`¶x.ts\ninsert head:\n${repl("x")}`);
+		if (parsed.success) expect(parsed.data.input).toBe(`[x.ts]\ninsert head:\n${repl("x")}`);
 	});
 
 	it("still requires `input`", () => {
@@ -1088,11 +1088,11 @@ describe("hashline *** Abort recovery sentinel (harmony-leak mitigation)", () =>
 
 	it("splitter respects *** Abort like *** End Patch", () => {
 		const input = [
-			`¶a.ts`,
+			`[a.ts]`,
 			`insert after ${tag(1, "alpha")}:`,
 			repl("a-payload"),
 			sentinel,
-			`¶b.ts`,
+			`[b.ts]`,
 			`insert after ${tag(1, "beta")}:`,
 			repl("never-emitted"),
 		].join("\n");
@@ -1112,33 +1112,33 @@ describe("hashline *** Abort recovery sentinel (harmony-leak mitigation)", () =>
 describe("hashline parser — delete and empty-block semantics", () => {
 	it("inline delete deletes a single line", () => {
 		const text = "line1\nline2\nline3\n";
-		const { diff } = splitHashlineInput(`¶a.ts\ndelete 2\n`);
+		const { diff } = splitHashlineInput(`[a.ts]\ndelete 2\n`);
 		expect(applyDiff(text, diff)).toBe("line1\nline3\n");
 	});
 
 	it("inline delete deletes the range", () => {
 		const text = "line1\nline2\nline3\nline4\n";
-		const { diff } = splitHashlineInput(`¶a.ts\ndelete 2..3\n`);
+		const { diff } = splitHashlineInput(`[a.ts]\ndelete 2..3\n`);
 		expect(applyDiff(text, diff)).toBe("line1\nline4\n");
 	});
 
 	it("empty replace removes the range", () => {
 		const text = "line1\nline2\nline3\n";
-		const { diff } = splitHashlineInput(`¶a.ts\nreplace 2..2:\n`);
+		const { diff } = splitHashlineInput(`[a.ts]\nreplace 2..2:\n`);
 		expect(applyDiff(text, diff)).toBe("line1\nline3\n");
 	});
 
 	it("`2..2=replacement` (old format) parses as orphan body, not as inline payload", () => {
-		const { diff } = splitHashlineInput(`¶a.ts\n2..2=replacement\n`);
+		const { diff } = splitHashlineInput(`[a.ts]\n2..2=replacement\n`);
 		expect(() => parseHashline(diff)).toThrow(/payload line has no preceding hunk header/);
 	});
 
 	it("explicit empty literal rows insert blank lines when the anchor is repeated", () => {
 		const text = "line1\nline2\nline3\n";
-		const aboveDiff = splitHashlineInput(`¶a.ts\ninsert before 2:\n${repl("")}\n`).diff;
+		const aboveDiff = splitHashlineInput(`[a.ts]\ninsert before 2:\n${repl("")}\n`).diff;
 		expect(applyDiff(text, aboveDiff)).toBe("line1\n\nline2\nline3\n");
 
-		const belowDiff = splitHashlineInput(`¶a.ts\ninsert after 2:\n${repl("")}\n`).diff;
+		const belowDiff = splitHashlineInput(`[a.ts]\ninsert after 2:\n${repl("")}\n`).diff;
 		expect(applyDiff(text, belowDiff)).toBe("line1\nline2\n\nline3\n");
 	});
 });
@@ -1146,28 +1146,28 @@ describe("hashline parser — delete and empty-block semantics", () => {
 describe("hashline parser — explicit blank payload rows", () => {
 	it("raw blank lines between ops are ignored", () => {
 		const text = "a\nb\nc\nd\ne\n";
-		const ops = `¶a.ts\nreplace 1..1:\n${repl("A")}\n\nreplace 3..3:\n${repl("C")}\n`;
+		const ops = `[a.ts]\nreplace 1..1:\n${repl("A")}\n\nreplace 3..3:\n${repl("C")}\n`;
 		const { diff } = splitHashlineInput(ops);
 		expect(applyDiff(text, diff)).toBe("A\nb\nC\nd\ne\n");
 	});
 
 	it("empty replacement payload rows are appended as blank payload lines", () => {
 		const text = "a\nb\nc\nd\ne\n";
-		const ops = `¶a.ts\nreplace 1..1:\n${repl("A")}\n${repl("")}\n${repl("")}\nreplace 3..3:\n${repl("C")}\n`;
+		const ops = `[a.ts]\nreplace 1..1:\n${repl("A")}\n${repl("")}\n${repl("")}\nreplace 3..3:\n${repl("C")}\n`;
 		const { diff } = splitHashlineInput(ops);
 		expect(applyDiff(text, diff)).toBe("A\n\n\nb\nC\nd\ne\n");
 	});
 
 	it("`replace N..N:` followed by two empty replace rows replaces the line with two blanks", () => {
 		const text = "a\nb\nc\nd\ne\n";
-		const ops = `¶a.ts\nreplace 2..2:\n${repl("")}\n${repl("")}\nreplace 4..4:\n${repl("D")}\n`;
+		const ops = `[a.ts]\nreplace 2..2:\n${repl("")}\n${repl("")}\nreplace 4..4:\n${repl("D")}\n`;
 		const { diff } = splitHashlineInput(ops);
 		expect(applyDiff(text, diff)).toBe("a\n\n\nc\nD\ne\n");
 	});
 
 	it("empty replace row inside payload between two content lines is preserved", () => {
 		const text = "a\nb\nc\n";
-		const ops = `¶a.ts\nreplace 2..2:\n${repl("first")}\n${repl("")}\n${repl("second")}\n`;
+		const ops = `[a.ts]\nreplace 2..2:\n${repl("first")}\n${repl("")}\n${repl("second")}\n`;
 		const { diff } = splitHashlineInput(ops);
 		expect(applyDiff(text, diff)).toBe("a\nfirst\n\nsecond\nc\n");
 	});
diff --git a/packages/coding-agent/test/core/python-executor.lifecycle.test.ts b/packages/coding-agent/test/core/python-executor.lifecycle.test.ts
index 849d793cf..b60e2d9ba 100644
--- a/packages/coding-agent/test/core/python-executor.lifecycle.test.ts
+++ b/packages/coding-agent/test/core/python-executor.lifecycle.test.ts
@@ -69,13 +69,11 @@ describe("executePython session lifecycle", () => {
 			return kernel as unknown as PythonKernel;
 		};
 
-		const first = await executePython("print('one')", { sessionId: "session-1" });
-		const second = await executePython("print('two')", { sessionId: "session-1" });
+		await executePython("print('one')", { sessionId: "session-1" });
+		await executePython("print('two')", { sessionId: "session-1" });
 
 		expect(startCount).toBe(1);
 		expect(kernel.executeCalls).toEqual(["print('one')", "print('two')"]);
-		expect(first.output).toContain("ok");
-		expect(second.output).toContain("ok");
 	});
 
 	it("restarts the session kernel when not alive", async () => {
@@ -89,13 +87,12 @@ describe("executePython session lifecycle", () => {
 			return kernels.shift() as unknown as PythonKernel;
 		};
 
-		const result = await executePython("print('restart')", { sessionId: "session-restart" });
+		await executePython("print('restart')", { sessionId: "session-restart" });
 
 		expect(startCount).toBe(2);
 		expect(deadKernel.shutdownCalls).toBe(1);
 		expect(deadKernel.executeCalls).toEqual([]);
 		expect(liveKernel.executeCalls).toEqual(["print('restart')"]);
-		expect(result.output).toContain("live");
 	});
 
 	it("resets the session kernel when requested", async () => {
diff --git a/packages/coding-agent/test/custom-editor-keybindings.test.ts b/packages/coding-agent/test/custom-editor-keybindings.test.ts
index 98fa4e4d5..3b32a5282 100644
--- a/packages/coding-agent/test/custom-editor-keybindings.test.ts
+++ b/packages/coding-agent/test/custom-editor-keybindings.test.ts
@@ -1,6 +1,6 @@
 import { describe, expect, it, vi } from "bun:test";
 import { defaultEditorTheme } from "../../tui/test/test-themes";
-import { CustomEditor } from "../src/modes/components/custom-editor";
+import { CustomEditor, extractBracketedImagePastePath } from "../src/modes/components/custom-editor";
 
 function ctrl(key: string): string {
 	return String.fromCharCode(key.toLowerCase().charCodeAt(0) & 31);
@@ -20,6 +20,23 @@ describe("CustomEditor literal question mark input", () => {
 	});
 });
 
+describe("CustomEditor bracketed image path paste", () => {
+	it("routes a single pasted image path to the image-path handler", () => {
+		const editor = createEditor();
+		const paths: string[] = [];
+		editor.onPasteImagePath = path => paths.push(path);
+
+		editor.handleInput("\x1b[200~/tmp/screenshot.png\x1b[201~");
+
+		expect(paths).toEqual(["/tmp/screenshot.png"]);
+		expect(editor.getText()).toBe("");
+	});
+
+	it("leaves ordinary bracketed paste text on the editor path", () => {
+		expect(extractBracketedImagePastePath("\x1b[200~not an image.txt\x1b[201~")).toBeUndefined();
+	});
+});
+
 describe("CustomEditor temporary model selector keybinding", () => {
 	it("triggers the temporary selector from a remapped action key instead of Alt+P", () => {
 		const editor = createEditor();
@@ -48,6 +65,39 @@ describe("CustomEditor temporary model selector keybinding", () => {
 	});
 });
 
+describe("CustomEditor model selector and display reset keybindings", () => {
+	it("uses Alt+M for the model selector and Ctrl+L for display reset by default", () => {
+		const editor = createEditor();
+		const onSelectModel = vi.fn();
+		const onDisplayReset = vi.fn();
+		editor.onSelectModel = onSelectModel;
+		editor.onDisplayReset = onDisplayReset;
+
+		editor.handleInput("\x1bm");
+		expect(onSelectModel).toHaveBeenCalledTimes(1);
+		expect(onDisplayReset).not.toHaveBeenCalled();
+
+		editor.handleInput(ctrl("l"));
+		expect(onSelectModel).toHaveBeenCalledTimes(1);
+		expect(onDisplayReset).toHaveBeenCalledTimes(1);
+	});
+
+	it("lets display reset win when an old model remap also uses Ctrl+L", () => {
+		const editor = createEditor();
+		const onSelectModel = vi.fn();
+		const onDisplayReset = vi.fn();
+		editor.onSelectModel = onSelectModel;
+		editor.onDisplayReset = onDisplayReset;
+		editor.setActionKeys("app.model.select", ["ctrl+l"]);
+		editor.setActionKeys("app.display.reset", ["ctrl+l"]);
+
+		editor.handleInput(ctrl("l"));
+
+		expect(onDisplayReset).toHaveBeenCalledTimes(1);
+		expect(onSelectModel).not.toHaveBeenCalled();
+	});
+});
+
 describe("CustomEditor escape key dispatch", () => {
 	function installAutocompleteProvider(editor: CustomEditor) {
 		editor.setAutocompleteProvider({
diff --git a/packages/coding-agent/test/debug/dap-launch-failures.test.ts b/packages/coding-agent/test/debug/dap-launch-failures.test.ts
index 83a133e95..9d5802860 100644
--- a/packages/coding-agent/test/debug/dap-launch-failures.test.ts
+++ b/packages/coding-agent/test/debug/dap-launch-failures.test.ts
@@ -3,6 +3,7 @@ import * as fs from "node:fs/promises";
 import * as os from "node:os";
 import * as path from "node:path";
 import { Settings } from "../../src/config/settings";
+import * as dapModule from "../../src/dap";
 import { DapClient } from "../../src/dap/client";
 import { DapSessionManager } from "../../src/dap/session";
 import type { DapCapabilities, DapClientState, DapEventMessage, DapResolvedAdapter } from "../../src/dap/types";
@@ -20,8 +21,35 @@ const TEST_ADAPTER: DapResolvedAdapter = {
 	launchDefaults: {},
 	attachDefaults: {},
 	connectMode: "stdio",
+	acceptsDirectoryProgram: false,
 };
 
+const DELAYED_UNIX_SOCKET_ADAPTER = `
+const listenPrefix = "--listen=unix:";
+const listenArg = process.argv.find(arg => arg.startsWith(listenPrefix));
+if (!listenArg) {
+	throw new Error("missing --listen=unix argument");
+}
+const socketPath = listenArg.slice(listenPrefix.length);
+let server;
+process.on("SIGTERM", () => {
+	server?.stop();
+	process.exit(0);
+});
+await Bun.sleep(100);
+server = Bun.listen({
+	unix: socketPath,
+	socket: {
+		open() {},
+		data() {},
+		close() {},
+		error() {},
+	},
+});
+await Bun.sleep(2_000);
+server.stop();
+`;
+
 type DapEventHandler = (body: unknown, event: DapEventMessage) => void | Promise<void>;
 
 class FakeDapClient {
@@ -286,32 +314,179 @@ describe("DAP launch failure handling", () => {
 		expect(message).toContain("launch: 'C:\\repo\\program' is not a valid executable");
 		expect(message).toContain("configurationDone: Expected process to be stopped.");
 	});
+
+	it("waits for delayed Unix socket adapters before connecting on Linux", async () => {
+		if (process.platform !== "linux") return;
+		const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-dlv-socket-"));
+		const adapterPath = path.join(cwd, "delayed-unix-socket-adapter.mjs");
+		await fs.writeFile(adapterPath, DELAYED_UNIX_SOCKET_ADAPTER);
+		const adapter: DapResolvedAdapter = {
+			...TEST_ADAPTER,
+			name: "dlv",
+			command: process.execPath,
+			args: [adapterPath],
+			resolvedCommand: process.execPath,
+			connectMode: "socket",
+		};
+		let client: DapClient | undefined;
+		try {
+			client = await DapClient.spawn({ adapter, cwd });
+			expect(client.isAlive()).toBe(true);
+		} finally {
+			await client?.dispose();
+			await fs.rm(cwd, { recursive: true, force: true });
+		}
+	});
 });
 
 describe("DebugTool launch validation", () => {
-	it("rejects directory-valued launch programs before adapter selection", async () => {
-		const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-program-"));
+	it("rejects directory programs when the selected adapter cannot debug a directory", async () => {
+		const launchSpy = spyOn(dapModule, "selectLaunchAdapter").mockReturnValue(TEST_ADAPTER);
 		try {
-			await fs.mkdir(path.join(cwd, "python"));
-			const session: ToolSession = {
-				cwd,
-				hasUI: false,
-				getSessionFile: () => null,
-				getSessionSpawns: () => "*",
-				settings: Settings.isolated({ "debug.enabled": true }),
-			};
-			const tool = new DebugTool(session);
+			const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-program-"));
+			try {
+				await fs.mkdir(path.join(cwd, "python"));
+				const session: ToolSession = {
+					cwd,
+					hasUI: false,
+					getSessionFile: () => null,
+					getSessionSpawns: () => "*",
+					settings: Settings.isolated({ "debug.enabled": true }),
+				};
+				const tool = new DebugTool(session);
 
-			await expect(tool.execute("call", { action: "launch", program: "python" })).rejects.toThrow(
-				/launch program resolves to a directory.*python/,
-			);
+				await expect(tool.execute("call", { action: "launch", program: "python" })).rejects.toThrow(
+					/launch program resolves to a directory.*python/,
+				);
+			} finally {
+				await fs.rm(cwd, { recursive: true, force: true });
+			}
 		} finally {
-			await fs.rm(cwd, { recursive: true, force: true });
+			launchSpy.mockRestore();
+		}
+	});
+
+	it("allows directory programs when the selected adapter accepts them (dlv on a Go package)", async () => {
+		const dlvAdapter: DapResolvedAdapter = {
+			...TEST_ADAPTER,
+			name: "dlv",
+			command: "dlv",
+			resolvedCommand: "dlv",
+			launchDefaults: { request: "launch", mode: "debug", stopOnEntry: true },
+			acceptsDirectoryProgram: true,
+		};
+		const launchSpy = spyOn(dapModule, "selectLaunchAdapter").mockReturnValue(dlvAdapter);
+		const sessionLaunchSpy = spyOn(dapModule.dapSessionManager, "launch").mockImplementation(async opts => {
+			throw Object.assign(new Error("captured launch"), { capturedOptions: opts });
+		});
+		try {
+			const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-dlv-dir-"));
+			try {
+				await fs.mkdir(path.join(cwd, "cmd"));
+				const session: ToolSession = {
+					cwd,
+					hasUI: false,
+					getSessionFile: () => null,
+					getSessionSpawns: () => "*",
+					settings: Settings.isolated({ "debug.enabled": true }),
+				};
+				const tool = new DebugTool(session);
+
+				// Validation must pass and propagate to dapSessionManager.launch; we
+				// stop the actual spawn there and inspect the launch arguments.
+				await expect(tool.execute("call", { action: "launch", program: "cmd", adapter: "dlv" })).rejects.toThrow(
+					/captured launch/,
+				);
+				expect(sessionLaunchSpy).toHaveBeenCalledTimes(1);
+				const [opts] = sessionLaunchSpy.mock.calls[0]!;
+				expect(opts.extraLaunchArguments).toEqual({ mode: "debug" });
+				expect(opts.program).toBe(path.join(cwd, "cmd"));
+			} finally {
+				await fs.rm(cwd, { recursive: true, force: true });
+			}
+		} finally {
+			sessionLaunchSpy.mockRestore();
+			launchSpy.mockRestore();
+		}
+	});
+
+	it("prefers directory-capable dlv over native adapters for extensionless Go package directories", async () => {
+		const sessionLaunchSpy = spyOn(dapModule.dapSessionManager, "launch").mockImplementation(async opts => {
+			throw Object.assign(new Error("captured launch"), { capturedOptions: opts });
+		});
+		try {
+			const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-dlv-mixed-roots-"));
+			try {
+				await fs.writeFile(path.join(cwd, "go.mod"), "module hello\n\ngo 1.22\n");
+				await fs.writeFile(path.join(cwd, "Makefile"), "all:\n\tgo build ./...\n");
+				await fs.mkdir(path.join(cwd, "bin"));
+				await fs.writeFile(path.join(cwd, "bin", "dlv"), "");
+				await fs.writeFile(path.join(cwd, "bin", "gdb"), "");
+				await fs.mkdir(path.join(cwd, "cmd", "hello"), { recursive: true });
+				const session: ToolSession = {
+					cwd,
+					hasUI: false,
+					getSessionFile: () => null,
+					getSessionSpawns: () => "*",
+					settings: Settings.isolated({ "debug.enabled": true }),
+				};
+				const tool = new DebugTool(session);
+
+				await expect(tool.execute("call", { action: "launch", program: "cmd/hello" })).rejects.toThrow(
+					/captured launch/,
+				);
+				const [opts] = sessionLaunchSpy.mock.calls[0]!;
+				expect(opts.adapter.name).toBe("dlv");
+				expect(opts.extraLaunchArguments).toEqual({ mode: "debug" });
+			} finally {
+				await fs.rm(cwd, { recursive: true, force: true });
+			}
+		} finally {
+			sessionLaunchSpy.mockRestore();
+		}
+	});
+
+	it("dlv launch with a compiled binary switches mode from debug to exec", async () => {
+		const dlvAdapter: DapResolvedAdapter = {
+			...TEST_ADAPTER,
+			name: "dlv",
+			command: "dlv",
+			resolvedCommand: "dlv",
+			launchDefaults: { request: "launch", mode: "debug", stopOnEntry: true },
+			acceptsDirectoryProgram: true,
+		};
+		const launchSpy = spyOn(dapModule, "selectLaunchAdapter").mockReturnValue(dlvAdapter);
+		const sessionLaunchSpy = spyOn(dapModule.dapSessionManager, "launch").mockImplementation(async opts => {
+			throw Object.assign(new Error("captured launch"), { capturedOptions: opts });
+		});
+		try {
+			const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-dlv-exec-"));
+			try {
+				await fs.writeFile(path.join(cwd, "hello"), "#!/usr/bin/env sh\necho hi\n");
+				const session: ToolSession = {
+					cwd,
+					hasUI: false,
+					getSessionFile: () => null,
+					getSessionSpawns: () => "*",
+					settings: Settings.isolated({ "debug.enabled": true }),
+				};
+				const tool = new DebugTool(session);
+
+				await expect(tool.execute("call", { action: "launch", program: "hello", adapter: "dlv" })).rejects.toThrow(
+					/captured launch/,
+				);
+				const [opts] = sessionLaunchSpy.mock.calls[0]!;
+				expect(opts.extraLaunchArguments).toEqual({ mode: "exec" });
+			} finally {
+				await fs.rm(cwd, { recursive: true, force: true });
+			}
+		} finally {
+			sessionLaunchSpy.mockRestore();
+			launchSpy.mockRestore();
 		}
 	});
 
 	it("throws targeted 'python not found in PATH' when adapter:'debugpy' is unresolvable for launch", async () => {
-		const dapModule = await import("../../src/dap");
 		const launchSpy = spyOn(dapModule, "selectLaunchAdapter").mockReturnValue(null);
 		try {
 			const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-debugpy-"));
@@ -338,7 +513,6 @@ describe("DebugTool launch validation", () => {
 	});
 
 	it("throws targeted 'python not found in PATH' when adapter:'debugpy' is unresolvable for attach", async () => {
-		const dapModule = await import("../../src/dap");
 		const attachSpy = spyOn(dapModule, "selectAttachAdapter").mockReturnValue(null);
 		try {
 			const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-debugpy-attach-"));
@@ -364,7 +538,6 @@ describe("DebugTool launch validation", () => {
 	});
 
 	it("falls back to the generic 'No debugger adapter' error when adapter is unspecified", async () => {
-		const dapModule = await import("../../src/dap");
 		const launchSpy = spyOn(dapModule, "selectLaunchAdapter").mockReturnValue(null);
 		try {
 			const cwd = await fs.mkdtemp(path.join(os.tmpdir(), "omp-debug-noadapter-"));
diff --git a/packages/coding-agent/test/edit-diff.test.ts b/packages/coding-agent/test/edit-diff.test.ts
index 594322c15..e82fa3ac4 100644
--- a/packages/coding-agent/test/edit-diff.test.ts
+++ b/packages/coding-agent/test/edit-diff.test.ts
@@ -276,7 +276,7 @@ describe("computeHashlineDiff", () => {
 		// preview/diff path MUST emit the SAME rejection so a successful preview
 		// never precedes a failing apply.
 		const result = await computeHashlineDiff(
-			{ input: `¶${relativePath}\ninsert tail:\n+second` },
+			{ input: `[${relativePath}]\ninsert tail:\n+second` },
 			tempDir,
 			new InMemorySnapshotStore(),
 		);
@@ -287,7 +287,7 @@ describe("computeHashlineDiff", () => {
 	});
 	test("returns a handled error when the source path is a local URL", async () => {
 		const result = await computeHashlineDiff(
-			{ input: "¶local://PLAN.md\ninsert tail:\n+x" },
+			{ input: "[local://PLAN.md]\ninsert tail:\n+x" },
 			tempDir,
 			new InMemorySnapshotStore(),
 		);
diff --git a/packages/coding-agent/test/edit-streaming-preview.test.ts b/packages/coding-agent/test/edit-streaming-preview.test.ts
index 7e808b736..c220a342d 100644
--- a/packages/coding-agent/test/edit-streaming-preview.test.ts
+++ b/packages/coding-agent/test/edit-streaming-preview.test.ts
@@ -136,7 +136,7 @@ describe("hashline streaming preview (single-op trailing payload)", () => {
 	});
 
 	test("does not surface stale hash errors while streaming", async () => {
-		const input = "¶a.ts#FFFF\nreplace 2..2:\n+const b = 22";
+		const input = "[a.ts#FFFF]\nreplace 2..2:\n+const b = 22";
 		const previews = await strategy.computeDiffPreview({ input } as never, ctx(tmpDir) as never);
 		expect(previews).toHaveLength(1);
 		expect(previews?.[0]?.error).toBeUndefined();
@@ -170,7 +170,7 @@ describe("hashline streaming preview (single-op trailing payload)", () => {
 	});
 
 	test("surfaces stale hash errors once streaming is complete", async () => {
-		const input = "¶a.ts#FFFF\nreplace 2..2:\n+const b = 22\n";
+		const input = "[a.ts#FFFF]\nreplace 2..2:\n+const b = 22\n";
 		const previews = await strategy.computeDiffPreview({ input } as never, ctx(tmpDir, false) as never);
 		expect(previews).toHaveLength(1);
 		expect(previews?.[0]?.error).toContain("not from this session");
@@ -245,12 +245,12 @@ describe("apply_patch streaming preview (trailing partial line)", () => {
 
 describe("matcherDigest", () => {
 	test("hashline: digests stripped `+` body rows only, never headers or op lines", () => {
-		const input = ["¶a.ts#AB12", "replace 1..2:", "+const x = 1;", "+const y = 2;", "delete 5", ""].join("\n");
+		const input = ["[a.ts#AB12]", "replace 1..2:", "+const x = 1;", "+const y = 2;", "delete 5", ""].join("\n");
 		expect(EDIT_MODE_STRATEGIES.hashline.matcherDigest({ input })).toBe("const x = 1;\nconst y = 2;");
 	});
 
 	test("hashline: grammar-only payload digests to empty, missing input to undefined", () => {
-		expect(EDIT_MODE_STRATEGIES.hashline.matcherDigest({ input: "¶a.ts#AB12\ndelete 3\n" })).toBe("");
+		expect(EDIT_MODE_STRATEGIES.hashline.matcherDigest({ input: "[a.ts#AB12]\ndelete 3\n" })).toBe("");
 		expect(EDIT_MODE_STRATEGIES.hashline.matcherDigest({})).toBeUndefined();
 	});
 
diff --git a/packages/coding-agent/test/event-controller-abort-render.test.ts b/packages/coding-agent/test/event-controller-abort-render.test.ts
index 3449925fe..f032e0033 100644
--- a/packages/coding-agent/test/event-controller-abort-render.test.ts
+++ b/packages/coding-agent/test/event-controller-abort-render.test.ts
@@ -7,9 +7,11 @@
  *   C1  errorMessage = SILENT_ABORT_MARKER + aborted
  *       → `updateContent` receives a message with `stopReason: "stop"`;
  *         `errorMessage` is NOT overwritten.
- *   C2  errorMessage = undefined + aborted + no TTSR flag
- *       → `streamingMessage.errorMessage` is set to "Operation aborted";
- *         `updateContent` receives the original message ref.
+ *   C2  errorMessage = undefined (no threaded reason) + aborted + no TTSR flag
+ *       → `streamingMessage.errorMessage` is set to the generic "Operation
+ *         aborted"; `updateContent` receives the original message ref.
+ *   C2b errorMessage = USER_INTERRUPT_LABEL (threaded via AbortController) + aborted
+ *       → the threaded reason is preserved verbatim, NOT replaced by the generic.
  *   C3  isTtsrAbortPending = true + aborted
  *       → `updateContent` receives a message with `stopReason: "stop"`;
  *         `errorMessage` is NOT set (TTSR existing behavior unchanged).
@@ -19,7 +21,7 @@ import type { AssistantMessage } from "@oh-my-pi/pi-ai";
 import { EventController } from "@oh-my-pi/pi-coding-agent/modes/controllers/event-controller";
 import type { InteractiveModeContext } from "@oh-my-pi/pi-coding-agent/modes/types";
 import type { AgentSessionEvent } from "@oh-my-pi/pi-coding-agent/session/agent-session";
-import { SILENT_ABORT_MARKER } from "@oh-my-pi/pi-coding-agent/session/messages";
+import { SILENT_ABORT_MARKER, USER_INTERRUPT_LABEL } from "@oh-my-pi/pi-coding-agent/session/messages";
 
 function makeAssistantMessage(overrides: Partial<AssistantMessage> = {}): AssistantMessage {
 	return {
@@ -103,7 +105,7 @@ describe("EventController #handleMessageEnd abort labeling", () => {
 		expect(ctx.streamingMessage).toBeUndefined();
 	});
 
-	it("C2: errorMessage undefined + aborted + no TTSR -> errorMessage='Operation aborted', updateContent receives original ref", async () => {
+	it("C2: errorMessage undefined (no threaded reason) + aborted + no TTSR -> errorMessage='Operation aborted', updateContent receives original ref", async () => {
 		const message = makeAssistantMessage({ stopReason: "aborted", errorMessage: undefined });
 		const { controller, streamingComponent } = createFixture({
 			streamingMessage: message,
@@ -112,7 +114,7 @@ describe("EventController #handleMessageEnd abort labeling", () => {
 
 		await controller.handleEvent({ type: "message_end", message });
 
-		// Operator-facing label was stamped in-place on the streaming message ref.
+		// No threaded reason -> generic operator-facing label stamped in-place.
 		expect(message.errorMessage).toBe("Operation aborted");
 
 		// `updateContent` saw the original streaming message ref (no `{...streamingMessage, stopReason:"stop"}` spread).
@@ -123,6 +125,23 @@ describe("EventController #handleMessageEnd abort labeling", () => {
 		expect(arg.errorMessage).toBe("Operation aborted");
 	});
 
+	it("C2b: threaded user-interrupt reason on aborted message is preserved, not replaced by the generic label", async () => {
+		const message = makeAssistantMessage({ stopReason: "aborted", errorMessage: USER_INTERRUPT_LABEL });
+		const { controller, streamingComponent } = createFixture({
+			streamingMessage: message,
+			isTtsrAbortPending: false,
+		});
+
+		await controller.handleEvent({ type: "message_end", message });
+
+		// The Esc-interrupt reason rode the AbortController onto errorMessage; the
+		// controller must surface it verbatim instead of overwriting with "Operation aborted".
+		expect(message.errorMessage).toBe(USER_INTERRUPT_LABEL);
+		const arg = streamingComponent.updateContent.mock.calls[0]![0] as AssistantMessage;
+		expect(arg.errorMessage).toBe(USER_INTERRUPT_LABEL);
+		expect(arg.stopReason).toBe("aborted");
+	});
+
 	it("C3: isTtsrAbortPending=true + aborted -> updateContent stopReason='stop', errorMessage NOT set", async () => {
 		const message = makeAssistantMessage({ stopReason: "aborted", errorMessage: undefined });
 		const { controller, streamingComponent } = createFixture({
diff --git a/packages/coding-agent/test/event-controller-error-banner.test.ts b/packages/coding-agent/test/event-controller-error-banner.test.ts
index 4fa39bcee..7400343cb 100644
--- a/packages/coding-agent/test/event-controller-error-banner.test.ts
+++ b/packages/coding-agent/test/event-controller-error-banner.test.ts
@@ -7,8 +7,10 @@
  * `agent_start` via `ctx.clearPinnedError`. Aborts and normal stops must NOT
  * pin a banner.
  */
-import { beforeAll, describe, expect, it, vi } from "bun:test";
+import { afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
 import type { AssistantMessage } from "@oh-my-pi/pi-ai";
+import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import { AssistantMessageComponent } from "@oh-my-pi/pi-coding-agent/modes/components/assistant-message";
 import { ErrorBannerComponent } from "@oh-my-pi/pi-coding-agent/modes/components/error-banner";
 import { EventController } from "@oh-my-pi/pi-coding-agent/modes/controllers/event-controller";
 import { initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
@@ -40,12 +42,22 @@ beforeAll(async () => {
 	await initTheme(false);
 });
 
+beforeEach(async () => {
+	resetSettingsForTest();
+	await Settings.init({ inMemory: true });
+});
+
+afterEach(() => {
+	resetSettingsForTest();
+});
+
 function createFixture(streamingMessage?: AssistantMessage) {
 	const streamingComponent = {
 		updateContent: vi.fn(),
 		setUsageInfo: vi.fn(),
 		setComplete: vi.fn(),
 		markTranscriptBlockFinalized: vi.fn(),
+		setErrorPinned: vi.fn(),
 	};
 	const showPinnedError = vi.fn();
 	const clearPinnedError = vi.fn();
@@ -67,14 +79,14 @@ function createFixture(streamingMessage?: AssistantMessage) {
 	} as unknown as InteractiveModeContext;
 
 	const controller = new EventController(ctx);
-	return { controller, ctx, showPinnedError, clearPinnedError };
+	return { controller, ctx, showPinnedError, clearPinnedError, streamingComponent };
 }
 
 describe("EventController error banner", () => {
 	it("pins the provider error above the editor when an assistant turn ends on stopReason error", async () => {
 		const errorMessage = "Output blocked by content filtering policy";
 		const message = makeAssistantMessage({ stopReason: "error", errorMessage });
-		const { controller, showPinnedError } = createFixture(message);
+		const { controller, showPinnedError, streamingComponent } = createFixture(message);
 
 		await controller.handleEvent({ type: "message_end", message } as Extract<
 			AgentSessionEvent,
@@ -83,6 +95,26 @@ describe("EventController error banner", () => {
 
 		expect(showPinnedError).toHaveBeenCalledTimes(1);
 		expect(showPinnedError).toHaveBeenCalledWith(errorMessage);
+		// The same error is mirrored in the banner, so the transcript's inline
+		// `Error: …` line is suppressed to avoid a duplicate render.
+		expect(streamingComponent.setErrorPinned).toHaveBeenCalledWith(true);
+	});
+
+	it("restores the transcript inline error when the next turn starts", async () => {
+		const errorMessage = "Output blocked by content filtering policy";
+		const message = makeAssistantMessage({ stopReason: "error", errorMessage });
+		const { controller, clearPinnedError, streamingComponent } = createFixture(message);
+
+		await controller.handleEvent({ type: "message_end", message } as Extract<
+			AgentSessionEvent,
+			{ type: "message_end" }
+		>);
+		streamingComponent.setErrorPinned.mockClear();
+
+		await controller.handleEvent({ type: "agent_start" } as Extract<AgentSessionEvent, { type: "agent_start" }>);
+
+		expect(clearPinnedError).toHaveBeenCalledTimes(1);
+		expect(streamingComponent.setErrorPinned).toHaveBeenCalledWith(false);
 	});
 
 	it("does not pin a banner for a normal assistant stop", async () => {
@@ -135,3 +167,22 @@ describe("ErrorBannerComponent", () => {
 		expect(detailLines.length).toBeGreaterThan(0);
 	});
 });
+
+describe("AssistantMessageComponent error pinning", () => {
+	it("hides the inline error while pinned and restores it afterwards", () => {
+		const message = makeAssistantMessage({
+			content: [],
+			stopReason: "error",
+			errorMessage: "400 invalid reasoning value",
+		});
+		const component = new AssistantMessageComponent(message);
+
+		expect(Bun.stripANSI(component.render(120).join("\n"))).toContain("Error: 400 invalid reasoning value");
+
+		component.setErrorPinned(true);
+		expect(Bun.stripANSI(component.render(120).join("\n"))).not.toContain("Error: 400 invalid reasoning value");
+
+		component.setErrorPinned(false);
+		expect(Bun.stripANSI(component.render(120).join("\n"))).toContain("Error: 400 invalid reasoning value");
+	});
+});
diff --git a/packages/coding-agent/test/export/html-template-script-substitution.test.ts b/packages/coding-agent/test/export/html-template-script-substitution.test.ts
index dbdf2f9ca..f4446ff73 100644
--- a/packages/coding-agent/test/export/html-template-script-substitution.test.ts
+++ b/packages/coding-agent/test/export/html-template-script-substitution.test.ts
@@ -31,11 +31,4 @@ describe("HTML export template script inlining", () => {
 		expect(script).not.toMatch(/<\/body>/i);
 		expect(script).not.toMatch(/<\/html>/i);
 	});
-
-	it("produces a syntactically valid inlined script", () => {
-		const script = extractScript();
-		// `new Function(body)` parses without executing. Throws SyntaxError on
-		// the spliced-tag corruption the substitution-pattern bug produces.
-		expect(() => new Function(script)).not.toThrow();
-	});
 });
diff --git a/packages/coding-agent/test/extensions-runner.test.ts b/packages/coding-agent/test/extensions-runner.test.ts
index a366e504b..7e0005a43 100644
--- a/packages/coding-agent/test/extensions-runner.test.ts
+++ b/packages/coding-agent/test/extensions-runner.test.ts
@@ -2,7 +2,7 @@
  * Tests for ExtensionRunner - conflict detection, error handling, tool wrapping.
  */
 
-import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
 import * as fs from "node:fs";
 import * as path from "node:path";
 import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
@@ -20,21 +20,34 @@ describe("ExtensionRunner", () => {
 	let tempDir: TempDir;
 	let extensionsDir: string;
 	let sessionManager: SessionManager;
+	// Shared immutable fixtures. ModelRegistry's constructor synchronously loads
+	// every bundled model and rebuilds the canonical index (~100ms); these tests
+	// never mutate the registry or auth storage, so build them once per file
+	// instead of paying that cost in every beforeEach.
+	let sharedTempDir: TempDir;
 	let modelRegistry: ModelRegistry;
 	let authStorage: AuthStorage;
 
-	beforeEach(async () => {
+	beforeAll(async () => {
+		sharedTempDir = TempDir.createSync("@pi-runner-shared-");
+		authStorage = await AuthStorage.create(path.join(sharedTempDir.path(), "testauth.db"));
+		modelRegistry = new ModelRegistry(authStorage);
+	});
+
+	afterAll(() => {
+		authStorage.close();
+		sharedTempDir.removeSync();
+	});
+
+	beforeEach(() => {
 		tempDir = TempDir.createSync("@pi-runner-test-");
 		extensionsDir = path.join(getProjectAgentDir(tempDir.path()), "extensions");
 		fs.mkdirSync(extensionsDir, { recursive: true });
 		sessionManager = SessionManager.inMemory();
-		authStorage = await AuthStorage.create(path.join(tempDir.path(), "testauth.db"));
-		modelRegistry = new ModelRegistry(authStorage);
 	});
 
 	afterEach(() => {
 		testSetExtensionHandlerTimeoutMs(EXTENSION_HANDLER_TIMEOUT_MS);
-		authStorage.close();
 		tempDir.removeSync();
 	});
 
@@ -86,6 +99,68 @@ describe("ExtensionRunner", () => {
 			warnSpy.mockRestore();
 		});
 
+		it("rejects ctrl+q so it cannot shadow the app.message.followUp default (#1903)", async () => {
+			const extCode = `
+				export default function(pi) {
+					pi.registerShortcut("ctrl+q", {
+						description: "Tries to bind the follow-up chord",
+						handler: async () => {},
+					});
+				}
+			`;
+			fs.writeFileSync(path.join(extensionsDir, "conflict-q.ts"), extCode);
+
+			const warnSpy = vi.spyOn(logger, "warn").mockImplementation(() => {});
+
+			const result = await loadTestExtensions();
+			const runner = new ExtensionRunner(
+				result.extensions,
+				result.runtime,
+				tempDir.path(),
+				sessionManager,
+				modelRegistry,
+			);
+			const shortcuts = runner.getShortcuts();
+
+			// Contract: ctrl+q is reserved because it is now a default chord for
+			// app.message.followUp. Without this guard, InputController registers
+			// the extension shortcut first and the follow-up handler silently
+			// overwrites it in the editor's custom-key map.
+			expect(warnSpy).toHaveBeenCalledWith(expect.stringContaining("conflicts with built-in"), expect.any(Object));
+			expect(shortcuts.has("ctrl+q")).toBe(false);
+
+			warnSpy.mockRestore();
+		});
+
+		it("rejects Alt+M so it cannot shadow the app.model.select default", async () => {
+			const extCode = `
+				export default function(pi) {
+					pi.registerShortcut("alt+m", {
+						description: "Tries to bind model select",
+						handler: async () => {},
+					});
+				}
+			`;
+			fs.writeFileSync(path.join(extensionsDir, "conflict-model.ts"), extCode);
+
+			const warnSpy = vi.spyOn(logger, "warn").mockImplementation(() => {});
+
+			const result = await loadTestExtensions();
+			const runner = new ExtensionRunner(
+				result.extensions,
+				result.runtime,
+				tempDir.path(),
+				sessionManager,
+				modelRegistry,
+			);
+			const shortcuts = runner.getShortcuts();
+
+			expect(warnSpy).toHaveBeenCalledWith(expect.stringContaining("conflicts with built-in"), expect.any(Object));
+			expect(shortcuts.has("alt+m")).toBe(false);
+
+			warnSpy.mockRestore();
+		});
+
 		it("warns when two extensions register same shortcut", async () => {
 			// Use a non-reserved shortcut
 			const extCode1 = `
diff --git a/packages/coding-agent/test/gallery-cli.test.ts b/packages/coding-agent/test/gallery-cli.test.ts
new file mode 100644
index 000000000..d7087f0b4
--- /dev/null
+++ b/packages/coding-agent/test/gallery-cli.test.ts
@@ -0,0 +1,79 @@
+import { beforeAll, describe, expect, it } from "bun:test";
+import { GALLERY_STATES, renderGalleryState, resolveFixture } from "../src/cli/gallery-cli";
+import type { GalleryFixture } from "../src/cli/gallery-fixtures";
+import { resetSettingsForTest, Settings } from "../src/config/settings";
+import { initTheme } from "../src/modes/theme/theme";
+import { toolRenderers } from "../src/tools/renderers";
+
+beforeAll(async () => {
+	resetSettingsForTest();
+	await Settings.init({ inMemory: true });
+	await initTheme(false, undefined, undefined, "dark", "light");
+});
+
+describe("gallery harness", () => {
+	it("renders every registered tool in every lifecycle state without throwing", async () => {
+		for (const name in toolRenderers) {
+			const fixture = resolveFixture(name);
+			for (const state of GALLERY_STATES) {
+				const lines = await renderGalleryState(name, fixture, state, 100);
+				// A renderer that produces no lines for a state is a regression: the
+				// component should always emit at least the call header or result.
+				expect(lines.length, `${name}/${state} rendered nothing`).toBeGreaterThan(0);
+			}
+		}
+	});
+
+	it("routes each state to the matching args/result (streaming args vs result, success vs error)", async () => {
+		const fixture: GalleryFixture = {
+			label: "Bash",
+			streamingArgs: { command: "echo STREAM_MARK" },
+			args: { command: "echo PROGRESS_MARK" },
+			result: { content: [{ type: "text", text: "SUCCESS_OUT" }], details: { exitCode: 0 } },
+			errorResult: { content: [{ type: "text", text: "ERROR_OUT" }], isError: true, details: { exitCode: 1 } },
+		};
+		const render = async (state: (typeof GALLERY_STATES)[number]) =>
+			Bun.stripANSI((await renderGalleryState("bash", fixture, state, 100)).join("\n"));
+
+		const streaming = await render("streaming");
+		expect(streaming).toContain("STREAM_MARK");
+		expect(streaming).not.toContain("PROGRESS_MARK");
+		expect(streaming).not.toContain("SUCCESS_OUT");
+
+		const progress = await render("progress");
+		expect(progress).toContain("PROGRESS_MARK");
+		expect(progress).not.toContain("SUCCESS_OUT");
+
+		const success = await render("success");
+		expect(success).toContain("SUCCESS_OUT");
+		expect(success).not.toContain("ERROR_OUT");
+
+		const error = await render("error");
+		expect(error).toContain("ERROR_OUT");
+		expect(error).not.toContain("SUCCESS_OUT");
+	});
+
+	it("routes customRendered tools (lsp, task) through the custom-tool branch", async () => {
+		// `lsp`/`task` attach their renderers on the real AgentTool, so the gallery
+		// must reproduce that path. With a result present and mergeCallAndResult, the
+		// custom branch must NOT emit a redundant tool-name line above the result box
+		// (regression guard for tool-execution's custom-branch fallback label).
+		const lsp = resolveFixture("lsp");
+		expect(lsp.customRendered).toBe(true);
+		const lines = await renderGalleryState("lsp", lsp, "error", 100);
+		const stripped = lines.map(line => Bun.stripANSI(line).trim());
+		// The framed result header is present...
+		expect(stripped.some(line => line.includes("LSP references"))).toBe(true);
+		// ...but no standalone "LSP" label line precedes it.
+		expect(stripped).not.toContain("LSP");
+	});
+
+	it("falls back to a generic fixture for registry tools without curated sample data", () => {
+		// resolveFixture never returns undefined for a registry tool, even one
+		// missing from the curated fixtures, so the gallery cannot crash on a newly
+		// added renderer.
+		const fixture = resolveFixture("a-tool-that-has-no-fixture");
+		expect(fixture.args).toBeDefined();
+		expect(fixture.result.content.length).toBeGreaterThan(0);
+	});
+});
diff --git a/packages/coding-agent/test/goals/goal-mode-integration.test.ts b/packages/coding-agent/test/goals/goal-mode-integration.test.ts
index 4eafffe64..83d34f4cc 100644
--- a/packages/coding-agent/test/goals/goal-mode-integration.test.ts
+++ b/packages/coding-agent/test/goals/goal-mode-integration.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
 import * as path from "node:path";
 import { Agent } from "@oh-my-pi/pi-agent-core";
 import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
@@ -25,7 +25,6 @@ function createToolSession(cwd: string, settings: Settings, overrides: Partial<T
 
 type GoalHarness = {
 	tempDir: TempDir;
-	authStorage: AuthStorage;
 	settings: Settings;
 	session: AgentSession;
 	mode: InteractiveMode;
@@ -33,16 +32,35 @@ type GoalHarness = {
 	cleanup: () => Promise<void>;
 };
 
-async function createGoalHarness(): Promise<GoalHarness> {
-	resetSettingsForTest();
-	const tempDir = TempDir.createSync("@pi-goal-mode-");
-	await Settings.init({ inMemory: true, cwd: tempDir.path() });
-	const authStorage = await AuthStorage.create(path.join(tempDir.path(), "testauth.db"));
+// Immutable, expensive fixtures shared across every test. `new ModelRegistry`
+// alone is ~110ms (loads + parses the bundled model catalog), which dominated
+// this file's wall time when rebuilt per test. The registry, its auth storage,
+// and the resolved model are never mutated by goal-mode flows, and
+// AgentSession.dispose() never closes authStorage — so a single shared instance
+// is safe and drops ~8×110ms of pure setup overhead.
+type SharedFixture = {
+	authStorage: AuthStorage;
+	modelRegistry: ModelRegistry;
+	model: NonNullable<ReturnType<ModelRegistry["find"]>>;
+	baseDir: TempDir;
+};
+
+async function createSharedFixture(): Promise<SharedFixture> {
+	const baseDir = TempDir.createSync("@pi-goal-mode-shared-");
+	const authStorage = await AuthStorage.create(path.join(baseDir.path(), "testauth.db"));
 	const modelRegistry = new ModelRegistry(authStorage);
 	const model = modelRegistry.find("anthropic", "claude-sonnet-4-5");
 	if (!model) {
 		throw new Error("Expected claude-sonnet-4-5 to exist in registry");
 	}
+	return { authStorage, modelRegistry, model, baseDir };
+}
+
+async function createGoalHarness(shared: SharedFixture): Promise<GoalHarness> {
+	resetSettingsForTest();
+	const tempDir = TempDir.createSync("@pi-goal-mode-");
+	await Settings.init({ inMemory: true, cwd: tempDir.path() });
+	const { modelRegistry, model } = shared;
 
 	const settings = Settings.isolated({
 		"compaction.enabled": false,
@@ -77,7 +95,6 @@ async function createGoalHarness(): Promise<GoalHarness> {
 
 	return {
 		tempDir,
-		authStorage,
 		settings,
 		session,
 		mode,
@@ -85,7 +102,6 @@ async function createGoalHarness(): Promise<GoalHarness> {
 		cleanup: async () => {
 			mode.stop();
 			await session.dispose();
-			authStorage.close();
 			tempDir.removeSync();
 			resetSettingsForTest();
 		},
@@ -98,13 +114,20 @@ async function toolNamesFor(harness: GoalHarness): Promise<string[]> {
 
 describe("InteractiveMode goal mode integration", () => {
 	let harness: GoalHarness;
+	let shared: SharedFixture;
 
-	beforeAll(() => {
+	beforeAll(async () => {
 		initTheme();
+		shared = await createSharedFixture();
+	});
+
+	afterAll(() => {
+		shared.authStorage.close();
+		shared.baseDir.removeSync();
 	});
 
 	beforeEach(async () => {
-		harness = await createGoalHarness();
+		harness = await createGoalHarness(shared);
 	});
 
 	afterEach(async () => {
diff --git a/packages/coding-agent/test/input-controller-escape.test.ts b/packages/coding-agent/test/input-controller-escape.test.ts
index 5231a1e3f..f458d9568 100644
--- a/packages/coding-agent/test/input-controller-escape.test.ts
+++ b/packages/coding-agent/test/input-controller-escape.test.ts
@@ -1,6 +1,7 @@
 import { describe, expect, it, type Mock, vi } from "bun:test";
 import { InputController } from "@oh-my-pi/pi-coding-agent/modes/controllers/input-controller";
 import type { InteractiveModeContext, SubmittedUserInput } from "@oh-my-pi/pi-coding-agent/modes/types";
+import { USER_INTERRUPT_LABEL } from "@oh-my-pi/pi-coding-agent/session/messages";
 
 type Spy = Mock<(...args: unknown[]) => unknown>;
 type StartPendingSubmissionSpy = Mock<InteractiveModeContext["startPendingSubmission"]>;
@@ -34,10 +35,12 @@ type FakeEditor = {
 function createSubmission(input: {
 	text: string;
 	images?: InteractiveModeContext["pendingImages"];
+	imageLinks?: InteractiveModeContext["pendingImageLinks"];
 }): SubmittedUserInput {
 	return {
 		text: input.text,
 		images: input.images,
+		imageLinks: input.imageLinks,
 		cancelled: false,
 		started: false,
 	};
@@ -80,10 +83,16 @@ function createContext(): {
 	const hasActiveBtw = vi.fn(() => false);
 	const handleOmfgEscape = vi.fn(() => true);
 	const hasActiveOmfg = vi.fn(() => false);
-	const startPendingSubmission = vi.fn((input: { text: string; images?: InteractiveModeContext["pendingImages"] }) => {
-		ensureLoadingAnimation();
-		return createSubmission(input);
-	});
+	const startPendingSubmission = vi.fn(
+		(input: {
+			text: string;
+			images?: InteractiveModeContext["pendingImages"];
+			imageLinks?: InteractiveModeContext["pendingImageLinks"];
+		}) => {
+			ensureLoadingAnimation();
+			return createSubmission(input);
+		},
+	);
 	const editor: FakeEditor = {
 		setText(text: string) {
 			editorText = text;
@@ -104,7 +113,11 @@ function createContext(): {
 
 	ctx = {
 		editor: editor as unknown as InteractiveModeContext["editor"],
-		ui: { requestRender } as unknown as InteractiveModeContext["ui"],
+		ui: {
+			requestRender,
+			addInputListener: vi.fn(),
+			addStartListener: vi.fn(),
+		} as unknown as InteractiveModeContext["ui"],
 		loadingAnimation: undefined,
 		autoCompactionLoader: undefined,
 		retryLoader: undefined,
@@ -132,6 +145,7 @@ function createContext(): {
 			getKeys: () => [],
 		} as unknown as InteractiveModeContext["keybindings"],
 		pendingImages: [],
+		pendingImageLinks: [],
 		isBashMode: false,
 		isPythonMode: false,
 		optimisticUserMessageSignature: undefined,
@@ -197,7 +211,11 @@ describe("InputController escape behavior", () => {
 		controller.setupEditorSubmitHandler();
 		await editor.onSubmit?.("hello");
 
-		expect(spies.startPendingSubmission).toHaveBeenCalledWith({ text: "hello", images: undefined });
+		expect(spies.startPendingSubmission).toHaveBeenCalledWith({
+			text: "hello",
+			images: undefined,
+			imageLinks: undefined,
+		});
 		expect(spies.onInputCallback).toHaveBeenCalledWith(submission);
 
 		editor.onEscape?.();
@@ -232,6 +250,9 @@ describe("InputController escape behavior", () => {
 		expect(spies.cancelPendingSubmission).toHaveBeenCalledTimes(1);
 		expect(spies.clearQueue).toHaveBeenCalledTimes(1);
 		expect(spies.abort).toHaveBeenCalledTimes(1);
+		// The Esc interrupt threads a user-facing reason so the aborted turn and its
+		// synthetic tool results read as a deliberate interrupt, not "Request was aborted".
+		expect(spies.abort).toHaveBeenCalledWith({ reason: USER_INTERRUPT_LABEL });
 	});
 
 	it("prefers aborting bash before aborting an overlapping stream", () => {
diff --git a/packages/coding-agent/test/input-controller-keybindings.test.ts b/packages/coding-agent/test/input-controller-keybindings.test.ts
index 4cc6aa421..7c773da8f 100644
--- a/packages/coding-agent/test/input-controller-keybindings.test.ts
+++ b/packages/coding-agent/test/input-controller-keybindings.test.ts
@@ -6,6 +6,7 @@ type FakeEditor = {
 	onEscape?: () => void;
 	onClear?: () => void;
 	onExit?: () => void;
+	onDisplayReset?: () => void;
 	onSuspend?: () => void;
 	onCycleThinkingLevel?: () => void;
 	onCycleModelForward?: () => void;
@@ -20,23 +21,41 @@ type FakeEditor = {
 	onExternalEditor?: () => void;
 	onDequeue?: () => void;
 	onChange?: (text: string) => void;
+	onSubmit?: (text: string) => Promise<void>;
 	setText(text: string): void;
 	getText(): string;
 	addToHistory(text: string): void;
 	setActionKeys(action: string, keys: string[]): void;
 	setCustomKeyHandler(key: string, handler: () => void): void;
 	clearCustomKeyHandlers(): void;
+	pasteText(text: string): void;
 };
 
 async function createContext() {
 	let editorText = "";
 	const keyMap: Record<string, string[]> = {
+		"app.display.reset": ["ctrl+l"],
 		"app.model.selectTemporary": ["ctrl+y"],
-		"app.model.select": ["ctrl+l"],
+		"app.model.select": ["alt+m"],
 	};
+	const customHandlers = new Map<string, () => void>();
 	const setActionKeys = vi.fn();
+	const setCustomKeyHandler = vi.fn((key: string, handler: () => void) => {
+		customHandlers.set(key, handler);
+	});
+	const clearCustomKeyHandlers = vi.fn(() => {
+		customHandlers.clear();
+	});
+	const resetDisplay = vi.fn();
 	const showModelSelector = vi.fn();
+	const requestRender = vi.fn();
+	const addInputListener = vi.fn();
+	const addStartListener = vi.fn();
+	const terminalWrite = vi.fn();
 	const prompt = vi.fn(async () => {});
+	const abort = vi.fn(async () => {});
+	const interruptAndFlushQueuedMessages = vi.fn(async () => {});
+	const getQueuedMessages = vi.fn(() => ({ steering: [] as string[], followUp: [] as string[] }));
 	const updatePendingMessagesDisplay = vi.fn();
 	const editor: FakeEditor = {
 		setText(text: string) {
@@ -46,13 +65,22 @@ async function createContext() {
 			return editorText;
 		},
 		addToHistory: vi.fn(),
+		pasteText(text: string) {
+			editorText += text;
+		},
 		setActionKeys,
-		setCustomKeyHandler: vi.fn(),
-		clearCustomKeyHandlers: vi.fn(),
+		setCustomKeyHandler,
+		clearCustomKeyHandlers,
 	};
 	const ctx = {
 		editor: editor as unknown as InteractiveModeContext["editor"],
-		ui: { requestRender: vi.fn() } as unknown as InteractiveModeContext["ui"],
+		ui: {
+			requestRender,
+			resetDisplay,
+			addInputListener,
+			addStartListener,
+			terminal: { write: terminalWrite },
+		} as unknown as InteractiveModeContext["ui"],
 		loadingAnimation: undefined,
 		autoCompactionLoader: undefined,
 		retryLoader: undefined,
@@ -66,6 +94,10 @@ async function createContext() {
 			isEvalRunning: false,
 			extensionRunner: undefined,
 			prompt,
+			queuedMessageCount: 0,
+			getQueuedMessages,
+			abort,
+			interruptAndFlushQueuedMessages,
 		} as unknown as InteractiveModeContext["session"],
 		keybindings: {
 			getKeys(action: string) {
@@ -122,33 +154,61 @@ async function createContext() {
 		InputController,
 		ctx,
 		editor,
+		customHandlers,
 		spies: {
 			setActionKeys,
 			showModelSelector,
 			prompt,
 			updatePendingMessagesDisplay,
+			requestRender,
+			abort,
+			interruptAndFlushQueuedMessages,
+			getQueuedMessages,
+			resetDisplay,
 		},
 	};
 }
 
 describe("InputController keybinding setup", () => {
-	it("registers temporary and persisted model selector actions separately", async () => {
+	it("registers model selector and display reset actions separately", async () => {
 		const { InputController, ctx, editor, spies } = await createContext();
 		const controller = new InputController(ctx);
 
 		controller.setupKeyHandlers();
 
+		expect(spies.setActionKeys).toHaveBeenCalledWith("app.display.reset", ["ctrl+l"]);
 		expect(spies.setActionKeys).toHaveBeenCalledWith("app.model.selectTemporary", ["ctrl+y"]);
-		expect(spies.setActionKeys).toHaveBeenCalledWith("app.model.select", ["ctrl+l"]);
+		expect(spies.setActionKeys).toHaveBeenCalledWith("app.model.select", ["alt+m"]);
+		expect(editor.onDisplayReset).toBeDefined();
 		expect(editor.onSelectModelTemporary).toBeDefined();
 		expect(editor.onSelectModel).toBeDefined();
 		expect(editor.onSelectModelTemporary).not.toBe(editor.onSelectModel);
 
+		editor.onDisplayReset?.();
 		editor.onSelectModelTemporary?.();
 		editor.onSelectModel?.();
 
 		expect(spies.showModelSelector).toHaveBeenNthCalledWith(1, { temporaryOnly: true });
 		expect(spies.showModelSelector).toHaveBeenNthCalledWith(2);
+		expect(spies.resetDisplay).toHaveBeenCalledTimes(1);
+	});
+
+	it("empty Enter interrupts and sends a queued steering message", async () => {
+		const { InputController, ctx, editor, spies } = await createContext();
+		const session = ctx.session as unknown as { isStreaming: boolean; queuedMessageCount: number };
+		session.isStreaming = true;
+		session.queuedMessageCount = 1;
+		spies.getQueuedMessages.mockReturnValue({ steering: ["Send this now"], followUp: [] });
+		const controller = new InputController(ctx);
+
+		controller.setupEditorSubmitHandler();
+		await editor.onSubmit?.("");
+
+		expect(spies.interruptAndFlushQueuedMessages).toHaveBeenCalledWith({ reason: "Interrupted by user" });
+		expect(spies.abort).not.toHaveBeenCalled();
+		expect(spies.updatePendingMessagesDisplay).toHaveBeenCalledTimes(1);
+		expect(spies.requestRender).toHaveBeenCalledTimes(1);
+		expect(spies.prompt).not.toHaveBeenCalled();
 	});
 
 	it("marks streaming follow-up submissions as local", async () => {
diff --git a/packages/coding-agent/test/input-controller-skill-queue.test.ts b/packages/coding-agent/test/input-controller-skill-queue.test.ts
index 1ec65d90b..93dbba9ef 100644
--- a/packages/coding-agent/test/input-controller-skill-queue.test.ts
+++ b/packages/coding-agent/test/input-controller-skill-queue.test.ts
@@ -94,7 +94,7 @@ function createStubInputControllerContext(opts: { skillCommands: Map<string, str
 		isBashMode: false,
 		isPythonMode: false,
 		pendingImages: [],
-		isBackgrounded: false,
+		pendingImageLinks: [],
 		loopModeEnabled: false,
 		compactionQueuedMessages: [],
 		locallySubmittedUserSignatures: new Set<string>(),
diff --git a/packages/coding-agent/test/interactive-mode-plan-review.test.ts b/packages/coding-agent/test/interactive-mode-plan-review.test.ts
index 6fb7b28eb..30086e8e3 100644
--- a/packages/coding-agent/test/interactive-mode-plan-review.test.ts
+++ b/packages/coding-agent/test/interactive-mode-plan-review.test.ts
@@ -6,8 +6,7 @@ import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config
 import { resolveLocalUrlToPath } from "@oh-my-pi/pi-coding-agent/internal-urls";
 import { AssistantMessageComponent } from "@oh-my-pi/pi-coding-agent/modes/components/assistant-message";
 import { initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
-import { SILENT_ABORT_MARKER } from "@oh-my-pi/pi-coding-agent/session/messages";
-import { Text } from "@oh-my-pi/pi-tui";
+import { SILENT_ABORT_MARKER, USER_INTERRUPT_LABEL } from "@oh-my-pi/pi-coding-agent/session/messages";
 import { formatNumber, TempDir } from "@oh-my-pi/pi-utils";
 import { ModelRegistry } from "../src/config/model-registry";
 import type { HookSelectorSlider } from "../src/modes/components/hook-selector";
@@ -68,7 +67,6 @@ describe("InteractiveMode plan review rendering", () => {
 	});
 
 	beforeEach(async () => {
-		Bun.gc(true);
 		resetSettingsForTest();
 		tempDir = TempDir.createSync("@pi-plan-review-");
 		await Settings.init({ inMemory: true, cwd: tempDir.path() });
@@ -111,10 +109,9 @@ describe("InteractiveMode plan review rendering", () => {
 		currentAuthStorage?.close();
 		currentTempDir?.removeSync();
 		resetSettingsForTest();
-		Bun.gc(true);
 	});
 
-	it("appends each submitted plan review preview to preserve scrollback", async () => {
+	it("forwards each submitted plan to the review overlay", async () => {
 		const planFilePath = "local://PLAN.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
@@ -124,38 +121,119 @@ describe("InteractiveMode plan review rendering", () => {
 
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Refine plan");
+		const review = vi.spyOn(mode, "showPlanReview").mockResolvedValue("Refine plan");
 
 		await mode.handlePlanApproval({
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath: "local://PLAN.md",
 		});
 
-		const firstPreview = mode.chatContainer.children.at(-1);
-		expect(firstPreview).toBeDefined();
-		expect(firstPreview!.render(120).join("\n")).toContain("First plan");
+		expect(review.mock.calls[0]?.[0]).toContain("First plan");
 
-		const marker = new Text("MARKER", 0, 0);
-		mode.chatContainer.addChild(marker);
 		await Bun.write(resolvedPlanPath, "# Second plan\n\nbeta");
 
 		await mode.handlePlanApproval({
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath: "local://PLAN.md",
 		});
 
-		const secondPreview = mode.chatContainer.children.at(-1);
-		expect(secondPreview).toBeDefined();
-		expect(secondPreview).not.toBe(firstPreview);
-		expect(mode.chatContainer.children.at(-2)).toBe(marker);
-		expect(mode.chatContainer.children.at(-3)).toBe(firstPreview);
-		expect(firstPreview!.render(120).join("\n")).toContain("First plan");
-		expect(firstPreview!.render(120).join("\n")).not.toContain("Second plan");
-		expect(secondPreview!.render(120).join("\n")).toContain("Second plan");
+		// Each approval shows the current plan in the overlay, not a stale one.
+		expect(review.mock.calls[1]?.[0]).toContain("Second plan");
+		expect(review.mock.calls[1]?.[0]).not.toContain("First plan");
+	});
+
+	it("re-prompts the model with annotation feedback when Refine is chosen", async () => {
+		const planFilePath = "local://PLAN.md";
+		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
+			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
+			getSessionId: () => session.sessionManager.getSessionId(),
+		});
+		await Bun.write(resolvedPlanPath, "# Plan\n\nbody");
+
+		mode.planModeEnabled = true;
+		mode.planModePlanFilePath = planFilePath;
+		const feedback = "Refinement feedback on the plan:\n\n## Goal\n- needs detail\n";
+		// The overlay reports annotation feedback through onFeedbackChange before the
+		// operator picks "Refine plan".
+		vi.spyOn(mode, "showPlanReview").mockImplementation(async (_plan, _title, _options, dialogOptions) => {
+			dialogOptions?.onFeedbackChange?.(feedback);
+			return "Refine plan";
+		});
+		const startSpy = vi
+			.spyOn(mode, "startPendingSubmission")
+			.mockReturnValue({ text: feedback, cancelled: false, started: false });
+		const onInput = vi.fn();
+		mode.onInputCallback = onInput;
+
+		await mode.handlePlanApproval({
+			planFilePath,
+			planExists: true,
+			title: "PLAN",
+		});
+
+		expect(startSpy).toHaveBeenCalledWith(expect.objectContaining({ text: expect.stringContaining("needs detail") }));
+		expect(onInput).toHaveBeenCalledTimes(1);
+	});
+
+	it("Refine with no annotations does not re-prompt the model", async () => {
+		const planFilePath = "local://PLAN.md";
+		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
+			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
+			getSessionId: () => session.sessionManager.getSessionId(),
+		});
+		await Bun.write(resolvedPlanPath, "# Plan\n\nbody");
+
+		mode.planModeEnabled = true;
+		mode.planModePlanFilePath = planFilePath;
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Refine plan");
+		const startSpy = vi.spyOn(mode, "startPendingSubmission");
+		const onInput = vi.fn();
+		mode.onInputCallback = onInput;
+
+		await mode.handlePlanApproval({
+			planFilePath,
+			planExists: true,
+			title: "PLAN",
+		});
+
+		expect(startSpy).not.toHaveBeenCalled();
+		expect(onInput).not.toHaveBeenCalled();
+	});
+
+	it("approves with in-overlay edits and mirrors them to the plan file", async () => {
+		const planFilePath = "local://PLAN.md";
+		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
+			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
+			getSessionId: () => session.sessionManager.getSessionId(),
+		});
+		await Bun.write(resolvedPlanPath, "# Plan\n\noriginal body\n");
+
+		mode.planModeEnabled = true;
+		mode.planModePlanFilePath = planFilePath;
+		const edited = "# Plan\n\nedited body\n";
+		vi.spyOn(mode, "showPlanReview").mockImplementation(async (_plan, _title, _options, dialogOptions) => {
+			dialogOptions?.onPlanEdited?.(edited);
+			return "Approve and execute";
+		});
+		vi.spyOn(mode, "handleClearCommand").mockResolvedValue();
+		const promptSpy = vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
+
+		await mode.handlePlanApproval({
+			planFilePath,
+			planExists: true,
+			title: "PLAN",
+		});
+
+		// The synthetic plan-approved prompt carries the in-overlay edit, not the
+		// stale on-disk content (preferring editedContent avoids the write race).
+		const call = promptSpy.mock.calls.find(isPlanApprovedCall);
+		expect(call).toBeDefined();
+		expect(call?.[0] as string).toContain("edited body");
+		expect(call?.[0] as string).not.toContain("original body");
+		// onPlanEdited mirrored the edit to the plan file.
+		expect(await Bun.file(resolvedPlanPath).text()).toContain("edited body");
 	});
 
 	it("offers approve-and-keep-context as a distinct plan approval path", async () => {
@@ -169,16 +247,16 @@ describe("InteractiveMode plan review rendering", () => {
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
 		vi.spyOn(session, "getContextUsage").mockReturnValue({ tokens: 7320, contextWindow: 10000, percent: 73.2 });
-		const selector = vi.spyOn(mode, "showHookSelector").mockResolvedValue("Refine plan");
+		const selector = vi.spyOn(mode, "showPlanReview").mockResolvedValue("Refine plan");
 
 		await mode.handlePlanApproval({
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath: "local://APPROVED.md",
 		});
 
 		expect(selector).toHaveBeenCalledWith(
+			expect.any(String),
 			"Plan mode - next step",
 			[
 				"Approve and execute",
@@ -246,17 +324,16 @@ describe("InteractiveMode plan review rendering", () => {
 				percent: (tokens / contextWindow) * 100,
 			};
 		});
-		const selector = vi.spyOn(mode, "showHookSelector").mockResolvedValue("Refine plan");
+		const selector = vi.spyOn(mode, "showPlanReview").mockResolvedValue("Refine plan");
 
 		await mode.handlePlanApproval({
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath: "local://APPROVED.md",
 		});
 
 		expect(contextSpy).toHaveBeenCalledWith({ contextWindow: executionModel.contextWindow });
-		expect(selector.mock.calls[0]?.[1]).toEqual([
+		expect(selector.mock.calls[0]?.[2]).toEqual([
 			"Approve and execute",
 			"Approve and compact context",
 			`Approve and keep context (~${compactNumber(tokens)} / ${compactNumber(executionModel.contextWindow)})`,
@@ -275,16 +352,15 @@ describe("InteractiveMode plan review rendering", () => {
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
 		vi.spyOn(session, "getContextUsage").mockReturnValue({ tokens: 9600, contextWindow: 10000, percent: 96 });
-		const selector = vi.spyOn(mode, "showHookSelector").mockResolvedValue("Refine plan");
+		const selector = vi.spyOn(mode, "showPlanReview").mockResolvedValue("Refine plan");
 
 		await mode.handlePlanApproval({
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath: "local://APPROVED.md",
 		});
 
-		expect(selector.mock.calls[0]?.[2]).toEqual(
+		expect(selector.mock.calls[0]?.[3]).toEqual(
 			expect.objectContaining({
 				disabledIndices: [2],
 			}),
@@ -302,16 +378,15 @@ describe("InteractiveMode plan review rendering", () => {
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
 		vi.spyOn(session, "getContextUsage").mockReturnValue({ tokens: 9500, contextWindow: 10000, percent: 95 });
-		const selector = vi.spyOn(mode, "showHookSelector").mockResolvedValue("Refine plan");
+		const selector = vi.spyOn(mode, "showPlanReview").mockResolvedValue("Refine plan");
 
 		await mode.handlePlanApproval({
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath: "local://APPROVED.md",
 		});
 
-		expect(selector.mock.calls[0]?.[2]).toEqual(
+		expect(selector.mock.calls[0]?.[3]).toEqual(
 			expect.objectContaining({
 				disabledIndices: undefined,
 			}),
@@ -330,16 +405,16 @@ describe("InteractiveMode plan review rendering", () => {
 		mode.planModePlanFilePath = planFilePath;
 		// Post-compaction: tokens unknown until the next LLM response.
 		vi.spyOn(session, "getContextUsage").mockReturnValue({ tokens: null, contextWindow: 200000, percent: null });
-		const selector = vi.spyOn(mode, "showHookSelector").mockResolvedValue("Refine plan");
+		const selector = vi.spyOn(mode, "showPlanReview").mockResolvedValue("Refine plan");
 
 		await mode.handlePlanApproval({
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath: "local://APPROVED.md",
 		});
 
 		expect(selector).toHaveBeenCalledWith(
+			expect.any(String),
 			"Plan mode - next step",
 			["Approve and execute", "Approve and compact context", "Approve and keep context", "Refine plan"],
 			expect.any(Object),
@@ -349,12 +424,11 @@ describe("InteractiveMode plan review rendering", () => {
 
 	it("approves a plan without clearing the session when keeping context", async () => {
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
 		});
-		const resolvedFinalPlanPath = resolveLocalUrlToPath(finalPlanFilePath, {
+		const resolvedFinalPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
 		});
@@ -363,7 +437,7 @@ describe("InteractiveMode plan review rendering", () => {
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
 		vi.spyOn(session, "getContextUsage").mockReturnValue({ tokens: null, contextWindow: 200000, percent: null });
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and keep context");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and keep context");
 		const clear = vi.spyOn(mode, "handleClearCommand").mockResolvedValue();
 		const prompt = vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
 
@@ -371,7 +445,6 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 
 		expect(clear).not.toHaveBeenCalled();
@@ -383,7 +456,6 @@ describe("InteractiveMode plan review rendering", () => {
 
 	it("keeps the existing approve-and-execute path clearing the session", async () => {
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -392,7 +464,7 @@ describe("InteractiveMode plan review rendering", () => {
 
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and execute");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and execute");
 		const clear = vi.spyOn(mode, "handleClearCommand").mockResolvedValue();
 		const prompt = vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
 
@@ -400,7 +472,6 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 
 		expect(clear).toHaveBeenCalledTimes(1);
@@ -428,7 +499,6 @@ describe("InteractiveMode plan review rendering", () => {
 		session.settings.setModelRole("plan", "anthropic/claude-sonnet-4-5");
 
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -445,8 +515,8 @@ describe("InteractiveMode plan review rendering", () => {
 		vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
 
 		let observedSegments: string[] = [];
-		vi.spyOn(mode, "showHookSelector").mockImplementation(
-			async (_title, _options, _dialogOptions, extra?: { slider?: HookSelectorSlider }) => {
+		vi.spyOn(mode, "showPlanReview").mockImplementation(
+			async (_planContent, _title, _options, _dialogOptions, extra?: { slider?: HookSelectorSlider }) => {
 				const slider = extra?.slider;
 				expect(slider).toBeDefined();
 				observedSegments = slider!.segments.map(segment => segment.label);
@@ -462,7 +532,6 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 
 		expect(observedSegments).toEqual(["default", "slow"]);
@@ -473,7 +542,6 @@ describe("InteractiveMode plan review rendering", () => {
 
 	it("re-enters plan mode on the approved titled artifact after approve-and-execute", async () => {
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -482,7 +550,7 @@ describe("InteractiveMode plan review rendering", () => {
 
 		await mode.handlePlanModeCommand();
 
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and execute");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and execute");
 		vi.spyOn(mode, "handleClearCommand").mockResolvedValue();
 		vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
 
@@ -490,23 +558,21 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "APPROVED",
-			finalPlanFilePath,
 		});
 
 		expect(mode.planModeEnabled).toBe(false);
-		expect(session.getPlanReferencePath()).toBe(finalPlanFilePath);
+		expect(session.getPlanReferencePath()).toBe(planFilePath);
 
 		await mode.handlePlanModeCommand();
 		expect(session.getPlanModeState()).toMatchObject({
 			enabled: true,
-			planFilePath: finalPlanFilePath,
+			planFilePath,
 			reentry: true,
 		});
 	});
 
 	it("Approve and compact context: ok outcome dispatches plan-approved after compaction", async () => {
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -515,7 +581,7 @@ describe("InteractiveMode plan review rendering", () => {
 
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and compact context");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and compact context");
 		const compactSpy = vi.spyOn(mode, "handleCompactCommand").mockResolvedValue("ok");
 		const markSentSpy = vi.spyOn(session, "markPlanReferenceSent");
 		const promptSpy = vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
@@ -524,14 +590,13 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 
 		// Compaction was run with the rendered planning-specific custom instruction.
 		expect(compactSpy).toHaveBeenCalledTimes(1);
 		const [compactInstruction] = compactSpy.mock.calls[0]!;
 		expect(typeof compactInstruction).toBe("string");
-		expect(compactInstruction as string).toContain(finalPlanFilePath);
+		expect(compactInstruction as string).toContain(planFilePath);
 
 		// Plan-approved synthetic prompt was dispatched.
 		const planApprovedIdx = promptSpy.mock.calls.findIndex(isPlanApprovedCall);
@@ -549,7 +614,6 @@ describe("InteractiveMode plan review rendering", () => {
 		// CompactionOutcome boundary; the underlying executeCompaction → sentinel
 		// classification path is producer-layer and not under T3's contract.)
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -558,7 +622,7 @@ describe("InteractiveMode plan review rendering", () => {
 
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and compact context");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and compact context");
 		vi.spyOn(mode, "handleCompactCommand").mockResolvedValue("cancelled");
 		const showWarningSpy = vi.spyOn(mode, "showWarning");
 		const setPlanRefSpy = vi.spyOn(session, "setPlanReferencePath");
@@ -569,7 +633,6 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 
 		// Operator was told the dispatch was deferred.
@@ -578,7 +641,7 @@ describe("InteractiveMode plan review rendering", () => {
 		);
 		// Plan reference path was recorded so the session knows about the approved
 		// plan at its final destination …
-		expect(setPlanRefSpy).toHaveBeenCalledWith(finalPlanFilePath);
+		expect(setPlanRefSpy).toHaveBeenCalledWith(planFilePath);
 		// … but markPlanReferenceSent was NOT called, so the next operator turn
 		// will inject the reference fresh via #buildPlanReferenceMessage. This is
 		// the load-bearing assertion that the cancel path leaves the executor
@@ -592,7 +655,6 @@ describe("InteractiveMode plan review rendering", () => {
 		// Mock `handleCompactCommand` to surface the "failed" outcome directly.
 		// Failure → approval intent stands → synthetic dispatch fires.
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -601,7 +663,7 @@ describe("InteractiveMode plan review rendering", () => {
 
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and compact context");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and compact context");
 		vi.spyOn(mode, "handleCompactCommand").mockResolvedValue("failed");
 		const markSentSpy = vi.spyOn(session, "markPlanReferenceSent");
 		const promptSpy = vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
@@ -610,7 +672,6 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 
 		// Plan-approved synthetic prompt WAS dispatched despite the failure.
@@ -625,7 +686,6 @@ describe("InteractiveMode plan review rendering", () => {
 		// hit #buildPlanReferenceMessage with the stale plan-mode path. Pin it
 		// before the compaction await.
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -634,13 +694,13 @@ describe("InteractiveMode plan review rendering", () => {
 
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and compact context");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and compact context");
 		vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
 
 		const setPlanRefSpy = vi.spyOn(session, "setPlanReferencePath");
 		let planRefSetWhenCompactionRan = false;
 		vi.spyOn(mode, "handleCompactCommand").mockImplementation(async () => {
-			planRefSetWhenCompactionRan = setPlanRefSpy.mock.calls.some(call => call[0] === finalPlanFilePath);
+			planRefSetWhenCompactionRan = setPlanRefSpy.mock.calls.some(call => call[0] === planFilePath);
 			return "ok";
 		});
 
@@ -648,7 +708,6 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 
 		// The contract: by the time handleCompactCommand runs (and flushes the
@@ -678,7 +737,6 @@ describe("InteractiveMode plan review rendering", () => {
 		throwError?: Error,
 	): Promise<void> {
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -687,7 +745,7 @@ describe("InteractiveMode plan review rendering", () => {
 
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and compact context");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and compact context");
 		if (compactOutcome === "throw") {
 			vi.spyOn(mode, "handleCompactCommand").mockRejectedValue(throwError ?? new Error("compact boom"));
 		} else {
@@ -699,7 +757,6 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 	}
 
@@ -737,7 +794,6 @@ describe("InteractiveMode plan review rendering", () => {
 
 	it("B5: Approve and execute (no compact) → markPlanCompactAbortPending never called; flag stays false", async () => {
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -745,7 +801,7 @@ describe("InteractiveMode plan review rendering", () => {
 		await Bun.write(resolvedPlanPath, "# Plan\n\nBody.");
 		mode.planModeEnabled = true;
 		mode.planModePlanFilePath = planFilePath;
-		vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and execute");
+		vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and execute");
 		const markSpy = vi.spyOn(session, "markPlanCompactAbortPending");
 		vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
 
@@ -753,7 +809,6 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "PLAN",
-			finalPlanFilePath,
 		});
 
 		expect(markSpy).not.toHaveBeenCalled();
@@ -762,7 +817,6 @@ describe("InteractiveMode plan review rendering", () => {
 
 	it("re-enters plan mode on the approved titled artifact after approval", async () => {
 		const planFilePath = "local://PLAN.md";
-		const finalPlanFilePath = "local://APPROVED.md";
 		const resolvedPlanPath = resolveLocalUrlToPath(planFilePath, {
 			getArtifactsDir: () => session.sessionManager.getArtifactsDir(),
 			getSessionId: () => session.sessionManager.getSessionId(),
@@ -773,7 +827,7 @@ describe("InteractiveMode plan review rendering", () => {
 		expect(session.getPlanModeState()?.planFilePath).toBe(planFilePath);
 
 		vi.spyOn(session, "getContextUsage").mockReturnValue({ tokens: null, contextWindow: 200000, percent: null });
-		const selector = vi.spyOn(mode, "showHookSelector").mockResolvedValue("Approve and keep context");
+		const selector = vi.spyOn(mode, "showPlanReview").mockResolvedValue("Approve and keep context");
 		const showError = vi.spyOn(mode, "showError");
 		vi.spyOn(session, "prompt").mockResolvedValue(undefined as never);
 
@@ -781,24 +835,22 @@ describe("InteractiveMode plan review rendering", () => {
 			planFilePath,
 			planExists: true,
 			title: "APPROVED",
-			finalPlanFilePath,
 		});
 
 		expect(mode.planModeEnabled).toBe(false);
-		expect(session.getPlanReferencePath()).toBe(finalPlanFilePath);
+		expect(session.getPlanReferencePath()).toBe(planFilePath);
 
 		await mode.handlePlanModeCommand();
 		expect(session.getPlanModeState()).toMatchObject({
 			enabled: true,
-			planFilePath: finalPlanFilePath,
+			planFilePath,
 			reentry: true,
 		});
 
 		await mode.handlePlanApproval({
-			planFilePath: finalPlanFilePath,
+			planFilePath,
 			planExists: true,
 			title: "APPROVED",
-			finalPlanFilePath,
 		});
 
 		expect(selector).toHaveBeenCalledTimes(2);
@@ -808,9 +860,10 @@ describe("InteractiveMode plan review rendering", () => {
 	// ==========================================================================
 	// Phase 6 — D layer: replay-side render branches in AssistantMessageComponent.
 	//
-	// D1 asserts that the persisted `SILENT_ABORT_MARKER` suppresses the red
-	// "Operation aborted" line. D2 is the over-suppression regression guard —
-	// an aborted message with NO marker must still render the line.
+	// D1 asserts that the persisted `SILENT_ABORT_MARKER` suppresses the red abort
+	// line. D2 is the over-suppression regression guard — an aborted message with
+	// NO marker still renders the generic label. D3 covers a threaded interrupt
+	// reason rendering verbatim.
 	// ==========================================================================
 
 	function renderAssistant(message: AssistantMessage, width = 120): string {
@@ -840,20 +893,30 @@ describe("InteractiveMode plan review rendering", () => {
 		};
 	}
 
-	it("D1: Replay of an assistant message with SILENT_ABORT_MARKER + aborted: rendered component contains no /Operation aborted/", () => {
+	it("D1: Replay of an assistant message with SILENT_ABORT_MARKER + aborted: rendered component contains no abort line", () => {
 		const message = buildAbortedAssistantMessage({ errorMessage: SILENT_ABORT_MARKER });
 		const rendered = renderAssistant(message);
-		expect(rendered).not.toMatch(/Operation aborted/);
+		expect(rendered).not.toContain("Operation aborted");
+		expect(rendered).not.toContain(USER_INTERRUPT_LABEL);
 		// The marker itself MUST NOT leak into rendered output either.
 		expect(rendered).not.toContain(SILENT_ABORT_MARKER);
 	});
 
-	it("D2: Replay of an aborted message with no marker + empty content: rendered component DOES contain 'Operation aborted'", () => {
+	it("D2: Replay of an aborted message with no threaded reason + empty content: rendered component DOES contain the generic label", () => {
 		// Over-suppression regression guard: silent path is opt-in via the
-		// persisted marker. A user-cancel abort with no marker and no content
-		// still surfaces the standard label.
+		// persisted marker. An abort with no marker and no threaded reason still
+		// surfaces the generic operator-facing label.
 		const message = buildAbortedAssistantMessage({ content: [], errorMessage: undefined });
 		const rendered = renderAssistant(message);
 		expect(rendered).toContain("Operation aborted");
 	});
+
+	it("D3: Replay of an aborted message carrying a user-interrupt reason renders it verbatim", () => {
+		// The Esc-interrupt reason persisted on errorMessage must render as-is,
+		// not collapse into the generic label.
+		const message = buildAbortedAssistantMessage({ content: [], errorMessage: USER_INTERRUPT_LABEL });
+		const rendered = renderAssistant(message);
+		expect(rendered).toContain(USER_INTERRUPT_LABEL);
+		expect(rendered).not.toContain("Operation aborted");
+	});
 });
diff --git a/packages/coding-agent/test/interactive-mode-shutdown.test.ts b/packages/coding-agent/test/interactive-mode-shutdown.test.ts
deleted file mode 100644
index 3d010d68d..000000000
--- a/packages/coding-agent/test/interactive-mode-shutdown.test.ts
+++ /dev/null
@@ -1,67 +0,0 @@
-import { afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
-import * as path from "node:path";
-import { Agent } from "@oh-my-pi/pi-agent-core";
-import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
-import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
-import { InteractiveMode } from "@oh-my-pi/pi-coding-agent/modes/interactive-mode";
-import { initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
-import { AgentSession } from "@oh-my-pi/pi-coding-agent/session/agent-session";
-import { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
-import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
-import { postmortem, TempDir } from "@oh-my-pi/pi-utils";
-
-describe("InteractiveMode shutdown", () => {
-	let authStorage: AuthStorage;
-	let mode: InteractiveMode;
-	let session: AgentSession;
-	let tempDir: TempDir;
-
-	beforeAll(() => {
-		initTheme();
-	});
-
-	beforeEach(async () => {
-		resetSettingsForTest();
-		tempDir = TempDir.createSync("@pi-shutdown-");
-		await Settings.init({ inMemory: true, cwd: tempDir.path() });
-		authStorage = await AuthStorage.create(path.join(tempDir.path(), "testauth.db"));
-		const modelRegistry = new ModelRegistry(authStorage);
-		const model = modelRegistry.find("anthropic", "claude-sonnet-4-5");
-		if (!model) throw new Error("Expected claude-sonnet-4-5 test model");
-
-		session = new AgentSession({
-			agent: new Agent({ initialState: { model, systemPrompt: ["Test"], tools: [], messages: [] } }),
-			sessionManager: SessionManager.create(tempDir.path(), tempDir.path()),
-			settings: Settings.isolated(),
-			modelRegistry,
-		});
-		mode = new InteractiveMode(session, "test");
-	});
-
-	afterEach(async () => {
-		mode?.stop();
-		vi.restoreAllMocks();
-		await session?.dispose();
-		authStorage?.close();
-		tempDir?.removeSync();
-		resetSettingsForTest();
-	});
-
-	it("stops from the last committed TUI frame without forcing a teardown repaint", async () => {
-		const requestRenderSpy = vi.spyOn(mode.ui, "requestRender").mockImplementation(() => {});
-		const stopSpy = vi.spyOn(mode.ui, "stop").mockImplementation(() => {});
-		const drainSpy = vi.spyOn(mode.ui.terminal, "drainInput").mockResolvedValue(undefined);
-		const disposeSpy = vi.spyOn(session, "dispose").mockResolvedValue(undefined);
-		const quitSpy = vi.spyOn(postmortem, "quit").mockResolvedValue(undefined);
-		vi.spyOn(session.sessionManager, "getSessionId").mockReturnValue("");
-		mode.isInitialized = true;
-
-		await mode.shutdown();
-
-		expect(disposeSpy).toHaveBeenCalled();
-		expect(requestRenderSpy.mock.calls.some(call => call[0] === true)).toBe(false);
-		expect(drainSpy).toHaveBeenCalledWith(1000);
-		expect(stopSpy).toHaveBeenCalled();
-		expect(quitSpy).toHaveBeenCalledWith(0);
-	});
-});
diff --git a/packages/coding-agent/test/interactive-mode-status.test.ts b/packages/coding-agent/test/interactive-mode-status.test.ts
index 92bfc8987..f32f13587 100644
--- a/packages/coding-agent/test/interactive-mode-status.test.ts
+++ b/packages/coding-agent/test/interactive-mode-status.test.ts
@@ -4,7 +4,7 @@ import { initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
 import type { InteractiveModeContext } from "@oh-my-pi/pi-coding-agent/modes/types";
 import { UiHelpers } from "@oh-my-pi/pi-coding-agent/modes/utils/ui-helpers";
 import { buildSessionContext, type SessionContext } from "@oh-my-pi/pi-coding-agent/session/session-manager";
-import { Container } from "@oh-my-pi/pi-tui";
+import { type Component, Container } from "@oh-my-pi/pi-tui";
 
 function renderLastLine(container: Container, width = 120): string {
 	const last = container.children[container.children.length - 1];
@@ -25,7 +25,11 @@ function createInitialRenderHarness(): { ctx: InteractiveModeContext; helpers: U
 		pendingPythonComponents: [],
 		pendingTools: new Map(),
 		ui: { requestRender: vi.fn() },
-		isBackgrounded: false,
+		present: (content: Component | readonly Component[]) => {
+			const items = Array.isArray(content) ? content : [content];
+			for (const item of items) ctx.chatContainer.addChild(item);
+			ctx.ui.requestRender();
+		},
 		sessionManager: {
 			buildSessionContext: () => buildSessionContext([]),
 			getEntries: () => [],
@@ -59,7 +63,11 @@ describe("InteractiveMode.showStatus", () => {
 		const ctx = {
 			chatContainer: new Container(),
 			ui: { requestRender: vi.fn() },
-			isBackgrounded: false,
+			present: (content: Component | readonly Component[]) => {
+				const items = Array.isArray(content) ? content : [content];
+				for (const item of items) ctx.chatContainer.addChild(item);
+				ctx.ui.requestRender();
+			},
 			lastStatusSpacer: undefined,
 			lastStatusText: undefined,
 		} as unknown as InteractiveModeContext;
@@ -80,7 +88,11 @@ describe("InteractiveMode.showStatus", () => {
 		const ctx = {
 			chatContainer: new Container(),
 			ui: { requestRender: vi.fn() },
-			isBackgrounded: false,
+			present: (content: Component | readonly Component[]) => {
+				const items = Array.isArray(content) ? content : [content];
+				for (const item of items) ctx.chatContainer.addChild(item);
+				ctx.ui.requestRender();
+			},
 			lastStatusSpacer: undefined,
 			lastStatusText: undefined,
 		} as unknown as InteractiveModeContext;
diff --git a/packages/coding-agent/test/issue-846-repro.test.ts b/packages/coding-agent/test/issue-846-repro.test.ts
index 13b3e1362..a1762a8d5 100644
--- a/packages/coding-agent/test/issue-846-repro.test.ts
+++ b/packages/coding-agent/test/issue-846-repro.test.ts
@@ -37,6 +37,7 @@ interface ModelRegistryLike {
 	find: (...args: unknown[]) => Model;
 	getAll: () => Model[];
 	getApiKey: (...args: unknown[]) => Promise<string>;
+	resolver: (...args: unknown[]) => () => Promise<string>;
 }
 
 const createdDirs = new Set<string>();
@@ -62,6 +63,7 @@ function createModelRegistry(model: Model): ModelRegistryLike {
 		find: vi.fn(() => model),
 		getAll: vi.fn(() => [model]),
 		getApiKey: vi.fn(async () => "test-api-key"),
+		resolver: vi.fn(() => async () => "test-api-key"),
 	};
 }
 
diff --git a/packages/coding-agent/test/issue-899-repro.test.ts b/packages/coding-agent/test/issue-899-repro.test.ts
index a63d2f887..03b3997ab 100644
--- a/packages/coding-agent/test/issue-899-repro.test.ts
+++ b/packages/coding-agent/test/issue-899-repro.test.ts
@@ -45,7 +45,6 @@ describe("issue #899 — sync git metadata reads must survive EINTR", () => {
 		}) as typeof fs.readFileSync);
 
 		// On main this throws EINTR; after fix it must return null (metadata unavailable).
-		expect(() => head.resolveSync(tempDir)).not.toThrow();
 		expect(head.resolveSync(tempDir)).toBeNull();
 		expect(spy).toHaveBeenCalled();
 	});
diff --git a/packages/coding-agent/test/issue-956-repro.test.ts b/packages/coding-agent/test/issue-956-repro.test.ts
index f7a1b38c7..1d1f4872b 100644
--- a/packages/coding-agent/test/issue-956-repro.test.ts
+++ b/packages/coding-agent/test/issue-956-repro.test.ts
@@ -80,6 +80,10 @@ describe("issue #956: interactive /mcp test", () => {
 		const disconnectServer = vi.spyOn(mcpClient, "disconnectServer").mockResolvedValue();
 		const controller = new MCPCommandController({
 			chatContainer: { addChild },
+			present: (content: unknown) => {
+				for (const item of Array.isArray(content) ? content : [content]) addChild(item);
+				requestRender();
+			},
 			ui: { requestRender },
 			editor: {},
 			showError,
diff --git a/packages/coding-agent/test/keybindings-migration.test.ts b/packages/coding-agent/test/keybindings-migration.test.ts
index 2dc8e0614..89564b84c 100644
--- a/packages/coding-agent/test/keybindings-migration.test.ts
+++ b/packages/coding-agent/test/keybindings-migration.test.ts
@@ -106,4 +106,71 @@ describe("KeybindingsManager.create", () => {
 			await fs.rm(agentDir, { recursive: true, force: true });
 		}
 	});
+
+	it("defaults model selection to Alt+M and display reset to Ctrl+L", () => {
+		const manager = KeybindingsManager.inMemory();
+
+		expect(manager.getKeys("app.model.select")).toEqual(["alt+m"]);
+		expect(manager.getKeys("app.display.reset")).toEqual(["ctrl+l"]);
+	});
+
+	it("keeps the Ctrl+L display reset default when an old model remap still claims Ctrl+L", () => {
+		const manager = KeybindingsManager.inMemory({
+			"app.model.select": "ctrl+l",
+		});
+
+		expect(manager.getKeys("app.model.select")).toEqual(["ctrl+l"]);
+		expect(manager.getKeys("app.display.reset")).toEqual(["ctrl+l"]);
+		expect(manager.getEffectiveConfig()["app.display.reset"]).toBe("ctrl+l");
+	});
+
+	it("keeps Ctrl+L when the user explicitly assigns it to display reset", () => {
+		const manager = KeybindingsManager.inMemory({
+			"app.display.reset": "ctrl+l",
+		});
+
+		expect(manager.getKeys("app.display.reset")).toEqual(["ctrl+l"]);
+	});
+
+	it("defaults the follow-up shortcut to both Ctrl+Q and Ctrl+Enter (#1903)", async () => {
+		const agentDir = await fs.mkdtemp(path.join(os.tmpdir(), "pi-keybindings-"));
+
+		try {
+			const manager = KeybindingsManager.create(agentDir);
+
+			// Both chords must be registered so Windows Terminal users (which swallow
+			// Ctrl+Enter at the terminal layer) get a working follow-up binding out
+			// of the box, without breaking users on Kitty/iTerm2/WezTerm/Ghostty.
+			expect(manager.getKeys("app.message.followUp")).toEqual(["ctrl+q", "ctrl+enter"]);
+		} finally {
+			await fs.rm(agentDir, { recursive: true, force: true });
+		}
+	});
+
+	it("removes the Ctrl+Q follow-up default when a user remap already claims it (#1903)", () => {
+		const manager = KeybindingsManager.inMemory({
+			"app.plan.toggle": "ctrl+q",
+		});
+
+		expect(manager.getKeys("app.plan.toggle")).toEqual(["ctrl+q"]);
+		expect(manager.getKeys("app.message.followUp")).toEqual(["ctrl+enter"]);
+		expect(manager.getDisplayString("app.message.followUp")).toBe("Ctrl+Enter");
+		expect(manager.getEffectiveConfig()["app.message.followUp"]).toBe("ctrl+enter");
+	});
+
+	it("keeps the Ctrl+Q follow-up default when only an unknown config key claims it (#1903)", () => {
+		const manager = KeybindingsManager.inMemory({
+			"unknown.action": "ctrl+q",
+		});
+
+		expect(manager.getKeys("app.message.followUp")).toEqual(["ctrl+q", "ctrl+enter"]);
+	});
+
+	it("keeps Ctrl+Q when the user explicitly assigns it to follow-up (#1903)", () => {
+		const manager = KeybindingsManager.inMemory({
+			"app.message.followUp": "ctrl+q",
+		});
+
+		expect(manager.getKeys("app.message.followUp")).toEqual(["ctrl+q"]);
+	});
 });
diff --git a/packages/coding-agent/test/keybindings-selector-navigation.test.ts b/packages/coding-agent/test/keybindings-selector-navigation.test.ts
index e79557af5..78dd76e76 100644
--- a/packages/coding-agent/test/keybindings-selector-navigation.test.ts
+++ b/packages/coding-agent/test/keybindings-selector-navigation.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, beforeAll, describe, expect, it } from "bun:test";
+import { afterEach, beforeAll, describe, expect, it, vi } from "bun:test";
 import * as fs from "node:fs/promises";
 import * as os from "node:os";
 import * as path from "node:path";
@@ -86,8 +86,15 @@ async function createHistoryStorage(prompts: string[]): Promise<HistoryStorage>
 	tempDirs.push(dir);
 	HistoryStorage.resetInstance();
 	const storage = HistoryStorage.open(path.join(dir, "history.db"));
-	for (const prompt of prompts) {
-		await storage.add(prompt);
+	// add() batches writes behind a 100ms AsyncDrain timer. Drive that timer with
+	// fake timers so the flush is instant instead of waiting real wall-clock time.
+	vi.useFakeTimers();
+	try {
+		const writes = prompts.map(prompt => storage.add(prompt));
+		vi.advanceTimersByTime(100);
+		await Promise.all(writes);
+	} finally {
+		vi.useRealTimers();
 	}
 	return storage;
 }
diff --git a/packages/coding-agent/test/loop-limit.test.ts b/packages/coding-agent/test/loop-limit.test.ts
index ff2033c19..9e660785c 100644
--- a/packages/coding-agent/test/loop-limit.test.ts
+++ b/packages/coding-agent/test/loop-limit.test.ts
@@ -16,7 +16,6 @@ describe("/loop slash command", () => {
 				handleLoopCommand,
 				editor: { setText: vi.fn() },
 			},
-			handleBackgroundCommand: vi.fn(),
 		} as unknown as BuiltinSlashCommandRuntime;
 		const result = await executeBuiltinSlashCommand("/loop 10min", runtime);
 
diff --git a/packages/coding-agent/test/marketplace/discovery.test.ts b/packages/coding-agent/test/marketplace/discovery.test.ts
index bff84d262..8356e014f 100644
--- a/packages/coding-agent/test/marketplace/discovery.test.ts
+++ b/packages/coding-agent/test/marketplace/discovery.test.ts
@@ -91,13 +91,6 @@ describe("OMP registry path contract", () => {
 		const expected = path.join(tmpHome, ".omp", "plugins", "installed_plugins.json");
 		expect(ompRegistryPath).toBe(expected);
 	});
-
-	it("OMP config dir name is .omp", () => {
-		// Validate our hardcoded constant matches getConfigDirName().
-		// If getConfigDirName() ever changes, this assertion will fail and
-		// we'll know the path constant here must be updated too.
-		expect(OMP_CONFIG_DIR).toBe(".omp");
-	});
 });
 
 // ── Format compatibility ───────────────────────────────────────────────────────
@@ -230,21 +223,4 @@ describe("OMP precedence contract (registry structure)", () => {
 		expect(id.slice(0, atIndex)).toBe("shared-plugin");
 		expect(id.slice(atIndex + 1)).toBe("common-mkt");
 	});
-
-	it("installPath deduplication: same path → one entry", () => {
-		// Mirrors the deduplication check: roots.some(r => r.id === pluginId && r.path === entry.installPath)
-		const id = buildPluginId("dup-plugin", "mkt");
-		const sharedPath = "/tmp/shared-install-path";
-
-		// Simulate what listClaudePluginRoots would do:
-		const roots: Array<{ id: string; path: string }> = [{ id, path: sharedPath }];
-
-		// Second entry with same installPath should be deduplicated
-		const isDuplicate = roots.some(r => r.id === id && r.path === sharedPath);
-		expect(isDuplicate).toBe(true);
-
-		// Entry with different installPath should NOT be deduplicated
-		const isDifferent = roots.some(r => r.id === id && r.path === "/tmp/other-path");
-		expect(isDifferent).toBe(false);
-	});
 });
diff --git a/packages/coding-agent/test/mcp-reconnect-storm.test.ts b/packages/coding-agent/test/mcp-reconnect-storm.test.ts
index 6f75e8955..87a5ef49a 100644
--- a/packages/coding-agent/test/mcp-reconnect-storm.test.ts
+++ b/packages/coding-agent/test/mcp-reconnect-storm.test.ts
@@ -52,10 +52,19 @@ describe("MCP reconnect storm (issue #1592)", () => {
 
 		try {
 			await manager.connectServers({ crashy: config }, {});
-			// Give the reconnect loop generous time to fire. With the bug this
-			// produced thousands of processes within a second; with the fix the
-			// circuit breaker caps the per-server spawn budget.
-			await Bun.sleep(3000);
+			// Wait for the circuit breaker to trip rather than blind-sleeping a
+			// fixed budget. During the storm `getConnectionStatus` is always
+			// "connected" or "connecting" (`#pendingReconnections` is set
+			// synchronously before any await in `#doReconnect`); it only reports
+			// "disconnected" once `#tripReconnectBreaker` opens, tears down the
+			// stale connection, and detaches `onClose` so no further spawns fire.
+			// That makes the terminal state a race-free signal: poll for it and
+			// return the instant the storm is capped instead of waiting out a
+			// fixed 3s. Generous deadline stays well under the 15s test timeout.
+			const deadline = Date.now() + 10_000;
+			while (manager.getConnectionStatus("crashy") !== "disconnected" && Date.now() < deadline) {
+				await Bun.sleep(5);
+			}
 
 			const spawns = countSpawns();
 			// `RECONNECT_BURST_LIMIT` (5) is the per-server reconnect cap inside
diff --git a/packages/coding-agent/test/memories-runtime.test.ts b/packages/coding-agent/test/memories-runtime.test.ts
index 394daebbb..ba2429d73 100644
--- a/packages/coding-agent/test/memories-runtime.test.ts
+++ b/packages/coding-agent/test/memories-runtime.test.ts
@@ -46,6 +46,7 @@ function createModelRegistry(model: Model): any {
 		find: vi.fn(() => model),
 		getAll: vi.fn(() => [model]),
 		getApiKey: vi.fn(async () => "test-api-key"),
+		resolver: vi.fn(() => async () => "test-api-key"),
 	};
 }
 
diff --git a/packages/coding-agent/test/memory-backend-resolve.test.ts b/packages/coding-agent/test/memory-backend-resolve.test.ts
index 075f05e8f..46845e103 100644
--- a/packages/coding-agent/test/memory-backend-resolve.test.ts
+++ b/packages/coding-agent/test/memory-backend-resolve.test.ts
@@ -11,10 +11,10 @@ describe("resolveMemoryBackend", () => {
 		resetSettingsForTest();
 	});
 
-	it("returns the hindsight backend when memory.backend is hindsight, regardless of legacy memories.enabled", () => {
+	it("returns the hindsight backend when memory.backend is hindsight, regardless of legacy memories.enabled", async () => {
 		const a = Settings.isolated({ "memory.backend": "hindsight", "memories.enabled": false });
 		const b = Settings.isolated({ "memory.backend": "hindsight", "memories.enabled": true });
-		expect(resolveMemoryBackend(a).id).toBe("hindsight");
-		expect(resolveMemoryBackend(b).id).toBe("hindsight");
+		expect((await resolveMemoryBackend(a)).id).toBe("hindsight");
+		expect((await resolveMemoryBackend(b)).id).toBe("hindsight");
 	});
 });
diff --git a/packages/coding-agent/test/model-id-affixes.test.ts b/packages/coding-agent/test/model-id-affixes.test.ts
new file mode 100644
index 000000000..b94a9bc48
--- /dev/null
+++ b/packages/coding-agent/test/model-id-affixes.test.ts
@@ -0,0 +1,66 @@
+import { describe, expect, test } from "bun:test";
+import {
+	getBracketStrippedModelIdCandidates,
+	getLongestModelLikeIdSegment,
+	getModelLikeIdSegments,
+	stripBracketedModelIdAffixes,
+} from "../src/config/model-id-affixes";
+
+describe("getModelLikeIdSegments", () => {
+	test("keeps only family-prefixed segments that carry a digit, deduped", () => {
+		expect(getModelLikeIdSegments("openrouter/anthropic/claude-3.5-sonnet")).toEqual(["claude-3.5-sonnet"]);
+		// `random-text` lacks a family prefix; `claude` (no digit) is dropped.
+		expect(getModelLikeIdSegments("random-text claude gemini-2")).toEqual(["gemini-2"]);
+	});
+
+	test("orders longest first with lexicographic tie-break", () => {
+		expect(getModelLikeIdSegments("claude-3 claude-3-5-haiku claude-2")).toEqual([
+			"claude-3-5-haiku",
+			"claude-2",
+			"claude-3",
+		]);
+	});
+
+	test("normalizes whitespace and case before matching", () => {
+		expect(getModelLikeIdSegments("  GLM-4.5-Air   GEMINI-2  ")).toEqual(["glm-4.5-air", "gemini-2"]);
+	});
+
+	test("returns empty for ids with no model-like segment", () => {
+		expect(getModelLikeIdSegments("")).toEqual([]);
+		expect(getModelLikeIdSegments("just some words")).toEqual([]);
+	});
+});
+
+describe("getLongestModelLikeIdSegment", () => {
+	test("matches getModelLikeIdSegments[0]", () => {
+		const id = "[Kiro] claude-3 claude-3-5-sonnet";
+		expect(getLongestModelLikeIdSegment(id)).toBe(getModelLikeIdSegments(id)[0]);
+		expect(getLongestModelLikeIdSegment(id)).toBe("claude-3-5-sonnet");
+	});
+
+	test("is undefined when nothing matches", () => {
+		expect(getLongestModelLikeIdSegment("vendor/unknown-tag")).toBeUndefined();
+	});
+});
+
+describe("getBracketStrippedModelIdCandidates", () => {
+	test("no brackets yields no candidates", () => {
+		expect(getBracketStrippedModelIdCandidates("claude-opus-4-8")).toEqual([]);
+	});
+
+	test("strips leading reseller tag", () => {
+		expect(getBracketStrippedModelIdCandidates("[Kiro] claude-opus-4-8")).toEqual(["claude-opus-4-8"]);
+	});
+
+	test("strips both ends first, then each side, in preference order", () => {
+		expect(getBracketStrippedModelIdCandidates("[gcli转] gemini-3.1-pro-preview [假流]")).toEqual([
+			"gemini-3.1-pro-preview",
+			"gemini-3.1-pro-preview [假流]",
+			"[gcli转] gemini-3.1-pro-preview",
+		]);
+	});
+
+	test("supports full-width brackets", () => {
+		expect(stripBracketedModelIdAffixes("【供应商】 deepseek-v3 【限时】")).toBe("deepseek-v3");
+	});
+});
diff --git a/packages/coding-agent/test/model-registry-runtime-provider.test.ts b/packages/coding-agent/test/model-registry-runtime-provider.test.ts
index 9538a0e34..30e40efbb 100644
--- a/packages/coding-agent/test/model-registry-runtime-provider.test.ts
+++ b/packages/coding-agent/test/model-registry-runtime-provider.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, beforeEach, describe, expect, test } from "bun:test";
+import { afterEach, beforeEach, describe, expect, type Mock, spyOn, test } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
@@ -13,6 +13,9 @@ describe("ModelRegistry runtime provider registration", () => {
 	let tempDir: string;
 	let modelsJsonPath: string;
 	let authStorage: AuthStorage;
+	// Neutralizes real network egress during "online" refresh tests so the merge
+	// path runs without wall-clock-bound DNS/socket latency. Restored in afterEach.
+	let fetchSpy: Mock<typeof fetch> | undefined;
 
 	const sourceIds = ["ext://atomic", "ext://runtime", "ext://oauth"];
 
@@ -24,6 +27,8 @@ describe("ModelRegistry runtime provider registration", () => {
 	});
 
 	afterEach(() => {
+		fetchSpy?.mockRestore();
+		fetchSpy = undefined;
 		clearCustomApis();
 		for (const sourceId of sourceIds) {
 			unregisterOAuthProviders(sourceId);
@@ -192,6 +197,11 @@ describe("ModelRegistry runtime provider registration", () => {
 	});
 
 	test("extension-registered models survive refresh('online') cycle", async () => {
+		// The contract is overlay survival through the full online refresh path
+		// (static reload + discovery + merge), not discovery success. Stub fetch so
+		// the online branch runs identically to production-with-no-reachable-providers
+		// without paying real network latency (~400ms of DNS/socket time otherwise).
+		fetchSpy = spyOn(globalThis, "fetch").mockRejectedValue(new Error("network disabled in test"));
 		const registry = new ModelRegistry(authStorage, modelsJsonPath);
 		const config: ProviderConfigInput = {
 			baseUrl: "https://runtime.example.com/v1",
diff --git a/packages/coding-agent/test/model-registry.test.ts b/packages/coding-agent/test/model-registry.test.ts
index 0cdb092ca..218d98a22 100644
--- a/packages/coding-agent/test/model-registry.test.ts
+++ b/packages/coding-agent/test/model-registry.test.ts
@@ -29,7 +29,10 @@ describe("ModelRegistry", () => {
 		fs.mkdirSync(tempDir, { recursive: true });
 		modelsJsonPath = path.join(tempDir, "models.json");
 		cacheDbPath = path.join(tempDir, "models.db");
-		authStorage = await AuthStorage.create(path.join(tempDir, "testauth.db"));
+		// In-memory auth DB: tests need a fresh, isolated credential store per case but
+		// never reopen it from disk, so :memory: avoids the WAL/chmod disk-open cost
+		// (~3ms/test) while preserving per-test isolation.
+		authStorage = await AuthStorage.create(":memory:");
 	});
 
 	afterEach(() => {
diff --git a/packages/coding-agent/test/model-selector-role-badge-thinking.test.ts b/packages/coding-agent/test/model-selector-role-badge-thinking.test.ts
index 384b13a89..44e65b20c 100644
--- a/packages/coding-agent/test/model-selector-role-badge-thinking.test.ts
+++ b/packages/coding-agent/test/model-selector-role-badge-thinking.test.ts
@@ -47,6 +47,47 @@ function createOllamaCloudModel(id: string): Model {
 		maxTokens: 8192,
 	};
 }
+function createContextTestModel(id: string, contextWindow: number): Model {
+	return {
+		id,
+		name: id,
+		api: "ollama-chat",
+		baseUrl: "https://example.com",
+		reasoning: false,
+		provider: "test",
+		input: ["text"],
+		cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
+		contextWindow,
+		maxTokens: 1024,
+	};
+}
+
+function createScopedSelector(
+	models: Model[],
+	settings: Settings,
+	onSelect: (model: Model) => void,
+	options?: { temporaryOnly?: boolean; currentContextTokens?: number },
+): ModelSelectorComponent {
+	const modelRegistry = {
+		getAll: () => models,
+		getDiscoverableProviders: () => [],
+		getCanonicalModels: () => [],
+		resolveCanonicalModel: () => undefined,
+	} as unknown as ModelRegistry;
+	const ui = {
+		requestRender: vi.fn(),
+	} as unknown as TUI;
+	return new ModelSelectorComponent(
+		ui,
+		undefined,
+		settings,
+		modelRegistry,
+		models.map(model => ({ model })),
+		model => onSelect(model),
+		() => {},
+		options,
+	);
+}
 let testTheme = await getThemeByName("dark");
 
 function installTestTheme(): void {
@@ -96,6 +137,48 @@ describe("ModelSelector role badge thinking display", () => {
 		expect(menuRendered).toContain("Set as SMOL (Quick)");
 	});
 
+	test("dims and disables models below the current context size", async () => {
+		installTestTheme();
+		const settings = Settings.isolated({});
+		const small = createContextTestModel("a-small", 4096);
+		const large = createContextTestModel("b-large", 128_000);
+		const selected: string[] = [];
+		const selector = createScopedSelector([small, large], settings, model => selected.push(model.id), {
+			temporaryOnly: true,
+			currentContextTokens: 6000,
+		});
+		await Bun.sleep(0);
+		installTestTheme();
+
+		const rendered = normalizeRenderedText(selector.render(220).join("\n"));
+		expect(rendered).toContain("a-small");
+		expect(rendered).toContain("context>4.1k");
+
+		selector.handleInput("\n");
+		expect(selected).toEqual(["b-large"]);
+	});
+
+	test("does not open the model menu when every candidate is disabled", async () => {
+		installTestTheme();
+		const settings = Settings.isolated({});
+		const small = createContextTestModel("only-small", 4096);
+		const onSelect = vi.fn();
+		const selector = createScopedSelector([small], settings, onSelect, {
+			currentContextTokens: 6000,
+		});
+		await Bun.sleep(0);
+		installTestTheme();
+
+		const rendered = normalizeRenderedText(selector.render(220).join("\n"));
+		expect(rendered).toContain("only-small");
+		expect(rendered).toContain("current context 6k > 4.1k limit");
+
+		selector.handleInput("\n");
+		const afterEnter = normalizeRenderedText(selector.render(220).join("\n"));
+		expect(afterEnter).not.toContain("Action for");
+		expect(onSelect).not.toHaveBeenCalled();
+	});
+
 	test("refreshes Ollama Cloud using provider id instead of tab label", async () => {
 		installTestTheme();
 		const settings = Settings.isolated({});
diff --git a/packages/coding-agent/test/modes/components/chat-block.test.ts b/packages/coding-agent/test/modes/components/chat-block.test.ts
new file mode 100644
index 000000000..703504b9d
--- /dev/null
+++ b/packages/coding-agent/test/modes/components/chat-block.test.ts
@@ -0,0 +1,143 @@
+import { beforeEach, describe, expect, it } from "bun:test";
+import { ChatBlock, type ChatBlockHost } from "@oh-my-pi/pi-coding-agent/modes/components/chat-block";
+import type { Component } from "@oh-my-pi/pi-tui";
+
+/** Concrete subclass exposing the protected lifecycle seams for assertions. */
+class TestBlock extends ChatBlock {
+	mountCount = 0;
+	cleanupCount = 0;
+
+	protected override onMount(): void {
+		this.mountCount++;
+		this.onCleanup(() => {
+			this.cleanupCount++;
+		});
+	}
+
+	/** Public proxy for the protected requestRender. */
+	ping(): void {
+		this.requestRender();
+	}
+
+	/** Public proxy for the protected onCleanup. */
+	register(cleanup: () => void): void {
+		this.onCleanup(cleanup);
+	}
+}
+
+describe("ChatBlock lifecycle", () => {
+	let renders: number;
+	let host: ChatBlockHost;
+
+	beforeEach(() => {
+		renders = 0;
+		host = {
+			requestRender: () => {
+				renders++;
+			},
+		};
+	});
+
+	it("runs onMount exactly once; a second mount is a no-op", () => {
+		const block = new TestBlock();
+		expect(block.mountCount).toBe(0);
+		block.mount(host);
+		block.mount(host);
+		expect(block.mountCount).toBe(1);
+	});
+
+	it("is finalized until mounted, live while active, finalized after finish", () => {
+		const block = new TestBlock();
+		expect(block.isTranscriptBlockFinalized()).toBe(true);
+		block.mount(host);
+		expect(block.isTranscriptBlockFinalized()).toBe(false);
+		block.finish();
+		expect(block.isTranscriptBlockFinalized()).toBe(true);
+	});
+
+	it("finish runs cleanups once and requests one render", () => {
+		const block = new TestBlock();
+		block.mount(host);
+		const before = renders;
+		block.finish();
+		expect(block.cleanupCount).toBe(1);
+		expect(renders).toBe(before + 1);
+		block.finish();
+		expect(block.cleanupCount).toBe(1);
+	});
+
+	it("dispose runs cleanups once, is idempotent, and finalizes", () => {
+		const block = new TestBlock();
+		block.mount(host);
+		block.dispose();
+		expect(block.cleanupCount).toBe(1);
+		expect(block.isTranscriptBlockFinalized()).toBe(true);
+		block.dispose();
+		expect(block.cleanupCount).toBe(1);
+	});
+
+	it("finish then dispose does not double-run cleanups", () => {
+		const block = new TestBlock();
+		block.mount(host);
+		block.finish();
+		block.dispose();
+		expect(block.cleanupCount).toBe(1);
+	});
+
+	it("requestRender routes to the host only between mount and dispose", () => {
+		const block = new TestBlock();
+		block.ping();
+		expect(renders).toBe(0);
+		block.mount(host);
+		block.ping();
+		expect(renders).toBe(1);
+		block.dispose();
+		const after = renders;
+		block.ping();
+		expect(renders).toBe(after);
+	});
+
+	it("onCleanup registered after dispose runs immediately so callers never leak", () => {
+		const block = new TestBlock();
+		block.mount(host);
+		block.dispose();
+		let ran = false;
+		block.register(() => {
+			ran = true;
+		});
+		expect(ran).toBe(true);
+	});
+
+	it("dispose propagates to child components", () => {
+		const block = new TestBlock();
+		let childDisposed = 0;
+		const child: Component = {
+			render: () => [],
+			invalidate: () => {},
+			dispose: () => {
+				childDisposed++;
+			},
+		};
+		block.addChild(child);
+		block.mount(host);
+		block.dispose();
+		expect(childDisposed).toBe(1);
+	});
+
+	it("tears down a timer effect started in onMount when finished", async () => {
+		class TimerBlock extends ChatBlock {
+			protected override onMount(): void {
+				const id = setInterval(() => this.requestRender(), 5);
+				this.onCleanup(() => clearInterval(id));
+			}
+		}
+		const block = new TimerBlock();
+		block.mount(host);
+		await Bun.sleep(25);
+		expect(renders).toBeGreaterThan(0); // timer fired while active
+		block.finish();
+		const settled = renders; // includes finish()'s own render
+		await Bun.sleep(25);
+		expect(renders).toBe(settled); // interval torn down — no further ticks
+	});
+});
diff --git a/packages/coding-agent/test/modes/components/copy-selector.test.ts b/packages/coding-agent/test/modes/components/copy-selector.test.ts
new file mode 100644
index 000000000..7b09f17be
--- /dev/null
+++ b/packages/coding-agent/test/modes/components/copy-selector.test.ts
@@ -0,0 +1,135 @@
+import { afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
+import { stripVTControlCharacters } from "node:util";
+import { KeybindingsManager } from "@oh-my-pi/pi-coding-agent/config/keybindings";
+import { CopySelectorComponent } from "@oh-my-pi/pi-coding-agent/modes/components/copy-selector";
+import { getThemeByName, setThemeInstance } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
+import type { CopyTarget } from "@oh-my-pi/pi-coding-agent/modes/utils/copy-targets";
+import { setKeybindings } from "@oh-my-pi/pi-tui";
+
+const UP = "\x1b[A";
+const DOWN = "\x1b[B";
+const ENTER = "\n";
+const CANCEL = "\x07"; // ctrl+g, remapped to tui.select.cancel below
+
+let darkTheme = await getThemeByName("dark");
+
+// Flatten order (always expanded): msg:1, Block 1, Block 2, msg:2.
+function makeRoots(): CopyTarget[] {
+	return [
+		{
+			id: "msg:1",
+			label: "Newest message",
+			hint: "5 lines · 2 code",
+			preview: "newest-preview-text",
+			content: "FULL_MESSAGE",
+			copyMessage: "Copied last message to clipboard",
+			children: [
+				{
+					id: "msg:1:code:0",
+					label: "Block 1",
+					hint: "ts",
+					language: "ts",
+					preview: "alpha()",
+					content: "BLOCK0",
+					copyMessage: "Copied block 1",
+				},
+				{
+					id: "msg:1:code:1",
+					label: "Block 2",
+					hint: "py",
+					language: "python",
+					preview: "beta()",
+					content: "BLOCK1",
+					copyMessage: "Copied block 2",
+				},
+			],
+		},
+		{
+			id: "msg:2",
+			label: "Older message",
+			hint: "3 lines",
+			preview: "older-text",
+			content: "OLDER",
+			copyMessage: "Copied message",
+		},
+	];
+}
+
+function render(component: CopySelectorComponent): string {
+	return stripVTControlCharacters(component.render(80).join("\n"));
+}
+
+describe("CopySelectorComponent", () => {
+	beforeAll(async () => {
+		darkTheme = await getThemeByName("dark");
+		if (!darkTheme) throw new Error("Failed to load dark theme");
+	});
+
+	beforeEach(() => {
+		setThemeInstance(darkTheme!);
+		setKeybindings(KeybindingsManager.inMemory({ "tui.select.cancel": "ctrl+g" }));
+	});
+
+	afterEach(() => {
+		setKeybindings(KeybindingsManager.inMemory());
+		vi.restoreAllMocks();
+	});
+
+	it("renders an outlined tree with code blocks nested under their message", () => {
+		const out = render(new CopySelectorComponent(makeRoots(), { onPick: vi.fn(), onCancel: vi.fn() }));
+		expect(out).toContain("┌");
+		expect(out).toContain("│");
+		expect(out).toContain("Copy to clipboard");
+		// Messages and their nested blocks are all visible (always expanded),
+		// connected with /tree-style branch glyphs.
+		expect(out).toContain("Newest message");
+		expect(out).toContain("Block 1");
+		expect(out).toContain("Block 2");
+		expect(out).toContain("Older message");
+		expect(out).toMatch(/[├└]/);
+	});
+
+	it("copies the message node itself on Enter", () => {
+		const onPick = vi.fn();
+		const component = new CopySelectorComponent(makeRoots(), { onPick, onCancel: vi.fn() });
+
+		component.handleInput(ENTER); // cursor starts on the message node
+
+		expect(onPick).toHaveBeenCalledTimes(1);
+		expect(onPick.mock.calls[0]![0].content).toBe("FULL_MESSAGE");
+	});
+
+	it("navigates into a nested code block and copies it", () => {
+		const onPick = vi.fn();
+		const component = new CopySelectorComponent(makeRoots(), { onPick, onCancel: vi.fn() });
+
+		component.handleInput(DOWN); // onto "Block 1"
+		component.handleInput(ENTER);
+
+		expect(onPick).toHaveBeenCalledTimes(1);
+		expect(onPick.mock.calls[0]![0].content).toBe("BLOCK0");
+	});
+
+	it("traverses past nested blocks to the older message, with the preview tracking the cursor", () => {
+		const component = new CopySelectorComponent(makeRoots(), { onPick: vi.fn(), onCancel: vi.fn() });
+
+		component.handleInput(DOWN); // Block 1
+		expect(render(component)).toContain("alpha()");
+
+		component.handleInput(DOWN); // Block 2
+		component.handleInput(DOWN); // Older message
+		expect(render(component)).toContain("older-text");
+
+		component.handleInput(UP); // back onto Block 2
+		expect(render(component)).toContain("beta()");
+	});
+
+	it("quits on the cancel key", () => {
+		const onCancel = vi.fn();
+		const component = new CopySelectorComponent(makeRoots(), { onPick: vi.fn(), onCancel });
+
+		component.handleInput(CANCEL);
+
+		expect(onCancel).toHaveBeenCalledTimes(1);
+	});
+});
diff --git a/packages/coding-agent/test/modes/components/plan-review-overlay.test.ts b/packages/coding-agent/test/modes/components/plan-review-overlay.test.ts
new file mode 100644
index 000000000..7bfea1265
--- /dev/null
+++ b/packages/coding-agent/test/modes/components/plan-review-overlay.test.ts
@@ -0,0 +1,447 @@
+import { afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
+import { stripVTControlCharacters } from "node:util";
+import { KeybindingsManager } from "@oh-my-pi/pi-coding-agent/config/keybindings";
+import type { HookSelectorSlider } from "@oh-my-pi/pi-coding-agent/modes/components/hook-selector";
+import { PlanReviewOverlay } from "@oh-my-pi/pi-coding-agent/modes/components/plan-review-overlay";
+import { getThemeByName, setThemeInstance } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
+import { setKeybindings } from "@oh-my-pi/pi-tui";
+
+const UP = "\x1b[A";
+const DOWN = "\x1b[B";
+const LEFT = "\x1b[D";
+const RIGHT = "\x1b[C";
+const ENTER = "\r";
+const TAB = "\t";
+const SHIFT_DOWN = "\x1b[1;2B";
+const CANCEL = "\x07"; // ctrl+g, remapped to tui.select.cancel below
+
+let darkTheme = await getThemeByName("dark");
+
+function render(component: PlanReviewOverlay): string {
+	return stripVTControlCharacters(component.render(80).join("\n"));
+}
+
+const APPROVAL_OPTIONS = [
+	"Approve and execute",
+	"Approve and compact context",
+	"Approve and keep context",
+	"Refine plan",
+];
+
+describe("PlanReviewOverlay", () => {
+	beforeAll(async () => {
+		darkTheme = await getThemeByName("dark");
+		if (!darkTheme) throw new Error("Failed to load dark theme");
+	});
+
+	beforeEach(() => {
+		setThemeInstance(darkTheme!);
+		setKeybindings(KeybindingsManager.inMemory({ "tui.select.cancel": "ctrl+g" }));
+	});
+
+	afterEach(() => {
+		setKeybindings(KeybindingsManager.inMemory());
+		vi.restoreAllMocks();
+	});
+
+	it("renders the plan body, prompt, options and footer inside one outlined box", () => {
+		const overlay = new PlanReviewOverlay(
+			"# My Plan\n\nstep one then step two",
+			{ promptTitle: "Plan mode - next step", options: APPROVAL_OPTIONS, helpText: "esc cancel" },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		const out = render(overlay);
+		expect(out).toContain("Plan Review");
+		expect(out).toContain("My Plan");
+		expect(out).toContain("step one then step two");
+		expect(out).toContain("Plan mode - next step");
+		for (const option of APPROVAL_OPTIONS) expect(out).toContain(option);
+		expect(out).toContain("esc cancel");
+		// Outlined like the /copy overlay.
+		expect(out).toContain("┌");
+		expect(out).toContain("│");
+		expect(out).toContain("└");
+	});
+
+	it("confirms the highlighted option on Enter", () => {
+		const onPick = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			"plan",
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick, onCancel: vi.fn() },
+		);
+		overlay.handleInput(ENTER);
+		expect(onPick).toHaveBeenCalledTimes(1);
+		expect(onPick).toHaveBeenCalledWith("Approve and execute");
+	});
+
+	it("moves the option cursor with up/down and confirms the new target", () => {
+		const onPick = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			"plan",
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick, onCancel: vi.fn() },
+		);
+		overlay.handleInput(DOWN);
+		overlay.handleInput(ENTER);
+		expect(onPick).toHaveBeenCalledWith("Approve and compact context");
+
+		onPick.mockClear();
+		overlay.handleInput(UP);
+		overlay.handleInput(ENTER);
+		expect(onPick).toHaveBeenCalledWith("Approve and execute");
+	});
+
+	it("skips disabled options and never confirms them", () => {
+		const onPick = vi.fn();
+		// Disable index 2 ("Approve and keep context").
+		const overlay = new PlanReviewOverlay(
+			"plan",
+			{ promptTitle: "next", options: APPROVAL_OPTIONS, disabledIndices: [2] },
+			{ onPick, onCancel: vi.fn() },
+		);
+		// 0 -> 1 -> (skip 2) -> 3.
+		overlay.handleInput(DOWN);
+		overlay.handleInput(DOWN);
+		overlay.handleInput(ENTER);
+		expect(onPick).toHaveBeenCalledTimes(1);
+		expect(onPick).toHaveBeenCalledWith("Refine plan");
+	});
+
+	it("cancels on the cancel key", () => {
+		const onCancel = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			"plan",
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel },
+		);
+		overlay.handleInput(CANCEL);
+		expect(onCancel).toHaveBeenCalledTimes(1);
+	});
+
+	it("drives the model-tier slider with left/right without changing the option cursor", () => {
+		const changes: number[] = [];
+		const slider: HookSelectorSlider = {
+			caption: "continue with",
+			index: 0,
+			segments: [{ label: "default" }, { label: "slow", detail: "opus" }],
+			onChange: index => changes.push(index),
+		};
+		const onPick = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			"plan",
+			{ promptTitle: "next", options: APPROVAL_OPTIONS, slider },
+			{ onPick, onCancel: vi.fn() },
+		);
+		overlay.handleInput(RIGHT);
+		expect(changes).toEqual([1]);
+		// Clamped at the right edge.
+		overlay.handleInput(RIGHT);
+		expect(changes).toEqual([1]);
+		overlay.handleInput(LEFT);
+		expect(changes).toEqual([1, 0]);
+
+		// The slider must not have moved the option cursor.
+		overlay.handleInput(ENTER);
+		expect(onPick).toHaveBeenCalledWith("Approve and execute");
+	});
+
+	it("invokes the external-editor callback on its key", () => {
+		setKeybindings(KeybindingsManager.inMemory({ "tui.select.cancel": "ctrl+g", "app.editor.external": "ctrl+e" }));
+		const onExternalEditor = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			"plan",
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn(), onExternalEditor },
+		);
+		overlay.handleInput("\x05"); // ctrl+e
+		expect(onExternalEditor).toHaveBeenCalledTimes(1);
+	});
+
+	it("scrolls a long plan to bottom and back to top", () => {
+		const longPlan = Array.from({ length: 200 }, (_, i) => `para ${i}`).join("\n\n");
+		const overlay = new PlanReviewOverlay(
+			longPlan,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		const top = render(overlay);
+		expect(top).toContain("para 0");
+		expect(top).not.toContain("para 199");
+
+		overlay.handleInput("G");
+		const bottom = render(overlay);
+		expect(bottom).toContain("para 199");
+		expect(bottom).not.toContain("para 0");
+
+		overlay.handleInput("g");
+		const backToTop = render(overlay);
+		expect(backToTop).toContain("para 0");
+		expect(backToTop).not.toContain("para 199");
+	});
+
+	it("swaps the displayed plan and resets scroll on setPlanContent", () => {
+		const longPlan = Array.from({ length: 200 }, (_, i) => `para ${i}`).join("\n\n");
+		const overlay = new PlanReviewOverlay(
+			longPlan,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		overlay.handleInput("G"); // scroll away from the top
+		overlay.setPlanContent("# Fresh plan\n\nbrand new body");
+		const out = render(overlay);
+		expect(out).toContain("Fresh plan");
+		expect(out).toContain("brand new body");
+		expect(out).not.toContain("para 199");
+	});
+
+	// Plan with ≥2 headings + nesting, wide enough for the sidebar at width 80.
+	const SECTION_PLAN =
+		"# Overview\n\nintro body\n\n## Goal\n\ngoal body\n\n## Steps\n\nstep body\n\n# Risks\n\nrisk body\n";
+	it("renders no per-line ellipsis in the plan body", () => {
+		const overlay = new PlanReviewOverlay(
+			"# Plan\n\nshort line one\n\nshort line two",
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		const body = render(overlay)
+			.split("\n")
+			.filter(line => line.includes("short line"));
+		expect(body.length).toBeGreaterThan(0);
+		for (const line of body) expect(line).not.toContain("…");
+	});
+
+	it("shows a header-less section sidebar and cycles focus regions with Tab", () => {
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		const out = render(overlay);
+		// Two-column split chrome (┬ joins the title rule over the divider) and the
+		// bare section list — no "Contents" label.
+		expect(out).toContain("┬");
+		expect(out).not.toContain("Contents");
+		expect(out).toContain("Overview");
+		// Tab into the ToC region surfaces its focus-specific help.
+		overlay.handleInput(TAB);
+		const tocFocused = render(overlay);
+		expect(tocFocused).toContain("a annotate");
+		expect(tocFocused).toContain("d delete");
+	});
+
+	it("omits the single plan-title heading from the ToC", () => {
+		// One shallow H1 title + two H2 sections: the title is redundant in the ToC.
+		const overlay = new PlanReviewOverlay(
+			"# Plan: build the thing\n\nintro\n\n## Design\n\nd\n\n## Rollout\n\nr\n",
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		const sidebar = render(overlay)
+			.split("\n")
+			.map(line => line.split("│")[1] ?? "")
+			.join("\n");
+		expect(sidebar).toContain("Design");
+		expect(sidebar).toContain("Rollout");
+		expect(sidebar).not.toContain("build the thing");
+	});
+
+	it("flows past the end of a region into the actions on Down", () => {
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		render(overlay);
+		overlay.handleInput(TAB); // -> toc, first section
+		// Walk to the last ToC entry, then one more Down drops into the actions.
+		for (let i = 0; i < 10; i++) overlay.handleInput(DOWN);
+		const out = render(overlay);
+		// Actions focus restores the option cursor highlight + actions help.
+		expect(out).toContain("⏎ confirm");
+		expect(out).not.toContain("a annotate");
+	});
+
+	it("scrolls the body exactly one line per keystroke in body focus", () => {
+		// Tall enough to overflow any test viewport, so the body genuinely scrolls.
+		const rows = Array.from({ length: 400 }, (_, i) => `L${String(i).padStart(3, "0")}`).join("\n");
+		const overlay = new PlanReviewOverlay(
+			`# Plan\n\n\`\`\`\n${rows}\n\`\`\`\n`,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		const visibleRows = (): string[] =>
+			render(overlay)
+				.split("\n")
+				.map(line => line.match(/L\d\d\d/)?.[0])
+				.filter((m): m is string => m !== undefined);
+		render(overlay); // first render loads the body lines into the ScrollView
+		overlay.handleInput(TAB); // actions -> body (no sidebar with a single heading)
+		// Scroll well past the heading/fence so the window is pure code rows.
+		for (let i = 0; i < 12; i++) overlay.handleInput(DOWN);
+		const before = visibleRows();
+		overlay.handleInput(DOWN);
+		const after = visibleRows();
+		// One-line scroll: the window advances by exactly one row.
+		expect(after[0]).toBe(before[1]);
+	});
+
+	it("jumps the body to a section when the ToC cursor moves", () => {
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		render(overlay);
+		overlay.handleInput(TAB); // -> toc (Overview)
+		overlay.handleInput(DOWN); // -> Goal, scrubbing the body to it
+		const body = render(overlay).split("\n").slice(1, 5).join(" ");
+		expect(body).toContain("Goal");
+	});
+
+	it("deletes the selected section and restores it with undo", () => {
+		const onPlanEdited = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn(), onPlanEdited },
+		);
+		render(overlay);
+		overlay.handleInput(TAB); // -> toc (Overview)
+		overlay.handleInput(DOWN); // -> Goal
+		overlay.handleInput("d");
+		expect(onPlanEdited).toHaveBeenCalled();
+		const edited = onPlanEdited.mock.calls.at(-1)?.[0] as string;
+		expect(edited).not.toContain("## Goal");
+		expect(edited).toContain("## Steps");
+		expect(render(overlay)).not.toContain("goal body");
+
+		overlay.handleInput("u");
+		const restored = render(overlay);
+		expect(restored).toContain("Goal");
+		expect(restored).toContain("goal body");
+	});
+
+	it("annotates a section and emits feedback for the Refine loop", () => {
+		const onFeedbackChange = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn(), onFeedbackChange },
+		);
+		render(overlay);
+		overlay.handleInput(TAB); // -> toc (Overview)
+		overlay.handleInput("a"); // annotate Overview
+		for (const ch of "needs detail") overlay.handleInput(ch);
+		overlay.handleInput(ENTER); // submit
+		const out = render(overlay);
+		expect(out).toContain("needs detail"); // callout in the body
+		expect(out).toContain("✎"); // marker in the sidebar
+		expect(onFeedbackChange).toHaveBeenCalled();
+		const feedback = onFeedbackChange.mock.calls.at(-1)?.[0] as string;
+		expect(feedback).toContain("Overview");
+		expect(feedback).toContain("needs detail");
+	});
+
+	// Click a rendered row. The fullscreen overlay paints from screen row 0, so a
+	// 1-based SGR mouse row equals the rendered-line index + 1.
+	const clickRow = (overlay: PlanReviewOverlay, needle: string, col = 4): boolean => {
+		const lines = overlay.render(80);
+		const row = lines.findIndex(line => stripVTControlCharacters(line).includes(needle));
+		if (row < 0) return false;
+		overlay.handleInput(`\x1b[<0;${col};${row + 1}M`);
+		return true;
+	};
+
+	it("activates an approval option on click", () => {
+		const onPick = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick, onCancel: vi.fn() },
+		);
+		render(overlay);
+		expect(clickRow(overlay, "Refine plan", 10)).toBe(true);
+		expect(onPick).toHaveBeenCalledWith("Refine plan");
+	});
+
+	it("selects a ToC section on click and scrubs the body to it", () => {
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		render(overlay);
+		// Click the "Steps" entry in the sidebar column.
+		expect(clickRow(overlay, "Steps", 4)).toBe(true);
+		const out = render(overlay);
+		expect(out).toContain("a annotate"); // ToC focus
+		// The body scrubbed to the clicked section.
+		expect(out.split("\n").slice(1, 5).join(" ")).toContain("Steps");
+	});
+
+	it("includes deleted sections in the refinement feedback", () => {
+		const onFeedbackChange = vi.fn();
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn(), onPlanEdited: vi.fn(), onFeedbackChange },
+		);
+		render(overlay);
+		overlay.handleInput(TAB); // -> toc (Overview)
+		overlay.handleInput(DOWN); // -> Goal
+		overlay.handleInput("d"); // delete Goal
+		const feedback = onFeedbackChange.mock.calls.at(-1)?.[0] as string;
+		expect(feedback).toContain("Remove these sections:");
+		expect(feedback).toContain("Goal");
+	});
+
+	it("drives the slider with both arrows even when a sidebar is present", () => {
+		const changes: number[] = [];
+		const slider: HookSelectorSlider = {
+			caption: "continue with",
+			index: 0,
+			segments: [{ label: "default" }, { label: "slow", detail: "opus" }],
+			onChange: index => changes.push(index),
+		};
+		const overlay = new PlanReviewOverlay(
+			SECTION_PLAN,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS, slider },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		render(overlay); // establish the sidebar
+		// The sidebar sits beside the body, not the slider, so left must still step
+		// the tier back instead of being stolen to focus the ToC.
+		overlay.handleInput(RIGHT); // 0 -> 1
+		overlay.handleInput(LEFT); // 1 -> 0
+		expect(changes).toEqual([1, 0]);
+		// Focus stayed on the actions region (left did not jump to the ToC).
+		expect(render(overlay)).not.toContain("a annotate");
+	});
+
+	it("fast-scrolls the body with Shift+Arrow", () => {
+		const rows = Array.from({ length: 400 }, (_, i) => `L${String(i).padStart(3, "0")}`).join("\n");
+		const overlay = new PlanReviewOverlay(
+			`# Plan\n\n\`\`\`\n${rows}\n\`\`\`\n`,
+			{ promptTitle: "next", options: APPROVAL_OPTIONS },
+			{ onPick: vi.fn(), onCancel: vi.fn() },
+		);
+		const firstRow = (): number =>
+			Number(
+				(
+					render(overlay)
+						.split("\n")
+						.map(line => line.match(/L\d\d\d/)?.[0])
+						.find((m): m is string => m !== undefined) ?? "L000"
+				).slice(1),
+			);
+		render(overlay); // first render loads the body lines into the ScrollView
+		overlay.handleInput(TAB); // -> body
+		// Scroll into the pure-code region so the leading heading rows don't skew
+		// the absolute row math, then compare a single step to a Shift step.
+		for (let i = 0; i < 10; i++) overlay.handleInput(DOWN);
+		const base = firstRow();
+		overlay.handleInput(SHIFT_DOWN); // Shift+Down — fastScrollLines (5) at once
+		expect(firstRow() - base).toBe(5);
+	});
+});
diff --git a/packages/coding-agent/test/modes/components/plan-toc.test.ts b/packages/coding-agent/test/modes/components/plan-toc.test.ts
new file mode 100644
index 000000000..16bb5b461
--- /dev/null
+++ b/packages/coding-agent/test/modes/components/plan-toc.test.ts
@@ -0,0 +1,91 @@
+import { describe, expect, it } from "bun:test";
+import {
+	joinPlanSections,
+	type PlanSection,
+	parsePlanSections,
+	sectionDeletionSpan,
+	stripInlineMarkdown,
+} from "@oh-my-pi/pi-coding-agent/modes/components/plan-toc";
+
+const titles = (sections: readonly PlanSection[]): string[] => sections.map(s => s.title);
+const levels = (sections: readonly PlanSection[]): number[] => sections.map(s => s.level);
+
+describe("parsePlanSections", () => {
+	it("splits a preamble and one section per ATX heading, tracking depth", () => {
+		const sections = parsePlanSections("intro\n\n# Overview\n\nbody\n\n## Goal\n\ngoal\n\n# Risks\n\nrisk\n");
+		expect(levels(sections)).toEqual([0, 1, 2, 1]);
+		expect(titles(sections)).toEqual(["", "Overview", "Goal", "Risks"]);
+		// Preamble is the only level-0 section; it carries no title.
+		expect(sections[0]!.raw).toBe("intro\n\n");
+	});
+
+	it("emits no preamble section when the document opens with a heading", () => {
+		const sections = parsePlanSections("# Top\n\nbody\n");
+		expect(levels(sections)).toEqual([1]);
+		expect(sections[0]!.title).toBe("Top");
+	});
+
+	it("does not treat '#' inside fenced code blocks as a heading", () => {
+		const sections = parsePlanSections("# Real\n\n```\n# not a heading\n```\n\n~~~\n## also not\n~~~\n");
+		expect(levels(sections)).toEqual([1]);
+		expect(titles(sections)).toEqual(["Real"]);
+	});
+
+	it("requires whitespace after the hashes, so '#tag' is body text", () => {
+		const sections = parsePlanSections("#tag is not a heading\nmore body\n");
+		expect(levels(sections)).toEqual([0]);
+	});
+
+	it("strips inline markdown and closing hashes from titles", () => {
+		const sections = parsePlanSections("## **Goal** & [docs](http://x) ##\n\nbody\n");
+		expect(sections[0]!.title).toBe("Goal & docs");
+	});
+});
+
+describe("joinPlanSections", () => {
+	it("round-trips a newline-terminated document", () => {
+		const text = "intro\n\n# A\n\nbody a\n\n## A1\n\nnested\n\n# B\n\nbody b\n";
+		expect(joinPlanSections(parsePlanSections(text))).toBe(text);
+	});
+
+	it("guarantees a single trailing newline when the source lacks one", () => {
+		expect(joinPlanSections(parsePlanSections("# A\n\nbody"))).toBe("# A\n\nbody\n");
+	});
+
+	it("returns an empty string for an empty document", () => {
+		expect(joinPlanSections(parsePlanSections(""))).toBe("");
+	});
+});
+
+describe("sectionDeletionSpan", () => {
+	const sections = parsePlanSections("intro\n\n# A\n\na\n\n## A1\n\na1\n\n## A2\n\na2\n\n# B\n\nb\n");
+	// Indices: 0 preamble, 1 A(L1), 2 A1(L2), 3 A2(L2), 4 B(L1)
+
+	it("removes a heading together with its deeper-nested children", () => {
+		expect(sectionDeletionSpan(sections, 1)).toEqual([1, 2, 3]);
+	});
+
+	it("removes only a leaf section", () => {
+		expect(sectionDeletionSpan(sections, 2)).toEqual([2]);
+	});
+
+	it("never targets the preamble", () => {
+		expect(sectionDeletionSpan(sections, 0)).toEqual([]);
+	});
+
+	it("joining the surviving sections drops the deleted subtree", () => {
+		const span = new Set(sectionDeletionSpan(sections, 1));
+		const survivors = sections.filter((_, i) => !span.has(i));
+		const result = joinPlanSections(survivors);
+		expect(result).toContain("# B");
+		expect(result).not.toContain("# A");
+		expect(result).not.toContain("## A1");
+	});
+});
+
+describe("stripInlineMarkdown", () => {
+	it("collapses emphasis, code, links, and whitespace to readable text", () => {
+		expect(stripInlineMarkdown("**bold** _it_ `code` [t](u)")).toBe("bold it code t");
+		expect(stripInlineMarkdown("a   b\tc")).toBe("a b c");
+	});
+});
diff --git a/packages/coding-agent/test/modes/components/session-selector-scrollbar.test.ts b/packages/coding-agent/test/modes/components/session-selector-scrollbar.test.ts
new file mode 100644
index 000000000..60d1b227d
--- /dev/null
+++ b/packages/coding-agent/test/modes/components/session-selector-scrollbar.test.ts
@@ -0,0 +1,52 @@
+import { beforeAll, describe, expect, it } from "bun:test";
+import { SessionSelectorComponent } from "../../../src/modes/components/session-selector";
+import { initTheme } from "../../../src/modes/theme/theme";
+import type { SessionInfo } from "../../../src/session/session-manager";
+
+beforeAll(() => {
+	initTheme();
+});
+
+const THUMB = "\u2588"; // ScrollView thumb glyph
+
+function makeSessions(count: number): SessionInfo[] {
+	return Array.from({ length: count }, (_, i) => ({
+		path: `/work/TITLE_${i}.jsonl`,
+		id: `id-${i}`,
+		cwd: "/work",
+		title: `TITLE_${i}`,
+		created: new Date("2024-01-01T00:00:00Z"),
+		modified: new Date("2024-01-02T00:00:00Z"),
+		messageCount: 1,
+		size: 1024,
+		firstMessage: `body content ${i}`,
+		allMessagesText: `body content ${i}`,
+	}));
+}
+
+function makeSelector(sessions: SessionInfo[], rows: number): SessionSelectorComponent {
+	return new SessionSelectorComponent(
+		sessions,
+		() => {},
+		() => {},
+		() => {},
+		{ getTerminalRows: () => rows },
+	);
+}
+
+describe("SessionSelectorComponent scrollbar", () => {
+	it("renders the ScrollView thumb when sessions overflow the viewport", () => {
+		// 50 titled sessions cannot fit a 30-row viewport, so the picker windows
+		// them and must surface the shared right-edge scrollbar (the /resume
+		// overflow path the user reported).
+		const out = makeSelector(makeSessions(50), 30).render(80).join("\n");
+		expect(out).toContain(THUMB);
+		// The old text position indicator must be gone.
+		expect(out).not.toContain("(1/50)");
+	});
+
+	it("omits the scrollbar when every session fits", () => {
+		const out = makeSelector(makeSessions(2), 40).render(80).join("\n");
+		expect(out).not.toContain(THUMB);
+	});
+});
diff --git a/packages/coding-agent/test/modes/components/transcript-container.test.ts b/packages/coding-agent/test/modes/components/transcript-container.test.ts
index 6baf1bfac..c97cbd5a5 100644
--- a/packages/coding-agent/test/modes/components/transcript-container.test.ts
+++ b/packages/coding-agent/test/modes/components/transcript-container.test.ts
@@ -6,6 +6,7 @@ import { resetSettingsForTest, Settings } from "../../../src/config/settings";
 import { AssistantMessageComponent } from "../../../src/modes/components/assistant-message";
 import { TranscriptContainer } from "../../../src/modes/components/transcript-container";
 import { initTheme } from "../../../src/modes/theme/theme";
+import { USER_INTERRUPT_LABEL } from "../../../src/session/messages";
 
 // Models a transcript block that re-lays-out (tool preview collapsing, assistant
 // message finalizing, late async result) after it has scrolled past the live
@@ -107,16 +108,16 @@ describe("TranscriptContainer", () => {
 		// A newer block makes `a` non-live; it now replays its last live render.
 		const b = new MutableBlock(["b1"]);
 		container.addChild(b);
-		expect(container.render(40)).toEqual(["a2", "b1"]);
+		expect(container.render(40)).toEqual(["a2", "", "b1"]);
 
 		// A post-freeze mutation of `a` (its collapse/re-layout) is NOT reflected —
 		// the committed rows stay stable so no stale duplicate enters scrollback.
 		a.set(["a3-collapsed"]);
-		expect(container.render(40)).toEqual(["a2", "b1"]);
+		expect(container.render(40)).toEqual(["a2", "", "b1"]);
 
 		// The live block still updates freely.
 		b.set(["b2"]);
-		expect(container.render(40)).toEqual(["a2", "b2"]);
+		expect(container.render(40)).toEqual(["a2", "", "b2"]);
 	});
 
 	it("reports the live block start for native scrollback pinning (ED3-risk)", () => {
@@ -127,12 +128,12 @@ describe("TranscriptContainer", () => {
 		container.addChild(a);
 		container.addChild(b);
 
-		expect(container.render(40)).toEqual(["a1", "a2", "b1"]);
-		expect(container.getNativeScrollbackLiveRegionStart()).toBe(2);
+		expect(container.render(40)).toEqual(["a1", "a2", "", "b1"]);
+		expect(container.getNativeScrollbackLiveRegionStart()).toBe(3);
 
 		b.set(["b1", "b2"]);
-		expect(container.render(40)).toEqual(["a1", "a2", "b1", "b2"]);
-		expect(container.getNativeScrollbackLiveRegionStart()).toBe(2);
+		expect(container.render(40)).toEqual(["a1", "a2", "", "b1", "b2"]);
+		expect(container.getNativeScrollbackLiveRegionStart()).toBe(3);
 	});
 
 	it("seals the prior block at its final content when finalize+append coalesce (ED3-risk)", () => {
@@ -151,11 +152,11 @@ describe("TranscriptContainer", () => {
 
 		// The transition frame must seal `a` at its final content, not the stale
 		// mid-stream snapshot ("Nat") it last rendered while live.
-		expect(container.render(40)).toEqual(["Natives built, now...", "b1"]);
+		expect(container.render(40)).toEqual(["Natives built, now...", "", "b1"]);
 
 		// Once sealed, a later re-layout of `a` stays frozen until the next thaw.
 		a.set(["a-collapsed"]);
-		expect(container.render(40)).toEqual(["Natives built, now...", "b1"]);
+		expect(container.render(40)).toEqual(["Natives built, now...", "", "b1"]);
 	});
 
 	it("thaw() reconciles frozen blocks to their current state", () => {
@@ -167,10 +168,29 @@ describe("TranscriptContainer", () => {
 		container.addChild(b);
 		container.render(40);
 		a.set(["a-final"]);
-		expect(container.render(40)).toEqual(["a1", "b1"]); // frozen
+		expect(container.render(40)).toEqual(["a1", "", "b1"]); // frozen
 
 		container.thaw();
-		expect(container.render(40)).toEqual(["a-final", "b1"]); // reconciled
+		expect(container.render(40)).toEqual(["a-final", "", "b1"]); // reconciled
+	});
+
+	it("invalidate() retires frozen snapshots so resetDisplay reflects current state", () => {
+		// resetDisplay() (Ctrl+L, and the Ctrl+O expand path) reflows by calling
+		// TUI.invalidate(), which propagates to this container. That must retire the
+		// frozen snapshots the same way thaw() does, or a forced full replay would
+		// still emit the pre-mutation (e.g. collapsed) render.
+		riskFlag.eagerEraseScrollbackRisk = true;
+		const container = new TranscriptContainer();
+		const a = new MutableBlock(["a-collapsed"]);
+		const b = new MutableBlock(["b1"]);
+		container.addChild(a);
+		container.addChild(b);
+		container.render(40);
+		a.set(["a-expanded-1", "a-expanded-2"]);
+		expect(container.render(40)).toEqual(["a-collapsed", "", "b1"]); // frozen
+
+		container.invalidate();
+		expect(container.render(40)).toEqual(["a-expanded-1", "a-expanded-2", "", "b1"]);
 	});
 
 	it("recomputes a frozen block on a width change", () => {
@@ -182,9 +202,9 @@ describe("TranscriptContainer", () => {
 		container.addChild(b);
 		container.render(40);
 		a.set(["a-reflowed"]);
-		expect(container.render(40)).toEqual(["a1", "b1"]); // frozen at width 40
+		expect(container.render(40)).toEqual(["a1", "", "b1"]); // frozen at width 40
 		// A resize is an explicit rebuild that reconciles history, so recompute.
-		expect(container.render(80)).toEqual(["a-reflowed", "b1"]);
+		expect(container.render(80)).toEqual(["a-reflowed", "", "b1"]);
 	});
 
 	it("renders every block live on terminals that can rebuild history", () => {
@@ -198,7 +218,7 @@ describe("TranscriptContainer", () => {
 		// No freezing: a non-live block's mutation is reflected (the renderer can
 		// rebuild committed history on these terminals).
 		a.set(["a-updated"]);
-		expect(container.render(40)).toEqual(["a-updated", "b1"]);
+		expect(container.render(40)).toEqual(["a-updated", "", "b1"]);
 	});
 
 	it("keeps an unfinalized block live when a finalized block is appended below it (ED3-risk)", () => {
@@ -213,7 +233,7 @@ describe("TranscriptContainer", () => {
 		// tool while it is still streaming. The tool must NOT freeze here.
 		const card = new MutableBlock(["rule card"]);
 		container.addChild(card);
-		expect(container.render(40)).toEqual(["write (streaming)", "rule card"]);
+		expect(container.render(40)).toEqual(["write (streaming)", "", "rule card"]);
 		// The live region begins at the unfinalized tool, not the bottom card.
 		expect(container.getNativeScrollbackLiveRegionStart()).toBe(0);
 
@@ -221,11 +241,11 @@ describe("TranscriptContainer", () => {
 		// tool was kept live, its final content is reflected — the bug was it
 		// freezing on the streaming preview and never showing the result.
 		tool.finalize(["✔ write: 4 lines"]);
-		expect(container.render(40)).toEqual(["✔ write: 4 lines", "rule card"]);
+		expect(container.render(40)).toEqual(["✔ write: 4 lines", "", "rule card"]);
 
 		// Now finalized, it freezes: a later re-layout stays put until the next thaw.
 		tool.set(["collapsed"]);
-		expect(container.render(40)).toEqual(["✔ write: 4 lines", "rule card"]);
+		expect(container.render(40)).toEqual(["✔ write: 4 lines", "", "rule card"]);
 	});
 
 	it("keeps a streaming assistant live so an abort label can land after status rows below it (ED3-risk)", () => {
@@ -251,14 +271,14 @@ describe("TranscriptContainer", () => {
 			makeAssistantMessage({
 				content: [{ type: "text", text: "The config file write went through despite the interruption." }],
 				stopReason: "aborted",
-				errorMessage: "Operation aborted",
+				errorMessage: USER_INTERRUPT_LABEL,
 			}),
 		);
 		assistant.markTranscriptBlockFinalized();
 
 		const rendered = plain(container.render(80));
 		expect(rendered).toContain("The config file write went through despite the interruption.");
-		expect(rendered).toContain("Operation aborted");
+		expect(rendered).toContain(USER_INTERRUPT_LABEL);
 		expect(rendered).toContain("Copied raw SSE stream");
 		expect(container.getNativeScrollbackLiveRegionStart()).not.toBe(0);
 	});
@@ -272,18 +292,86 @@ describe("TranscriptContainer", () => {
 		container.addChild(sealed);
 		container.addChild(pending);
 		container.addChild(card);
-		expect(container.render(40)).toEqual(["done", "pending", "card"]);
+		expect(container.render(40)).toEqual(["done", "", "pending", "", "card"]);
 		// Live region starts at the pending block (offset 1), so the already-sealed
 		// leading block can commit while pending + card stay repaintable.
-		expect(container.getNativeScrollbackLiveRegionStart()).toBe(1);
+		expect(container.getNativeScrollbackLiveRegionStart()).toBe(2);
 
 		// The leading sealed block freezes; its re-layout is not reflected.
 		sealed.set(["done-collapsed"]);
-		expect(container.render(40)).toEqual(["done", "pending", "card"]);
+		expect(container.render(40)).toEqual(["done", "", "pending", "", "card"]);
 
 		// The pending block updates freely while live.
 		pending.finalize(["pending-final"]);
-		expect(container.render(40)).toEqual(["done", "pending-final", "card"]);
-		expect(container.getNativeScrollbackLiveRegionStart()).toBe(2);
+		expect(container.render(40)).toEqual(["done", "", "pending-final", "", "card"]);
+		expect(container.getNativeScrollbackLiveRegionStart()).toBe(4);
+	});
+});
+
+describe("TranscriptContainer spacing", () => {
+	it("inserts exactly one blank line between consecutive blocks", () => {
+		riskFlag.eagerEraseScrollbackRisk = false;
+		const container = new TranscriptContainer();
+		container.addChild(new MutableBlock(["a"]));
+		container.addChild(new MutableBlock(["b"]));
+		container.addChild(new MutableBlock(["c"]));
+		// One separator between each block; none above the first.
+		expect(container.render(40)).toEqual(["a", "", "b", "", "c"]);
+	});
+
+	it("strips a block's plain-blank top/bottom padding", () => {
+		riskFlag.eagerEraseScrollbackRisk = false;
+		const container = new TranscriptContainer();
+		container.addChild(new MutableBlock(["a"]));
+		// Leading Spacer rows + a trailing paddingY row collapse to just the body.
+		container.addChild(new MutableBlock(["", "   ", "body", ""]));
+		expect(container.render(40)).toEqual(["a", "", "body"]);
+	});
+
+	it("preserves background-colored padding rows (block-internal design)", () => {
+		riskFlag.eagerEraseScrollbackRisk = false;
+		const bgPad = "\x1b[48;2;0;0;0m   \x1b[0m";
+		const container = new TranscriptContainer();
+		container.addChild(new MutableBlock(["a"]));
+		// The ANSI-bearing padding row is not "plain blank", so it survives stripping.
+		container.addChild(new MutableBlock([bgPad, "x", bgPad]));
+		expect(container.render(40)).toEqual(["a", "", bgPad, "x", bgPad]);
+	});
+
+	it("does not double the gap when a block carries its own trailing blank", () => {
+		riskFlag.eagerEraseScrollbackRisk = false;
+		const container = new TranscriptContainer();
+		// The trailing blank is stripped, so only the container's separator remains.
+		container.addChild(new MutableBlock(["note", ""]));
+		container.addChild(new MutableBlock(["b"]));
+		expect(container.render(40)).toEqual(["note", "", "b"]);
+	});
+
+	it("does not inject separators within a single block's rows", () => {
+		riskFlag.eagerEraseScrollbackRisk = false;
+		const container = new TranscriptContainer();
+		// An IRC card / file-mention list wrapped as one block stays tight inside.
+		container.addChild(new MutableBlock(["header", "  body1", "  body2"]));
+		expect(container.render(40)).toEqual(["header", "  body1", "  body2"]);
+	});
+
+	it("drops a blank-only block without leaving a stray gap", () => {
+		riskFlag.eagerEraseScrollbackRisk = false;
+		const container = new TranscriptContainer();
+		container.addChild(new MutableBlock(["a"]));
+		container.addChild(new MutableBlock(["", "  "]));
+		container.addChild(new MutableBlock(["b"]));
+		expect(container.render(40)).toEqual(["a", "", "b"]);
+	});
+
+	it("counts the separator into the committed prefix below the live region (ED3-risk)", () => {
+		riskFlag.eagerEraseScrollbackRisk = true;
+		const container = new TranscriptContainer();
+		// A finalized block, then a still-live block below it.
+		container.addChild(new MutableBlock(["a1", "a2"]));
+		container.addChild(new StreamingBlock(["b"]));
+		// Separator sits at index 2; the live block's content begins at index 3.
+		expect(container.render(40)).toEqual(["a1", "a2", "", "b"]);
+		expect(container.getNativeScrollbackLiveRegionStart()).toBe(3);
 	});
 });
diff --git a/packages/coding-agent/test/modes/controllers/command-controller-hotkeys.test.ts b/packages/coding-agent/test/modes/controllers/command-controller-hotkeys.test.ts
index 3bc19bfef..ee9cfc284 100644
--- a/packages/coding-agent/test/modes/controllers/command-controller-hotkeys.test.ts
+++ b/packages/coding-agent/test/modes/controllers/command-controller-hotkeys.test.ts
@@ -6,8 +6,9 @@ describe("buildHotkeysMarkdown", () => {
 		const displayStrings: Record<string, string> = {
 			"app.clipboard.copyLine": "Alt+Shift+L",
 			"app.clipboard.copyPrompt": "Ctrl+Shift+P",
-			"app.plan.toggle": "Alt+M",
+			"app.plan.toggle": "Alt+Shift+P",
 			"app.tools.expand": "Ctrl+O",
+			"app.display.reset": "Ctrl+L",
 			"app.interrupt": "Esc",
 			"app.clear": "Ctrl+C",
 			"app.exit": "Ctrl+D",
@@ -16,7 +17,7 @@ describe("buildHotkeysMarkdown", () => {
 			"app.model.cycleForward": "Ctrl+P",
 			"app.model.cycleBackward": "Shift+Ctrl+P",
 			"app.model.selectTemporary": "Ctrl+Shift+L",
-			"app.model.select": "Ctrl+L",
+			"app.model.select": "Alt+M",
 			"app.history.search": "Ctrl+R",
 			"app.thinking.toggle": "Ctrl+T",
 			"app.editor.external": "Ctrl+G",
@@ -35,8 +36,9 @@ describe("buildHotkeysMarkdown", () => {
 		expect(lines[0]).toBe("**Navigation**");
 		expect(markdown).toContain("| `Ctrl+Shift+P` | Copy whole prompt |");
 		expect(markdown).toContain("| `Ctrl+Shift+L` | Select model (temporary) |");
-		expect(markdown).toContain("| `Ctrl+L` | Select model (set roles) |");
-		expect(markdown).toContain("| `Alt+M` | Toggle plan mode |");
+		expect(markdown).toContain("| `Alt+M` | Select model (set roles) |");
+		expect(markdown).toContain("| `Ctrl+L` | Reset terminal display |");
+		expect(markdown).toContain("| `Alt+Shift+P` | Toggle plan mode |");
 		expect(markdown).toContain("| `#` | Open prompt actions |");
 		for (const line of lines) {
 			if (line.length === 0) continue;
@@ -53,6 +55,9 @@ describe("buildHotkeysMarkdown", () => {
 						return "";
 					}
 					if (action === "app.model.select") {
+						return "Alt+M";
+					}
+					if (action === "app.display.reset") {
 						return "Ctrl+L";
 					}
 					return "Ctrl+K";
@@ -61,6 +66,6 @@ describe("buildHotkeysMarkdown", () => {
 		});
 
 		expect(markdown).toContain("| `Disabled` | Select model (temporary) |");
-		expect(markdown).toContain("| `Ctrl+L` | Select model (set roles) |");
+		expect(markdown).toContain("| `Alt+M` | Select model (set roles) |");
 	});
 });
diff --git a/packages/coding-agent/test/modes/controllers/copy-command.test.ts b/packages/coding-agent/test/modes/controllers/copy-command.test.ts
deleted file mode 100644
index b87f4d178..000000000
--- a/packages/coding-agent/test/modes/controllers/copy-command.test.ts
+++ /dev/null
@@ -1,53 +0,0 @@
-import { afterEach, describe, expect, it, vi } from "bun:test";
-import { CommandController } from "@oh-my-pi/pi-coding-agent/modes/controllers/command-controller";
-import type { InteractiveModeContext } from "@oh-my-pi/pi-coding-agent/modes/types";
-import * as native from "@oh-my-pi/pi-natives";
-
-function createController(options: { assistantText?: string; hasAssistantMessage?: boolean; handoffText?: string }) {
-	const showStatus = vi.fn();
-	const showError = vi.fn();
-	const ctx = {
-		session: {
-			getLastAssistantText: () => options.assistantText,
-			hasCopyCandidateAssistantMessage: () => options.hasAssistantMessage ?? options.assistantText !== undefined,
-			getLastVisibleHandoffText: () => options.handoffText,
-		},
-		showStatus,
-		showError,
-	} as unknown as InteractiveModeContext;
-
-	return { controller: new CommandController(ctx), showStatus, showError };
-}
-
-describe("/copy command", () => {
-	afterEach(() => {
-		vi.restoreAllMocks();
-	});
-
-	it("falls back to the fresh handoff context when no assistant message exists", () => {
-		const copySpy = vi.spyOn(native, "copyToClipboard").mockImplementation(() => undefined);
-		const { controller, showStatus, showError } = createController({
-			handoffText: "<handoff-context>\n## Goal\nContinue\n</handoff-context>",
-		});
-
-		controller.handleCopyCommand();
-
-		expect(copySpy).toHaveBeenCalledWith("<handoff-context>\n## Goal\nContinue\n</handoff-context>");
-		expect(showStatus).toHaveBeenCalledWith("Copied handoff context to clipboard");
-		expect(showError).not.toHaveBeenCalled();
-	});
-
-	it("does not fall back to stale handoff context after a textless assistant response", () => {
-		const copySpy = vi.spyOn(native, "copyToClipboard").mockImplementation(() => undefined);
-		const { controller, showStatus, showError } = createController({
-			hasAssistantMessage: true,
-			handoffText: "<handoff-context>\n## Goal\nContinue\n</handoff-context>",
-		});
-
-		controller.handleCopyCommand();
-
-		expect(copySpy).not.toHaveBeenCalled();
-		expect(showStatus).not.toHaveBeenCalled();
-		expect(showError).toHaveBeenCalledWith("No agent messages to copy yet.");
-	});
-});
diff --git a/packages/coding-agent/test/modes/controllers/event-controller-abort-guard.test.ts b/packages/coding-agent/test/modes/controllers/event-controller-abort-guard.test.ts
index b35db298f..e25eab0a4 100644
--- a/packages/coding-agent/test/modes/controllers/event-controller-abort-guard.test.ts
+++ b/packages/coding-agent/test/modes/controllers/event-controller-abort-guard.test.ts
@@ -2,12 +2,12 @@
  * Regression test for the abort-guard on `EventController.sendCompletionNotification`.
  *
  * Bug: a user Ctrl+C on the `ask` tool selector throws `ToolAbortError`,
- * the turn ends with `stopReason === "aborted"`, and `handleBackgroundEvent`
- * fires `sendCompletionNotification()` unconditionally. The pre-fix code
- * then produced a misleading "Task complete" desktop toast for a turn that
- * never actually completed. The fix mirrors the `stopReason !== "aborted"`
- * pattern already used by `#currentContextTokens`, `#handleMessageEnd`, and
- * the retry / TTSR / compaction skip paths in `agent-session.ts`.
+ * the turn ends with `stopReason === "aborted"`, and `#handleAgentEnd`
+ * fires `sendCompletionNotification()`. Without a guard this produced a
+ * misleading "Task complete" desktop toast for a turn that never actually
+ * completed. The fix mirrors the `stopReason !== "aborted"` pattern already
+ * used by `#currentContextTokens`, `#handleMessageEnd`, and the
+ * retry / TTSR / compaction skip paths in `agent-session.ts`.
  */
 import { afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
 import * as fs from "node:fs";
@@ -49,8 +49,6 @@ function makeAssistantMessage(stopReason: StopReason): AssistantMessage {
 
 function makeContext(lastMessage: AssistantMessage | undefined): InteractiveModeContext {
 	return {
-		// sendCompletionNotification only fires when backgrounded.
-		isBackgrounded: true,
 		sessionManager: {
 			getSessionName: () => "test-session",
 		},
@@ -96,16 +94,6 @@ describe("EventController.sendCompletionNotification — abort guard", () => {
 		expect(spy).toHaveBeenCalledTimes(1);
 	});
 
-	it("honors the existing isBackgrounded gate (no notification when foreground)", () => {
-		const spy = vi.spyOn(TERMINAL, "sendNotification").mockImplementation(() => {});
-		settings.override("completion.notify", "on");
-		const ctx = makeContext(makeAssistantMessage("stop"));
-		(ctx as unknown as { isBackgrounded: boolean }).isBackgrounded = false;
-		const controller = new EventController(ctx);
-		controller.sendCompletionNotification();
-		expect(spy).toHaveBeenCalledTimes(0);
-	});
-
 	it("honors the existing completion.notify=off gate", () => {
 		const spy = vi.spyOn(TERMINAL, "sendNotification").mockImplementation(() => {});
 		settings.override("completion.notify", "off");
diff --git a/packages/coding-agent/test/modes/controllers/event-controller-idle-compaction.test.ts b/packages/coding-agent/test/modes/controllers/event-controller-idle-compaction.test.ts
index 48f4e51b1..39b37e38c 100644
--- a/packages/coding-agent/test/modes/controllers/event-controller-idle-compaction.test.ts
+++ b/packages/coding-agent/test/modes/controllers/event-controller-idle-compaction.test.ts
@@ -48,7 +48,6 @@ describe("EventController idle compaction teardown", () => {
 		const runIdleCompaction = vi.fn();
 		const context = {
 			isInitialized: true,
-			isBackgrounded: false,
 			loadingAnimation: undefined,
 			streamingComponent: undefined,
 			streamingMessage: undefined,
diff --git a/packages/coding-agent/test/modes/controllers/event-controller-message-start.test.ts b/packages/coding-agent/test/modes/controllers/event-controller-message-start.test.ts
index 8e80c6c28..f5a796393 100644
--- a/packages/coding-agent/test/modes/controllers/event-controller-message-start.test.ts
+++ b/packages/coding-agent/test/modes/controllers/event-controller-message-start.test.ts
@@ -194,11 +194,11 @@ describe("EventController IRC expiry", () => {
 
 		await controller.handleEvent({ type: "irc_message", message });
 
-		expect(chatContainer.children).toHaveLength(2);
+		expect(chatContainer.children).toHaveLength(1);
 		expect(requestRender).toHaveBeenCalledTimes(1);
 
 		vi.advanceTimersByTime(9_999);
-		expect(chatContainer.children).toHaveLength(2);
+		expect(chatContainer.children).toHaveLength(1);
 
 		vi.advanceTimersByTime(1);
 		expect(chatContainer.children).toHaveLength(0);
@@ -215,7 +215,7 @@ describe("EventController IRC expiry", () => {
 		await controller.handleEvent({ type: "irc_message", message });
 
 		expect(addMessageToChat).toHaveBeenCalledTimes(1);
-		expect(chatContainer.children).toHaveLength(2);
+		expect(chatContainer.children).toHaveLength(1);
 		vi.advanceTimersByTime(10_000);
 		expect(chatContainer.children).toHaveLength(0);
 	});
@@ -230,7 +230,7 @@ describe("EventController IRC expiry", () => {
 		controller.dispose();
 		vi.advanceTimersByTime(10_000);
 
-		expect(chatContainer.children).toHaveLength(2);
+		expect(chatContainer.children).toHaveLength(1);
 		expect(requestRender).toHaveBeenCalledTimes(1);
 	});
 });
diff --git a/packages/coding-agent/test/modes/controllers/event-controller-read-grouping.test.ts b/packages/coding-agent/test/modes/controllers/event-controller-read-grouping.test.ts
new file mode 100644
index 000000000..a815d5e69
--- /dev/null
+++ b/packages/coding-agent/test/modes/controllers/event-controller-read-grouping.test.ts
@@ -0,0 +1,147 @@
+/**
+ * Read-group accretion across assistant completions.
+ *
+ * Reasoning models (and codex-style providers) frequently emit one `read` per
+ * completion as `[thinking?, toolCall]` rather than batching parallel calls.
+ * The transcript should still collapse an uninterrupted run of those reads into
+ * a single {@link ReadToolGroupComponent}; a completion that renders visible
+ * content (non-empty text/thinking) is the only thing that breaks the run, so a
+ * fresh group starts after it.
+ *
+ * Regression: every completion used to reset the active group at `message_start`,
+ * so consecutive single-read completions never grouped (each rendered as its own
+ * one-entry block).
+ */
+import { afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "bun:test";
+import type { AssistantMessage } from "@oh-my-pi/pi-ai";
+import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import { ReadToolGroupComponent } from "@oh-my-pi/pi-coding-agent/modes/components/read-tool-group";
+import { EventController } from "@oh-my-pi/pi-coding-agent/modes/controllers/event-controller";
+import { initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
+import type { InteractiveModeContext } from "@oh-my-pi/pi-coding-agent/modes/types";
+import type { AgentSessionEvent } from "@oh-my-pi/pi-coding-agent/session/agent-session";
+import { Container } from "@oh-my-pi/pi-tui";
+
+beforeAll(async () => {
+	await initTheme(false, undefined, undefined, "dark", "light");
+});
+
+beforeEach(async () => {
+	resetSettingsForTest();
+	await Settings.init({ inMemory: true });
+});
+
+afterEach(() => {
+	resetSettingsForTest();
+	vi.restoreAllMocks();
+});
+
+type Block = AssistantMessage["content"][number];
+
+function read(path: string): Block {
+	return { type: "toolCall", id: `read-${path}`, name: "read", arguments: { path } } as Block;
+}
+
+function thinking(text: string): Block {
+	return { type: "thinking", thinking: text } as Block;
+}
+
+function assistantMessage(content: Block[]): AssistantMessage {
+	return {
+		role: "assistant",
+		content,
+		api: "openai-codex-responses",
+		provider: "openai-codex",
+		model: "gpt-5.5",
+		stopReason: "toolUse",
+		usage: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+			totalTokens: 0,
+			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+		},
+		timestamp: Date.now(),
+	};
+}
+
+function createFixture() {
+	const chatContainer = new Container();
+	const ctx = {
+		isInitialized: true,
+		init: vi.fn(async () => {}),
+		statusLine: { invalidate: vi.fn() },
+		updateEditorTopBorder: vi.fn(),
+		ui: { requestRender: vi.fn(), setEagerNativeScrollbackRebuild: vi.fn(), imageBudget: undefined },
+		chatContainer,
+		pendingTools: new Map(),
+		settings: { get: () => false },
+		toolOutputExpanded: false,
+		hideThinkingBlock: false,
+		setWorkingMessage: vi.fn(),
+		session: { getToolByName: () => undefined, extensionRunner: undefined },
+	} as unknown as InteractiveModeContext;
+	return { controller: new EventController(ctx), chatContainer };
+}
+
+/** Drive one assistant completion: message_start then a single full message_update. */
+async function streamCompletion(controller: EventController, content: Block[]): Promise<void> {
+	const message = assistantMessage(content);
+	await controller.handleEvent({ type: "message_start", message } as AgentSessionEvent);
+	await controller.handleEvent({ type: "message_update", message } as AgentSessionEvent);
+}
+
+function readGroups(chatContainer: Container): ReadToolGroupComponent[] {
+	return chatContainer.children.filter((c): c is ReadToolGroupComponent => c instanceof ReadToolGroupComponent);
+}
+
+function header(group: ReadToolGroupComponent): string {
+	return Bun.stripANSI(group.render(120).join("\n")).split("\n")[0] ?? "";
+}
+
+describe("EventController read-group accretion", () => {
+	it("collapses a run of single-read completions into one group (mixed/empty thinking)", async () => {
+		const { controller, chatContainer } = createFixture();
+
+		// Mirrors the reported session: first read carries reasoning, the rest have
+		// empty or absent thinking. None of them should break the run.
+		await streamCompletion(controller, [thinking("Considering performance optimizations"), read("a.ts:180-250")]);
+		await streamCompletion(controller, [thinking(""), read("a.ts:1-120")]);
+		await streamCompletion(controller, [read("b.ts:1-220")]);
+		await streamCompletion(controller, [read("b.ts:450-535")]);
+
+		const groups = readGroups(chatContainer);
+		expect(groups.length).toBe(1);
+		expect(header(groups[0]!)).toContain("Read (4)");
+	});
+
+	it("starts a new group after a completion that renders visible reasoning", async () => {
+		const { controller, chatContainer } = createFixture();
+
+		await streamCompletion(controller, [read("a.ts:1-50")]);
+		await streamCompletion(controller, [read("a.ts:51-100")]);
+		// Visible reasoning is a separator: the next reads form a distinct group.
+		await streamCompletion(controller, [thinking("Now let me check the other files"), read("c.ts:1-40")]);
+		await streamCompletion(controller, [read("c.ts:41-80")]);
+
+		const groups = readGroups(chatContainer);
+		expect(groups.length).toBe(2);
+		expect(header(groups[0]!)).toContain("Read (2)");
+		expect(header(groups[1]!)).toContain("Read (2)");
+	});
+
+	it("keeps the active group repaintable until it is finalized", async () => {
+		const { controller, chatContainer } = createFixture();
+
+		await streamCompletion(controller, [read("a.ts:1-50")]);
+		const [group] = readGroups(chatContainer);
+		// While it is the active run the block must stay in the live region so its
+		// header can re-layout from `Read <path>` to `Read (N)` on risk terminals.
+		expect(group!.isTranscriptBlockFinalized()).toBe(false);
+
+		// A visible-reasoning completion breaks the run and finalizes the prior group.
+		await streamCompletion(controller, [thinking("done exploring"), read("b.ts:1-50")]);
+		expect(group!.isTranscriptBlockFinalized()).toBe(true);
+	});
+});
diff --git a/packages/coding-agent/test/modes/controllers/event-controller-tool-render-mode.test.ts b/packages/coding-agent/test/modes/controllers/event-controller-tool-render-mode.test.ts
index f0bb4d15c..c63eb3d82 100644
--- a/packages/coding-agent/test/modes/controllers/event-controller-tool-render-mode.test.ts
+++ b/packages/coding-agent/test/modes/controllers/event-controller-tool-render-mode.test.ts
@@ -6,11 +6,12 @@ import type { AgentSessionEvent } from "@oh-my-pi/pi-coding-agent/session/agent-
 
 function createContext() {
 	const setEagerNativeScrollbackRebuild = vi.fn();
+	const ensureLoadingAnimation = vi.fn();
 	const pendingTools = new Map<string, unknown>();
 	const chatContainer = { addChild: vi.fn(), removeChild: vi.fn() };
 	const ctx = {
 		isInitialized: true,
-		isBackgrounded: false,
+		settings: { get: () => false },
 		statusLine: { invalidate: vi.fn() },
 		updateEditorTopBorder: vi.fn(),
 		pendingTools,
@@ -18,6 +19,7 @@ function createContext() {
 		hideThinkingBlock: false,
 		editor: { getText: vi.fn(() => "") },
 		flushPendingModelSwitch: vi.fn(),
+		sessionManager: { getSessionName: () => undefined },
 		session: {
 			agent: { state: { messages: [] } },
 			isCompacting: false,
@@ -25,8 +27,10 @@ function createContext() {
 			retryAttempt: 0,
 		},
 		ui: { setEagerNativeScrollbackRebuild, requestRender: vi.fn() },
+		clearPinnedError: vi.fn(),
+		ensureLoadingAnimation,
 	} as unknown as InteractiveModeContext;
-	return { ctx, pendingTools, setEagerNativeScrollbackRebuild };
+	return { ctx, pendingTools, setEagerNativeScrollbackRebuild, ensureLoadingAnimation };
 }
 
 // A tool_execution_update for an id that is not pending is a no-op in its handler,
@@ -67,6 +71,17 @@ describe("EventController tool render mode", () => {
 		vi.restoreAllMocks();
 	});
 
+	it("enables eager native scrollback rebuild before starting the idle Working loader", async () => {
+		const { ctx, ensureLoadingAnimation, setEagerNativeScrollbackRebuild } = createContext();
+		const controller = new EventController(ctx);
+
+		await controller.handleEvent({ type: "agent_start" } as unknown as AgentSessionEvent);
+
+		expect(setEagerNativeScrollbackRebuild).toHaveBeenCalledWith(true);
+		expect(setEagerNativeScrollbackRebuild.mock.invocationCallOrder[0]!).toBeLessThan(
+			ensureLoadingAnimation.mock.invocationCallOrder[0]!,
+		);
+	});
 	it("enables eager native scrollback rebuild while a foreground tool is pending", async () => {
 		const { ctx, pendingTools, setEagerNativeScrollbackRebuild } = createContext();
 		const controller = new EventController(ctx);
diff --git a/packages/coding-agent/test/modes/controllers/input-controller-tool-expansion.test.ts b/packages/coding-agent/test/modes/controllers/input-controller-tool-expansion.test.ts
index bcf3900ea..38370dca3 100644
--- a/packages/coding-agent/test/modes/controllers/input-controller-tool-expansion.test.ts
+++ b/packages/coding-agent/test/modes/controllers/input-controller-tool-expansion.test.ts
@@ -3,20 +3,25 @@ import { InputController } from "../../../src/modes/controllers/input-controller
 import type { InteractiveModeContext } from "../../../src/modes/types";
 
 describe("InputController tool output expansion", () => {
-	it("allows unknown viewport mutation when toggling tool output expansion", () => {
+	it("expands children and forces a full display reset to bypass frozen snapshots", () => {
 		const expandable = { setExpanded: vi.fn() };
 		const inert = { render: vi.fn(() => []) };
 		const requestRender = vi.fn();
+		const resetDisplay = vi.fn();
 		const ctx = {
 			toolOutputExpanded: false,
 			chatContainer: { children: [expandable, inert] },
-			ui: { requestRender },
+			ui: { requestRender, resetDisplay },
 		} as unknown as InteractiveModeContext;
 
 		new InputController(ctx).toggleToolOutputExpansion();
 
 		expect(ctx.toolOutputExpanded).toBe(true);
 		expect(expandable.setExpanded).toHaveBeenCalledWith(true);
-		expect(requestRender).toHaveBeenCalledWith(false, { allowUnknownViewportMutation: true });
+		// resetDisplay() is the only path that retires the transcript's frozen
+		// block snapshots and re-emits the whole transcript at its new heights.
+		// A plain requestRender would replay the stale (collapsed) snapshots.
+		expect(resetDisplay).toHaveBeenCalledTimes(1);
+		expect(requestRender).not.toHaveBeenCalled();
 	});
 });
diff --git a/packages/coding-agent/test/modes/controllers/tan-command-controller.test.ts b/packages/coding-agent/test/modes/controllers/tan-command-controller.test.ts
new file mode 100644
index 000000000..ee41a1887
--- /dev/null
+++ b/packages/coding-agent/test/modes/controllers/tan-command-controller.test.ts
@@ -0,0 +1,219 @@
+import { afterEach, describe, expect, it, vi } from "bun:test";
+import * as path from "node:path";
+import type { AssistantMessage, Model } from "@oh-my-pi/pi-ai";
+import type { AsyncJobRegisterOptions } from "@oh-my-pi/pi-coding-agent/async/job-manager";
+import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import { TanCommandController } from "@oh-my-pi/pi-coding-agent/modes/controllers/tan-command-controller";
+import type { InteractiveModeContext } from "@oh-my-pi/pi-coding-agent/modes/types";
+import { MAIN_AGENT_ID } from "@oh-my-pi/pi-coding-agent/registry/agent-registry";
+import type { CreateAgentSessionResult } from "@oh-my-pi/pi-coding-agent/sdk";
+import * as sdkModule from "@oh-my-pi/pi-coding-agent/sdk";
+import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
+import { TempDir } from "@oh-my-pi/pi-utils";
+
+interface CapturedJobRunContext {
+	jobId: string;
+	signal: AbortSignal;
+	reportProgress: (text: string, details?: Record<string, unknown>) => Promise<void>;
+}
+
+type CapturedJobRun = (ctx: CapturedJobRunContext) => Promise<string>;
+
+const model = { provider: "anthropic", id: "claude-sonnet-4-5" } as Model;
+
+function assistantText(text: string): AssistantMessage {
+	return {
+		role: "assistant",
+		content: [{ type: "text", text }],
+		api: "anthropic-messages",
+		provider: model.provider,
+		model: model.id,
+		usage: {
+			input: 0,
+			output: 0,
+			cacheRead: 0,
+			cacheWrite: 0,
+			totalTokens: 0,
+			cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+		},
+		stopReason: "stop",
+		timestamp: 0,
+	};
+}
+
+function createContext(overrides?: {
+	isStreaming?: boolean;
+	model?: Model;
+	agentId?: string;
+	register?: (run: CapturedJobRun, options?: AsyncJobRegisterOptions) => string;
+}) {
+	const tempDir = TempDir.createSync("@omp-tan-controller-");
+	const parentFile = path.join(tempDir.path(), "parent.jsonl");
+	// The clone nests inside the parent's artifact directory, like a subagent.
+	const cloneFile = path.join(parentFile.slice(0, -6), "clone.jsonl");
+	let capturedRun: CapturedJobRun | undefined;
+	let capturedOptions: AsyncJobRegisterOptions | undefined;
+	const sequence: string[] = [];
+	const register = vi.fn(
+		(_type: "bash" | "task", _label: string, run: CapturedJobRun, options?: AsyncJobRegisterOptions): string => {
+			sequence.push("register");
+			capturedRun = run;
+			capturedOptions = options;
+			return overrides?.register ? overrides.register(run, options) : "job-123";
+		},
+	);
+	const session = {
+		isStreaming: overrides?.isStreaming ?? false,
+		model: overrides?.model ?? model,
+		asyncJobManager: { register },
+		sessionId: "parent-session",
+		configuredThinkingLevel: vi.fn(() => undefined),
+		systemPrompt: ["system prompt"],
+		getActiveToolNames: vi.fn(() => ["read", "bash"]),
+		modelRegistry: { authStorage: { marker: "auth" } },
+		getAgentId: vi.fn(() => overrides?.agentId),
+		sendCustomMessage: vi.fn(async () => {
+			sequence.push("sendCustomMessage");
+		}),
+	} as unknown as InteractiveModeContext["session"];
+	const sessionManager = {
+		getSessionFile: vi.fn(() => parentFile),
+		getCwd: vi.fn(() => tempDir.path()),
+		getSessionDir: vi.fn(() => tempDir.path()),
+		ensureOnDisk: vi.fn(async () => {}),
+		flush: vi.fn(async () => {}),
+	} as unknown as InteractiveModeContext["sessionManager"];
+	const cloneManager = {
+		getSessionFile: vi.fn(() => cloneFile),
+	} as unknown as SessionManager;
+	const ctx = {
+		session,
+		sessionManager,
+		settings: Settings.isolated({ "task.enableLsp": true }),
+		showStatus: vi.fn(),
+		showWarning: vi.fn(),
+		showError: vi.fn(),
+		rebuildChatFromMessages: vi.fn(),
+	} as unknown as InteractiveModeContext;
+	return {
+		tempDir,
+		parentFile,
+		cloneFile,
+		cloneManager,
+		ctx,
+		register,
+		sequence,
+		get capturedRun() {
+			return capturedRun;
+		},
+		get capturedOptions() {
+			return capturedOptions;
+		},
+	};
+}
+
+describe("TanCommandController", () => {
+	afterEach(() => {
+		vi.restoreAllMocks();
+	});
+
+	it("rejects empty work before forking", async () => {
+		const harness = createContext();
+		const forkSpy = vi.spyOn(SessionManager, "forkFrom").mockResolvedValue(harness.cloneManager);
+		const controller = new TanCommandController(harness.ctx);
+
+		await controller.start("   ");
+
+		expect(forkSpy).not.toHaveBeenCalled();
+		expect(harness.ctx.showStatus).toHaveBeenCalledWith("Usage: /tan <work>");
+	});
+
+	it("rejects while the parent session is streaming", async () => {
+		const harness = createContext({ isStreaming: true });
+		const forkSpy = vi.spyOn(SessionManager, "forkFrom").mockResolvedValue(harness.cloneManager);
+		const controller = new TanCommandController(harness.ctx);
+
+		await controller.start("check something");
+
+		expect(forkSpy).not.toHaveBeenCalled();
+		expect(harness.ctx.showWarning).toHaveBeenCalled();
+	});
+
+	it("forks with breadcrumb suppression, registers under Main, and dispatches after receiving the job id", async () => {
+		const harness = createContext();
+		const forkSpy = vi.spyOn(SessionManager, "forkFrom").mockResolvedValue(harness.cloneManager);
+		const controller = new TanCommandController(harness.ctx);
+
+		await controller.start("write the release note");
+
+		expect(forkSpy).toHaveBeenCalledWith(
+			harness.parentFile,
+			harness.tempDir.path(),
+			harness.parentFile.slice(0, -6),
+			undefined,
+			{ suppressBreadcrumb: true },
+		);
+		expect(harness.register).toHaveBeenCalledWith("task", "/tan write the release note", expect.any(Function), {
+			ownerId: MAIN_AGENT_ID,
+		});
+		expect(harness.capturedOptions?.ownerId).toBe(MAIN_AGENT_ID);
+		expect(harness.sequence).toEqual(["register", "sendCustomMessage"]);
+		expect(harness.ctx.session.sendCustomMessage).toHaveBeenCalledWith(
+			expect.objectContaining({
+				customType: "background-tan-dispatch",
+				details: { jobId: "job-123", work: "write the release note", sessionFile: harness.cloneFile },
+			}),
+			{ triggerTurn: false },
+		);
+		expect(harness.ctx.rebuildChatFromMessages).toHaveBeenCalled();
+		expect(harness.ctx.showStatus).toHaveBeenCalledWith("Dispatched background tan job-123");
+	});
+
+	it("aborts the cloned agent when the background job signal aborts", async () => {
+		const harness = createContext({ agentId: MAIN_AGENT_ID });
+		vi.spyOn(SessionManager, "forkFrom").mockResolvedValue(harness.cloneManager);
+		const promptStarted = Promise.withResolvers<void>();
+		const abortObserved = Promise.withResolvers<void>();
+		const clone = {
+			prompt: vi.fn(async () => {
+				promptStarted.resolve();
+				await abortObserved.promise;
+			}),
+			waitForIdle: vi.fn(async () => {}),
+			getLastAssistantMessage: vi.fn(() => assistantText("finished")),
+			abort: vi.fn(() => {
+				abortObserved.resolve();
+			}),
+			dispose: vi.fn(async () => {}),
+		};
+		const createAgentSessionSpy = vi
+			.spyOn(sdkModule, "createAgentSession")
+			.mockResolvedValue({ session: clone } as unknown as CreateAgentSessionResult);
+		const controller = new TanCommandController(harness.ctx);
+		await controller.start("follow the tangent");
+		const capturedRun = harness.capturedRun;
+		expect(capturedRun).toBeDefined();
+		if (!capturedRun) throw new Error("run function was not captured");
+		const abortController = new AbortController();
+
+		const resultPromise = capturedRun({
+			jobId: "job-123",
+			signal: abortController.signal,
+			reportProgress: async () => {},
+		});
+		await promptStarted.promise;
+		abortController.abort();
+		const result = await resultPromise;
+
+		expect(result).toBe("finished");
+		expect(clone.abort).toHaveBeenCalled();
+		expect(clone.dispose).toHaveBeenCalled();
+		expect(createAgentSessionSpy.mock.calls[0]?.[0]).toEqual(
+			expect.objectContaining({
+				providerPromptCacheKey: "parent-session",
+				parentTaskPrefix: expect.stringMatching(/^Tan-/) as unknown as string,
+				agentDisplayName: "tan",
+			}),
+		);
+	});
+});
diff --git a/packages/coding-agent/test/modes/utils/copy-targets.test.ts b/packages/coding-agent/test/modes/utils/copy-targets.test.ts
new file mode 100644
index 000000000..d041ce4df
--- /dev/null
+++ b/packages/coding-agent/test/modes/utils/copy-targets.test.ts
@@ -0,0 +1,207 @@
+import { describe, expect, it } from "bun:test";
+import type { AgentMessage } from "@oh-my-pi/pi-agent-core";
+import {
+	buildCopyTargets,
+	type CopySource,
+	type CopyTarget,
+	extractCodeBlocks,
+	extractLastCommand,
+	extractQuoteBlocks,
+} from "@oh-my-pi/pi-coding-agent/modes/utils/copy-targets";
+
+function source(overrides: Partial<CopySource>): CopySource {
+	return {
+		messages: [],
+		getLastVisibleHandoffText: () => undefined,
+		...overrides,
+	};
+}
+
+function byId(targets: CopyTarget[], id: string): CopyTarget | undefined {
+	return targets.find(t => t.id === id);
+}
+
+function assistantText(text: string): AgentMessage {
+	return { role: "assistant", content: [{ type: "text", text }] } as unknown as AgentMessage;
+}
+
+function assistantCalls(toolCalls: Array<{ name: string; arguments: Record<string, unknown> }>): AgentMessage {
+	return {
+		role: "assistant",
+		content: toolCalls.map((tc, i) => ({ type: "toolCall", id: `tc-${i}`, name: tc.name, arguments: tc.arguments })),
+	} as unknown as AgentMessage;
+}
+
+describe("extractCodeBlocks", () => {
+	it("captures the language id and strips the trailing newline", () => {
+		expect(extractCodeBlocks("intro\n```ts\nconst x = 1;\n```\ntail")).toEqual([
+			{ lang: "ts", code: "const x = 1;" },
+		]);
+	});
+
+	it("returns blocks in document order with empty lang for bare fences", () => {
+		const blocks = extractCodeBlocks("```\nplain\n```\n\n```py\nprint(1)\n```");
+		expect(blocks.map(b => b.lang)).toEqual(["", "py"]);
+		expect(blocks.map(b => b.code)).toEqual(["plain", "print(1)"]);
+	});
+});
+
+describe("extractQuoteBlocks", () => {
+	it("collects a `>`-prefixed run and strips the marker plus one space", () => {
+		const text = "intro\n> line one\n> line two\ntail";
+		expect(extractQuoteBlocks(text)).toEqual([{ text: "line one\nline two" }]);
+	});
+
+	it("keeps bare `>` separator lines as blank lines and splits on plain text", () => {
+		const text = "> first\n>\n> second\n\nbreak\n> later";
+		expect(extractQuoteBlocks(text).map(b => b.text)).toEqual(["first\n\nsecond", "later"]);
+	});
+
+	it("does not treat `>` lines inside a fenced code block as a quote", () => {
+		const text = "> real quote\n```\n> not a quote\n```";
+		expect(extractQuoteBlocks(text)).toEqual([{ text: "real quote" }]);
+	});
+});
+
+describe("extractLastCommand", () => {
+	it("returns the most recent bash command, walking backwards", () => {
+		const messages = [
+			assistantCalls([{ name: "bash", arguments: { command: "echo old" } }]),
+			assistantCalls([{ name: "read", arguments: { path: "x" } }]),
+			assistantCalls([
+				{ name: "bash", arguments: { command: "echo a" } },
+				{ name: "bash", arguments: { command: "echo b" } },
+			]),
+		] as unknown as AgentMessage[];
+		expect(extractLastCommand(messages)).toEqual({ kind: "bash", code: "echo b", language: "bash" });
+	});
+
+	it("joins eval cell code and reports the cell language", () => {
+		const py = [
+			assistantCalls([
+				{ name: "eval", arguments: { cells: [{ language: "py", code: "print(1)" }, { code: "print(2)" }] } },
+			]),
+		] as unknown as AgentMessage[];
+		expect(extractLastCommand(py)).toEqual({ kind: "eval", code: "print(1)\n\nprint(2)", language: "python" });
+
+		const js = [
+			assistantCalls([{ name: "eval", arguments: { cells: [{ language: "js", code: "log(1)" }] } }]),
+		] as unknown as AgentMessage[];
+		expect(extractLastCommand(js)?.language).toBe("javascript");
+	});
+});
+
+describe("buildCopyTargets", () => {
+	it("lists assistant messages most-recent-first, drilling code-bearing ones", () => {
+		const newer = "Newer message\n```ts\nconst a = 1;\n```\nand\n```py\nprint(2)\n```";
+		const targets = buildCopyTargets(
+			source({
+				messages: [assistantText("Older message"), assistantText(newer)] as unknown as AgentMessage[],
+			}),
+		);
+
+		// Newest first.
+		expect(targets[0]?.id).toBe("msg:1");
+		expect(targets[0]?.label).toBe("Newer message");
+		expect(targets[1]?.id).toBe("msg:2");
+
+		// The newer message is itself a copy target (full text) AND a tree node
+		// exposing each code block as a child copy target.
+		const group = targets[0]!;
+		expect(group.content).toBe(newer);
+		expect(group.children?.map(c => c.label)).toEqual(["Block 1", "Block 2", "All 2 blocks"]);
+		expect(group.children?.[0]?.content).toBe("const a = 1;");
+		expect(group.children?.[0]?.language).toBe("ts"); // drives preview syntax highlighting
+		expect(group.children?.at(-1)?.content).toBe("const a = 1;\n\nprint(2)");
+
+		// The older, code-free message is a leaf that copies its full text.
+		expect(targets[1]?.children).toBeUndefined();
+		expect(targets[1]?.content).toBe("Older message");
+	});
+
+	it("exposes a single-block message as content plus one block child (no 'all')", () => {
+		const targets = buildCopyTargets(
+			source({ messages: [assistantText("Just one\n```js\nfoo();\n```")] as unknown as AgentMessage[] }),
+		);
+		const msg = byId(targets, "msg:1");
+		expect(msg?.content).toBe("Just one\n```js\nfoo();\n```");
+		expect(msg?.children?.map(c => c.label)).toEqual(["Block 1"]);
+	});
+
+	it("drills a quoted message into a de-prefixed quote child", () => {
+		const text = "Copy-paste to the other agent:\n\n> relay this\n> across agents";
+		const targets = buildCopyTargets(source({ messages: [assistantText(text)] as unknown as AgentMessage[] }));
+		const msg = byId(targets, "msg:1");
+		// The message node still copies the full markdown (with markers).
+		expect(msg?.content).toBe(text);
+		expect(msg?.hint).toBe("4 lines · 1 quote");
+		const quote = msg?.children?.find(c => c.id === "msg:1:quote:0");
+		expect(quote?.label).toBe("Quote 1");
+		// The drilled child copies the un-prefixed quote, ready to paste onward.
+		expect(quote?.content).toBe("relay this\nacross agents");
+		expect(quote?.language).toBeUndefined();
+		expect(quote?.copyMessage).toBe("Copied quote block 1 to clipboard");
+	});
+
+	it("interleaves code and quote children in document order with combined nodes", () => {
+		const text = "intro\n```ts\na;\n```\n> q one\n```py\nb\n```\n> q two";
+		const targets = buildCopyTargets(source({ messages: [assistantText(text)] as unknown as AgentMessage[] }));
+		const msg = byId(targets, "msg:1");
+		expect(msg?.children?.map(c => c.id)).toEqual([
+			"msg:1:code:0",
+			"msg:1:quote:0",
+			"msg:1:code:1",
+			"msg:1:quote:1",
+			"msg:1:all",
+			"msg:1:all-quotes",
+		]);
+		expect(msg?.hint).toBe("9 lines · 2 code · 2 quote");
+		expect(msg?.children?.find(c => c.id === "msg:1:all-quotes")?.content).toBe("q one\n\nq two");
+	});
+
+	it("skips tool-only assistant turns and non-assistant messages", () => {
+		const messages = [
+			{ role: "user", content: [{ type: "text", text: "hi" }] },
+			assistantCalls([{ name: "read", arguments: { path: "x" } }]),
+			assistantText("real answer"),
+		] as unknown as AgentMessage[];
+		const targets = buildCopyTargets(source({ messages }));
+		expect(targets.filter(t => t.id.startsWith("msg:")).map(t => t.label)).toEqual(["real answer"]);
+	});
+
+	it("falls back to handoff context only when there are no assistant messages", () => {
+		const withMessages = buildCopyTargets(
+			source({
+				messages: [assistantText("answer")] as unknown as AgentMessage[],
+				getLastVisibleHandoffText: () => "<handoff>",
+			}),
+		);
+		expect(byId(withMessages, "handoff")).toBeUndefined();
+
+		const fresh = buildCopyTargets(source({ getLastVisibleHandoffText: () => "<handoff>\nGoal" }));
+		expect(byId(fresh, "handoff")?.content).toBe("<handoff>\nGoal");
+		expect(byId(fresh, "handoff")?.copyMessage).toBe("Copied handoff context to clipboard");
+	});
+
+	it("interleaves runnable commands after the assistant message that issued them", () => {
+		const targets = buildCopyTargets(
+			source({
+				messages: [
+					assistantText("older answer"),
+					assistantCalls([{ name: "bash", arguments: { command: "echo old" } }]),
+					assistantText("newer answer"),
+					assistantCalls([{ name: "bash", arguments: { command: "bun check" } }]),
+				] as unknown as AgentMessage[],
+			}),
+		);
+
+		expect(targets.map(t => t.id)).toEqual(["msg:1", "cmd:1", "msg:2", "cmd:2"]);
+
+		const cmd = byId(targets, "cmd:1");
+		expect(cmd?.label).toBe("bun check");
+		expect(cmd?.hint).toBe("bash · 1 line");
+		expect(cmd?.content).toBe("bun check");
+		expect(cmd?.language).toBe("bash");
+		expect(byId(targets, "cmd:2")?.content).toBe("echo old");
+	});
+});
diff --git a/packages/coding-agent/test/modes/utils/render-initial-messages-dedupe.test.ts b/packages/coding-agent/test/modes/utils/render-initial-messages-dedupe.test.ts
index be1d43af3..52de0a13e 100644
--- a/packages/coding-agent/test/modes/utils/render-initial-messages-dedupe.test.ts
+++ b/packages/coding-agent/test/modes/utils/render-initial-messages-dedupe.test.ts
@@ -65,6 +65,7 @@ function makeCtx(sessionManager?: Pick<SessionManager, "buildSessionContext" | "
 		renderSessionContext: renderSessionContextSpy,
 		showStatus: vi.fn(),
 		ui: { requestRender: vi.fn() },
+		resetTranscript: () => ctx.chatContainer.clear(),
 	} as unknown as InteractiveModeContext;
 
 	return { ctx, buildSessionContextSpy, renderSessionContextSpy };
diff --git a/packages/coding-agent/test/otel-export-probe.ts b/packages/coding-agent/test/otel-export-probe.ts
index 2ada60550..2b285bcf8 100644
--- a/packages/coding-agent/test/otel-export-probe.ts
+++ b/packages/coding-agent/test/otel-export-probe.ts
@@ -35,7 +35,7 @@ const server = Bun.serve({
 process.env.OTEL_EXPORTER_OTLP_TRACES_ENDPOINT = `http://localhost:${server.port}/v1/traces`;
 process.env.OTEL_SERVICE_NAME = "oh-my-pi-export-probe";
 
-initTelemetryExport();
+await initTelemetryExport();
 if (!isTelemetryExportEnabled()) {
 	console.error("PROBE: provider did not register");
 	await server.stop(true);
diff --git a/packages/coding-agent/test/plan-mode-thinking-level.test.ts b/packages/coding-agent/test/plan-mode-thinking-level.test.ts
index b5eee3bf3..548593336 100644
--- a/packages/coding-agent/test/plan-mode-thinking-level.test.ts
+++ b/packages/coding-agent/test/plan-mode-thinking-level.test.ts
@@ -6,7 +6,7 @@
  * calls resolveModelRoleValue() but only returns .model, dropping the thinking level.
  * #applyPlanModeModel() therefore has no thinking level to apply.
  */
-import { afterEach, beforeEach, describe, expect, it } from "bun:test";
+import { afterAll, afterEach, beforeAll, describe, expect, it } from "bun:test";
 import * as path from "node:path";
 import { Agent, ThinkingLevel } from "@oh-my-pi/pi-agent-core";
 import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
@@ -22,7 +22,7 @@ describe("plan mode thinking level", () => {
 	let modelRegistry: ModelRegistry;
 	let authStorage: AuthStorage;
 
-	beforeEach(async () => {
+	beforeAll(async () => {
 		tempDir = TempDir.createSync("@pi-plan-thinking-");
 		authStorage = await AuthStorage.create(path.join(tempDir.path(), "testauth.db"));
 		authStorage.setRuntimeApiKey("anthropic", "test-key");
@@ -33,6 +33,9 @@ describe("plan mode thinking level", () => {
 		if (session) {
 			await session.dispose();
 		}
+	});
+
+	afterAll(() => {
 		authStorage.close();
 		tempDir.removeSync();
 	});
diff --git a/packages/coding-agent/test/plan-mode/approved-plan.test.ts b/packages/coding-agent/test/plan-mode/approved-plan.test.ts
index 6b1d1d98b..1c454d292 100644
--- a/packages/coding-agent/test/plan-mode/approved-plan.test.ts
+++ b/packages/coding-agent/test/plan-mode/approved-plan.test.ts
@@ -1,53 +1,71 @@
-import { afterEach, beforeEach, describe, expect, it } from "bun:test";
-import * as fs from "node:fs/promises";
-import * as os from "node:os";
-import * as path from "node:path";
+import { describe, expect, it } from "bun:test";
 import {
 	humanizePlanTitle,
 	normalizePlanTitle,
-	renameApprovedPlanFile,
+	planFileUrlForSlug,
+	resolveApprovedPlan,
 	resolvePlanTitle,
 } from "@oh-my-pi/pi-coding-agent/plan-mode/approved-plan";
 
-describe("renameApprovedPlanFile", () => {
-	let tmpDir: string;
-	let artifactsDir: string;
-
-	beforeEach(async () => {
-		tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), "approved-plan-"));
-		artifactsDir = path.join(tmpDir, "artifacts");
-		await fs.mkdir(path.join(artifactsDir, "local"), { recursive: true });
+describe("planFileUrlForSlug", () => {
+	it("maps a slug to its local plan URL", () => {
+		expect(planFileUrlForSlug("auth-refactor")).toBe("local://auth-refactor-plan.md");
 	});
+});
 
-	afterEach(async () => {
-		await fs.rm(tmpDir, { recursive: true, force: true });
-	});
-
-	function options(planFilePath: string, finalPlanFilePath: string) {
-		return {
-			planFilePath,
-			finalPlanFilePath,
-			getArtifactsDir: () => artifactsDir,
-			getSessionId: () => "session-z",
-		};
+describe("resolveApprovedPlan", () => {
+	/** A `readPlan` backed by an in-memory map of `local://` URL → content. */
+	function reader(files: Record<string, string>) {
+		return async (url: string) => (url in files ? files[url] : null);
 	}
 
-	it("fails with actionable error when destination already exists", async () => {
-		await Bun.write(path.join(artifactsDir, "local", "PLAN.md"), "draft");
-		await Bun.write(path.join(artifactsDir, "local", "WP_MIGRATION_PLAN.md"), "existing");
-
-		await expect(renameApprovedPlanFile(options("local://PLAN.md", "local://WP_MIGRATION_PLAN.md"))).rejects.toThrow(
-			"Plan destination already exists at local://WP_MIGRATION_PLAN.md",
-		);
+	it("locates the plan from the supplied title's slug — no rename", async () => {
+		const result = await resolveApprovedPlan({
+			suppliedTitle: "auth-refactor",
+			statePlanFilePath: "local://PLAN.md",
+			readPlan: reader({ "local://auth-refactor-plan.md": "# Auth refactor\n\nbody" }),
+		});
+		expect(result.planFilePath).toBe("local://auth-refactor-plan.md");
+		expect(result.planContent).toContain("body");
+		expect(result.title).toBe("auth-refactor");
 	});
 
-	it("renames PLAN.md to titled artifact path", async () => {
-		await Bun.write(path.join(artifactsDir, "local", "PLAN.md"), "draft body");
+	it("strips a trailing -plan from the supplied title before reconstructing the file", async () => {
+		const result = await resolveApprovedPlan({
+			suppliedTitle: "auth-plan",
+			statePlanFilePath: "local://PLAN.md",
+			readPlan: reader({ "local://auth-plan.md": "# Auth\n\nbody" }),
+		});
+		expect(result.planFilePath).toBe("local://auth-plan.md");
+	});
 
-		await renameApprovedPlanFile(options("local://PLAN.md", "local://WP_MIGRATION_PLAN.md"));
+	it("falls back to the plan-mode state path when the slug file is absent", async () => {
+		const result = await resolveApprovedPlan({
+			suppliedTitle: "mismatch",
+			statePlanFilePath: "local://existing-plan.md",
+			readPlan: reader({ "local://existing-plan.md": "# Existing\n\nbody" }),
+		});
+		expect(result.planFilePath).toBe("local://existing-plan.md");
+	});
 
-		expect(await Bun.file(path.join(artifactsDir, "local", "WP_MIGRATION_PLAN.md")).text()).toBe("draft body");
-		await expect(fs.stat(path.join(artifactsDir, "local", "PLAN.md"))).rejects.toThrow();
+	it("scans listed plan files when the title was dropped and state path is empty", async () => {
+		const result = await resolveApprovedPlan({
+			suppliedTitle: undefined,
+			statePlanFilePath: "local://PLAN.md",
+			readPlan: reader({ "local://discovered-plan.md": "# Discovered\n\nbody" }),
+			listPlanFiles: async () => ["local://discovered-plan.md"],
+		});
+		expect(result.planFilePath).toBe("local://discovered-plan.md");
+	});
+
+	it("throws an actionable error when no plan file exists", async () => {
+		await expect(
+			resolveApprovedPlan({
+				suppliedTitle: "ghost",
+				statePlanFilePath: "local://PLAN.md",
+				readPlan: reader({}),
+			}),
+		).rejects.toThrow("Plan file not found at local://ghost-plan.md");
 	});
 });
 
diff --git a/packages/coding-agent/test/read-column-truncation-snapshot.test.ts b/packages/coding-agent/test/read-column-truncation-snapshot.test.ts
index a01c8113d..2c4c037a8 100644
--- a/packages/coding-agent/test/read-column-truncation-snapshot.test.ts
+++ b/packages/coding-agent/test/read-column-truncation-snapshot.test.ts
@@ -23,7 +23,7 @@ import type { ToolSession } from "@oh-my-pi/pi-coding-agent/tools";
 import type { ReadToolDetails } from "@oh-my-pi/pi-coding-agent/tools/read";
 import { ReadTool } from "@oh-my-pi/pi-coding-agent/tools/read";
 
-const HASHLINE_HEADER_LINE = /^¶(\S+)#([0-9A-F]{4})$/m;
+const HASHLINE_HEADER_LINE = /^\[([^#\r\n]+)#([0-9A-F]{4})\]$/m;
 const COLUMN_CAP = 64;
 const LONG_LINE_LEN = COLUMN_CAP * 3;
 
diff --git a/packages/coding-agent/test/read-tool-group.test.ts b/packages/coding-agent/test/read-tool-group.test.ts
index 38884ccee..2ae827823 100644
--- a/packages/coding-agent/test/read-tool-group.test.ts
+++ b/packages/coding-agent/test/read-tool-group.test.ts
@@ -32,6 +32,42 @@ describe("ReadToolGroupComponent", () => {
 		expect(rendered.toLowerCase()).not.toContain("ctrl+o");
 	});
 
+	it("uses the read-specific success mark for completed reads", () => {
+		const component = new ReadToolGroupComponent();
+		component.updateArgs({ path: "/tmp/example.ts" }, "read-success");
+		component.updateResult(
+			{
+				content: [{ type: "text", text: "line 1" }],
+			},
+			false,
+			"read-success",
+		);
+
+		const rendered = component.render(120).join("\n");
+		const plain = Bun.stripANSI(rendered);
+
+		expect(plain).toContain(themeModule.theme.status.enabled);
+		expect(plain).not.toContain(themeModule.theme.status.success);
+		expect(rendered).toContain(themeModule.theme.fg("text", themeModule.theme.status.enabled));
+		expect(rendered).not.toContain(themeModule.theme.fg("success", themeModule.theme.status.enabled));
+	});
+
+	it("omits duplicate success marks from multi-read child rows", () => {
+		const component = new ReadToolGroupComponent();
+		component.updateArgs({ path: "/tmp/one.ts" }, "read-one");
+		component.updateArgs({ path: "/tmp/two.ts" }, "read-two");
+		component.updateResult({ content: [{ type: "text", text: "one" }] }, false, "read-one");
+		component.updateResult({ content: [{ type: "text", text: "two" }] }, false, "read-two");
+
+		const plain = Bun.stripANSI(component.render(120).join("\n"));
+
+		expect(plain).toContain("Read (2)");
+		expect(plain).toContain(`${themeModule.theme.tree.branch} /tmp/one.ts`);
+		expect(plain).toContain(`${themeModule.theme.tree.last} /tmp/two.ts`);
+		expect(plain).not.toContain(`${themeModule.theme.tree.branch} ${themeModule.theme.status.enabled}`);
+		expect(plain).not.toContain(`${themeModule.theme.tree.last} ${themeModule.theme.status.enabled}`);
+	});
+
 	it("renders warning previews with warning styling instead of success styling", () => {
 		const component = new ReadToolGroupComponent({ showContentPreview: true });
 		component.updateArgs({ path: "/tmp/example.ts" }, "read-1");
diff --git a/packages/coding-agent/test/role-thinking-helper-propagation.test.ts b/packages/coding-agent/test/role-thinking-helper-propagation.test.ts
index 37e4b7b54..f094340bc 100644
--- a/packages/coding-agent/test/role-thinking-helper-propagation.test.ts
+++ b/packages/coding-agent/test/role-thinking-helper-propagation.test.ts
@@ -43,6 +43,7 @@ describe("role thinking helper propagation", () => {
 		const registry = {
 			getAvailable: () => [model],
 			getApiKey: async () => "test-key",
+			resolver: vi.fn(() => async () => "test-key"),
 		};
 		const completeSimpleMock = vi.spyOn(ai, "completeSimple").mockResolvedValue({
 			stopReason: "end_turn",
@@ -66,6 +67,7 @@ describe("role thinking helper propagation", () => {
 		const registry = {
 			getAvailable: () => [model],
 			getApiKey: async () => "test-key",
+			resolver: vi.fn(() => async () => "test-key"),
 		};
 		const completeSimpleMock = vi.spyOn(ai, "completeSimple").mockResolvedValue({
 			stopReason: "end_turn",
diff --git a/packages/coding-agent/test/sdk-async-job-manager-singleton.test.ts b/packages/coding-agent/test/sdk-async-job-manager-singleton.test.ts
index 0d4a8029d..903c00722 100644
--- a/packages/coding-agent/test/sdk-async-job-manager-singleton.test.ts
+++ b/packages/coding-agent/test/sdk-async-job-manager-singleton.test.ts
@@ -1,14 +1,35 @@
-import { afterEach, describe, expect, it } from "bun:test";
+import { afterAll, afterEach, beforeAll, describe, expect, it } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
 import { AsyncJobManager } from "@oh-my-pi/pi-coding-agent/async/job-manager";
+import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
 import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
 import { createAgentSession } from "@oh-my-pi/pi-coding-agent/sdk";
+import { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
 import { Snowflake } from "@oh-my-pi/pi-utils";
 
 describe("AsyncJobManager singleton across concurrent top-level sessions", () => {
 	const tempDirs: string[] = [];
+	// Building a ModelRegistry per session is the dominant cost here: createAgentSession
+	// otherwise runs discoverAuthStorage (a fresh AuthStorage DB create+reload) and a
+	// background online model refresh for every spawn (~450ms each). The singleton
+	// ownership behavior under test is independent of model resolution, so we hand every
+	// session one shared, network-free registry built once (~10ms/session instead).
+	let sharedTempDir: string;
+	let sharedAuthStorage: AuthStorage;
+	let sharedModelRegistry: ModelRegistry;
+
+	beforeAll(async () => {
+		sharedTempDir = fs.mkdtempSync(path.join(os.tmpdir(), "pi-sdk-async-singleton-shared-"));
+		sharedAuthStorage = await AuthStorage.create(path.join(sharedTempDir, "auth.db"));
+		sharedModelRegistry = new ModelRegistry(sharedAuthStorage, path.join(sharedTempDir, "models.yml"));
+	});
+
+	afterAll(() => {
+		sharedAuthStorage.close();
+		fs.rmSync(sharedTempDir, { recursive: true, force: true });
+	});
 
 	afterEach(async () => {
 		for (const tempDir of tempDirs.splice(0)) {
@@ -34,6 +55,7 @@ describe("AsyncJobManager singleton across concurrent top-level sessions", () =>
 			slashCommands: [],
 			enableMCP: false,
 			enableLsp: false,
+			modelRegistry: sharedModelRegistry,
 		});
 		return session;
 	}
@@ -65,7 +87,7 @@ describe("AsyncJobManager singleton across concurrent top-level sessions", () =>
 		// Once the owning primary session disposes the singleton clears, matching
 		// the documented single-owner invariant.
 		expect(AsyncJobManager.instance()).toBeUndefined();
-	});
+	}, 60000);
 
 	it("does not cancel the primary session's running jobs when a secondary session disposes", async () => {
 		const primary = await spawnTopLevelSession();
@@ -106,7 +128,7 @@ describe("AsyncJobManager singleton across concurrent top-level sessions", () =>
 		} finally {
 			await primary.dispose();
 		}
-	});
+	}, 60000);
 
 	it("refuses async bash from a secondary session instead of routing it to the primary's manager", async () => {
 		const primary = await spawnTopLevelSession({ "async.enabled": true });
@@ -132,7 +154,7 @@ describe("AsyncJobManager singleton across concurrent top-level sessions", () =>
 		} finally {
 			await primary.dispose();
 		}
-	});
+	}, 60000);
 
 	it("clears a manager installed before a top-level session startup failure takes ownership", async () => {
 		const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), `pi-sdk-async-startup-failure-${Snowflake.next()}-`));
@@ -153,6 +175,7 @@ describe("AsyncJobManager singleton across concurrent top-level sessions", () =>
 				slashCommands: [],
 				enableMCP: false,
 				enableLsp: false,
+				modelRegistry: sharedModelRegistry,
 				systemPrompt: () => {
 					throw new Error("forced startup failure");
 				},
@@ -168,5 +191,5 @@ describe("AsyncJobManager singleton across concurrent top-level sessions", () =>
 		} finally {
 			await replacement.dispose();
 		}
-	});
+	}, 60000);
 });
diff --git a/packages/coding-agent/test/sdk-credential-disabled-bridge.test.ts b/packages/coding-agent/test/sdk-credential-disabled-bridge.test.ts
index 0001a780a..a01522c99 100644
--- a/packages/coding-agent/test/sdk-credential-disabled-bridge.test.ts
+++ b/packages/coding-agent/test/sdk-credential-disabled-bridge.test.ts
@@ -93,10 +93,17 @@ describe("createAgentSession credential_disabled subscription", () => {
 		cwd: dirs.cwd,
 		agentDir: dirs.agentDir,
 		authStorage,
+		// Pin the model registry at a temp models.json. Without an explicit path, ModelRegistry
+		// loads the developer's real ~/.omp models config on every construction (~100ms each,
+		// and non-isolated). Pointing it at the (absent) temp file keeps construction at ~2ms and
+		// avoids leaking host config into the test. Providing the registry also skips the
+		// fire-and-forget background model discovery, which is irrelevant to credential_disabled.
+		modelRegistry: new ModelRegistry(authStorage, path.join(dirs.agentDir, "models.json")),
 		settings: Settings.isolated(),
 		disableExtensionDiscovery: true,
 		extensions,
 		skills: [],
+		rules: [],
 		contextFiles: [],
 		promptTemplates: [],
 		workspaceTree: emptyWorkspaceTree(dirs.cwd),
@@ -406,7 +413,7 @@ describe("createAgentSession credential_disabled subscription", () => {
 				embedderEvents.push(event);
 			},
 		});
-		const modelRegistry = new ModelRegistry(authStorage);
+		const modelRegistry = new ModelRegistry(authStorage, path.join(dirs.agentDir, "models.json"));
 		const ext = makeRecordingExtension();
 
 		const { session } = await createAgentSession({
@@ -448,7 +455,7 @@ describe("createAgentSession credential_disabled subscription", () => {
 		const dirs = makeDirs("mismatch");
 		const registryStorage = await AuthStorage.create(path.join(dirs.agentDir, "agent-registry.db"));
 		const otherStorage = await AuthStorage.create(path.join(dirs.agentDir, "agent-other.db"));
-		const modelRegistry = new ModelRegistry(registryStorage);
+		const modelRegistry = new ModelRegistry(registryStorage, path.join(dirs.agentDir, "models-registry.json"));
 
 		await expect(
 			createAgentSession({
@@ -477,7 +484,7 @@ describe("createAgentSession credential_disabled subscription", () => {
 		// by one microtask so a sync onError() registration lands in time.
 		const dirs = makeDirs("error-routing");
 		const authStorage = await AuthStorage.create(path.join(dirs.agentDir, "agent.db"));
-		const modelRegistry = new ModelRegistry(authStorage);
+		const modelRegistry = new ModelRegistry(authStorage, path.join(dirs.agentDir, "models.json"));
 		try {
 			const throwingExtension: Extension = {
 				path: "test://throwing-credential-disabled",
diff --git a/packages/coding-agent/test/sdk-mcp-discovery.test.ts b/packages/coding-agent/test/sdk-mcp-discovery.test.ts
index 2d5d754c5..ed866c6b2 100644
--- a/packages/coding-agent/test/sdk-mcp-discovery.test.ts
+++ b/packages/coding-agent/test/sdk-mcp-discovery.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, beforeEach, describe, expect, it } from "bun:test";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
@@ -11,6 +11,7 @@ import { createAgentSession } from "@oh-my-pi/pi-coding-agent/sdk";
 import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
 import { Snowflake } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
+import { TOOL_DISCOVERY_AUTO_THRESHOLD } from "../src/tool-discovery/mode";
 
 function createMcpCustomTool(name: string, serverName: string, mcpToolName: string): CustomTool {
 	return {
@@ -46,18 +47,34 @@ const oldSessionMtime = new Date("2000-01-01T00:00:00.000Z");
 
 describe("createAgentSession MCP discovery prompt gating", () => {
 	let tempDir: string;
+	let registryDir: string;
 	let authStorage: AuthStorage;
 	let modelRegistry: ModelRegistry;
 
-	beforeEach(async () => {
-		tempDir = path.join(os.tmpdir(), `pi-sdk-mcp-discovery-${Snowflake.next()}`);
-		fs.mkdirSync(tempDir, { recursive: true });
-		authStorage = await AuthStorage.create(path.join(tempDir, "auth.db"));
+	// Immutable across tests: ModelRegistry's constructor eagerly loads the bundled
+	// model catalog (~120ms). The tests pass models explicitly and never mutate the
+	// registry (refreshInBackground is skipped when modelRegistry is supplied, and
+	// extension source sync is empty under disableExtensionDiscovery), so build it once.
+	beforeAll(async () => {
+		registryDir = path.join(os.tmpdir(), `pi-sdk-mcp-discovery-registry-${Snowflake.next()}`);
+		fs.mkdirSync(registryDir, { recursive: true });
+		authStorage = await AuthStorage.create(path.join(registryDir, "auth.db"));
 		modelRegistry = new ModelRegistry(authStorage);
 	});
 
-	afterEach(() => {
+	afterAll(() => {
 		authStorage.close();
+		if (registryDir && fs.existsSync(registryDir)) {
+			fs.rmSync(registryDir, { recursive: true, force: true });
+		}
+	});
+
+	beforeEach(() => {
+		tempDir = path.join(os.tmpdir(), `pi-sdk-mcp-discovery-${Snowflake.next()}`);
+		fs.mkdirSync(tempDir, { recursive: true });
+	});
+
+	afterEach(() => {
 		if (tempDir && fs.existsSync(tempDir)) {
 			fs.rmSync(tempDir, { recursive: true, force: true });
 		}
@@ -88,6 +105,34 @@ describe("createAgentSession MCP discovery prompt gating", () => {
 		);
 	});
 
+	it("default auto discovery hides MCP tools once the total tool set is too large", async () => {
+		const mcpTools = Array.from({ length: TOOL_DISCOVERY_AUTO_THRESHOLD + 1 }, (_, index) =>
+			createMcpCustomTool(`mcp__auto_tool_${index}`, "auto", `tool_${index}`),
+		);
+		const { session } = await createAgentSession({
+			cwd: tempDir,
+			agentDir: tempDir,
+			modelRegistry,
+			sessionManager: SessionManager.inMemory(),
+			settings: Settings.isolated({}),
+			model: getBundledModel("openai", "gpt-4o-mini"),
+			disableExtensionDiscovery: true,
+			skills: [],
+			contextFiles: [],
+			promptTemplates: [],
+			slashCommands: [],
+			enableMCP: false,
+			enableLsp: false,
+			customTools: mcpTools,
+		});
+
+		const activeNames = session.getActiveToolNames();
+		expect(session.isToolDiscoveryEnabled()).toBe(true);
+		expect(activeNames).toContain("search_tool_bm25");
+		expect(activeNames).not.toContain("mcp__auto_tool_0");
+		expect(session.getDiscoverableTools({ source: "mcp" })).toHaveLength(TOOL_DISCOVERY_AUTO_THRESHOLD + 1);
+	});
+
 	it("advertises discovery guidance for builtin-only tools.discoveryMode all sessions", async () => {
 		const { session } = await createAgentSession({
 			cwd: tempDir,
diff --git a/packages/coding-agent/test/sdk-model-selection.test.ts b/packages/coding-agent/test/sdk-model-selection.test.ts
index c702edf29..89f072394 100644
--- a/packages/coding-agent/test/sdk-model-selection.test.ts
+++ b/packages/coding-agent/test/sdk-model-selection.test.ts
@@ -12,6 +12,7 @@ import { Snowflake } from "@oh-my-pi/pi-utils";
 
 describe("createAgentSession deferred model pattern resolution", () => {
 	let tempDir: string;
+	const authStoragesToClose: AuthStorage[] = [];
 
 	beforeEach(() => {
 		tempDir = path.join(os.tmpdir(), `pi-sdk-model-selection-${Snowflake.next()}`);
@@ -19,6 +20,10 @@ describe("createAgentSession deferred model pattern resolution", () => {
 	});
 
 	afterEach(() => {
+		for (const authStorage of authStoragesToClose) {
+			authStorage.close();
+		}
+		authStoragesToClose.length = 0;
 		if (tempDir && fs.existsSync(tempDir)) {
 			fs.rmSync(tempDir, { recursive: true, force: true });
 		}
@@ -52,10 +57,20 @@ describe("createAgentSession deferred model pattern resolution", () => {
 		});
 	};
 
-	function buildSessionOptions(modelPattern: string) {
+	async function buildSessionOptions(modelPattern: string) {
+		// Pass an explicit ModelRegistry so createAgentSession skips its implicit
+		// ModelRegistry.refreshInBackground() — a network model-discovery pass
+		// (~250ms/session) that contributes nothing here: the model resolves from
+		// the inline extension provider, never from network catalogs. Mirrors the
+		// explicit-registry pattern the resume tests below already rely on.
+		const authStorage = await AuthStorage.create(path.join(tempDir, "auth.db"));
+		authStoragesToClose.push(authStorage);
+		const modelRegistry = new ModelRegistry(authStorage, path.join(tempDir, "models.yml"));
 		return {
 			cwd: tempDir,
 			agentDir: tempDir,
+			authStorage,
+			modelRegistry,
 			sessionManager: SessionManager.inMemory(),
 			disableExtensionDiscovery: true,
 			extensions: [providerExtension],
@@ -71,7 +86,7 @@ describe("createAgentSession deferred model pattern resolution", () => {
 
 	test("resolves explicit modelPattern after extension providers register", async () => {
 		const { session, modelFallbackMessage } = await createAgentSession(
-			buildSessionOptions("runtime-provider/runtime-model"),
+			await buildSessionOptions("runtime-provider/runtime-model"),
 		);
 
 		expect(session.model).toBeDefined();
@@ -82,7 +97,7 @@ describe("createAgentSession deferred model pattern resolution", () => {
 
 	test("does not silently fallback when explicit modelPattern is unresolved", async () => {
 		const { session, modelFallbackMessage } = await createAgentSession(
-			buildSessionOptions("missing-provider/missing-model"),
+			await buildSessionOptions("missing-provider/missing-model"),
 		);
 
 		expect(session.model).toBeUndefined();
@@ -95,7 +110,7 @@ describe("createAgentSession deferred model pattern resolution", () => {
 		settings.setModelRole("default", "pi/smol:high");
 
 		const { session } = await createAgentSession({
-			...buildSessionOptions("runtime-provider/runtime-reasoning-model"),
+			...(await buildSessionOptions("runtime-provider/runtime-reasoning-model")),
 			settings,
 		});
 
diff --git a/packages/coding-agent/test/sdk-session-isolation.test.ts b/packages/coding-agent/test/sdk-session-isolation.test.ts
index 91355299c..7bac57a38 100644
--- a/packages/coding-agent/test/sdk-session-isolation.test.ts
+++ b/packages/coding-agent/test/sdk-session-isolation.test.ts
@@ -1,12 +1,14 @@
-import { afterEach, describe, expect, it } from "bun:test";
+import { afterAll, afterEach, beforeAll, describe, expect, it } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
 import { type AssistantMessage, getBundledModel } from "@oh-my-pi/pi-ai";
 import type { Rule } from "@oh-my-pi/pi-coding-agent/capability/rule";
+import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
 import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
 import { createAgentSession } from "@oh-my-pi/pi-coding-agent/sdk";
 import { SecretObfuscator } from "@oh-my-pi/pi-coding-agent/secrets";
+import { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
 import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
 import { getSessionsDir, Snowflake } from "@oh-my-pi/pi-utils";
 
@@ -55,6 +57,23 @@ function getAssistantText(message: AssistantMessage | undefined): string {
 
 describe("createAgentSession session storage isolation", () => {
 	const tempDirs: string[] = [];
+	// One shared, fully-populated (bundled models load synchronously in the
+	// constructor) registry for every case. Passing it via options skips the
+	// per-call discoverAuthStorage() SQLite open and the refreshInBackground()
+	// network model probe inside createAgentSession — the two real wall-clock
+	// sinks here. None of these cases assert on model discovery, so an
+	// ambient-credential-free in-memory auth store keeps them deterministic.
+	let sharedAuthStorage: AuthStorage;
+	let sharedModelRegistry: ModelRegistry;
+
+	beforeAll(async () => {
+		sharedAuthStorage = await AuthStorage.create(":memory:");
+		sharedModelRegistry = new ModelRegistry(sharedAuthStorage);
+	});
+
+	afterAll(() => {
+		sharedAuthStorage.close();
+	});
 
 	afterEach(async () => {
 		for (const tempDir of tempDirs.splice(0)) {
@@ -72,6 +91,7 @@ describe("createAgentSession session storage isolation", () => {
 		const { session } = await createAgentSession({
 			cwd,
 			agentDir,
+			modelRegistry: sharedModelRegistry,
 			settings: Settings.isolated(),
 			disableExtensionDiscovery: true,
 			skills: [],
@@ -105,6 +125,7 @@ describe("createAgentSession session storage isolation", () => {
 		const { session } = await createAgentSession({
 			cwd,
 			agentDir,
+			modelRegistry: sharedModelRegistry,
 			settings: Settings.isolated(),
 			rules: [rule],
 			disableExtensionDiscovery: true,
@@ -137,6 +158,7 @@ describe("createAgentSession session storage isolation", () => {
 			const commonOptions = {
 				cwd,
 				agentDir,
+				modelRegistry: sharedModelRegistry,
 				settings: Settings.isolated({ "secrets.enabled": true }),
 				disableExtensionDiscovery: true,
 				skills: [],
@@ -206,6 +228,7 @@ describe("createAgentSession session storage isolation", () => {
 			const { session } = await createAgentSession({
 				cwd,
 				agentDir,
+				modelRegistry: sharedModelRegistry,
 				sessionManager: resumedManager,
 				model,
 				settings: Settings.isolated({ "secrets.enabled": true }),
diff --git a/packages/coding-agent/test/sdk-skills.test.ts b/packages/coding-agent/test/sdk-skills.test.ts
index 81ad42029..15bccd54a 100644
--- a/packages/coding-agent/test/sdk-skills.test.ts
+++ b/packages/coding-agent/test/sdk-skills.test.ts
@@ -1,10 +1,12 @@
-import { afterEach, beforeEach, describe, expect, it } from "bun:test";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
+import { ModelRegistry } from "@oh-my-pi/pi-coding-agent/config/model-registry";
 import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
 import type { Skill } from "@oh-my-pi/pi-coding-agent/sdk";
 import { createAgentSession } from "@oh-my-pi/pi-coding-agent/sdk";
+import { AuthStorage } from "@oh-my-pi/pi-coding-agent/session/auth-storage";
 import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
 import { cleanupTempHome } from "./helpers/temp-home-cleanup";
 
@@ -24,6 +26,25 @@ describe("createAgentSession skills option", () => {
 	let skillsDir: string;
 	let tempHomeDir = "";
 	let originalHome: string | undefined;
+	// Auth storage (SQLite DB) and the model registry are immutable across these tests: skill
+	// discovery never touches models, and building them per test would make createAgentSession call
+	// modelRegistry.refreshInBackground(), whose online model discovery saturates the event loop and
+	// serializes the otherwise-parallel capability scans (~340ms/call). Supplying a prebuilt registry
+	// skips that refresh entirely (~24ms/call).
+	let sharedDir: string;
+	let sharedAuthStorage: AuthStorage;
+	let sharedModelRegistry: ModelRegistry;
+
+	beforeAll(async () => {
+		sharedDir = fs.mkdtempSync(path.join(os.tmpdir(), "pi-sdk-skills-shared-"));
+		sharedAuthStorage = await AuthStorage.create(path.join(sharedDir, "auth.db"));
+		sharedModelRegistry = new ModelRegistry(sharedAuthStorage, path.join(sharedDir, "models.yml"));
+	});
+
+	afterAll(() => {
+		sharedAuthStorage.close();
+		fs.rmSync(sharedDir, { recursive: true, force: true });
+	});
 
 	beforeEach(() => {
 		tempDir = path.join(os.tmpdir(), `pi-sdk-test-${Date.now()}-${Math.random().toString(36).slice(2)}`);
@@ -74,6 +95,7 @@ Loaded via symbolic link.
 			cwd: tempDir,
 			agentDir: tempDir,
 			sessionManager: SessionManager.inMemory(),
+			modelRegistry: sharedModelRegistry,
 			settings: createIsolatedSkillsSettings(),
 		});
 
@@ -87,6 +109,7 @@ Loaded via symbolic link.
 			cwd: tempDir,
 			agentDir: tempDir,
 			sessionManager: SessionManager.inMemory(),
+			modelRegistry: sharedModelRegistry,
 			settings: createIsolatedSkillsSettings(),
 		});
 
@@ -102,6 +125,7 @@ Loaded via symbolic link.
 			cwd: tempDir,
 			agentDir: tempDir,
 			sessionManager: SessionManager.inMemory(),
+			modelRegistry: sharedModelRegistry,
 			settings: createIsolatedSkillsSettings(),
 		});
 
@@ -112,6 +136,7 @@ Loaded via symbolic link.
 			cwd: tempDir,
 			agentDir: tempDir,
 			sessionManager: SessionManager.inMemory(),
+			modelRegistry: sharedModelRegistry,
 			skills: [], // Explicitly empty - like --no-skills
 			settings: createIsolatedSkillsSettings(),
 		});
@@ -135,6 +160,7 @@ Loaded via symbolic link.
 			cwd: tempDir,
 			agentDir: tempDir,
 			sessionManager: SessionManager.inMemory(),
+			modelRegistry: sharedModelRegistry,
 			skills: [customSkill],
 			settings: createIsolatedSkillsSettings(),
 		});
diff --git a/packages/coding-agent/test/sdk-tool-activation.test.ts b/packages/coding-agent/test/sdk-tool-activation.test.ts
index 6758acde1..a4b8da987 100644
--- a/packages/coding-agent/test/sdk-tool-activation.test.ts
+++ b/packages/coding-agent/test/sdk-tool-activation.test.ts
@@ -4,7 +4,11 @@ import * as os from "node:os";
 import * as path from "node:path";
 import { getBundledModel } from "@oh-my-pi/pi-ai";
 import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
-import { createAgentSession, type ExtensionFactory } from "@oh-my-pi/pi-coding-agent/sdk";
+import {
+	type CreateAgentSessionOptions,
+	createAgentSession,
+	type ExtensionFactory,
+} from "@oh-my-pi/pi-coding-agent/sdk";
 import { SessionManager } from "@oh-my-pi/pi-coding-agent/session/session-manager";
 import { Snowflake } from "@oh-my-pi/pi-utils";
 import * as z from "zod/v4";
@@ -34,6 +38,35 @@ const toolActivationExtension: ExtensionFactory = pi => {
 describe("createAgentSession defaultInactive tool activation", () => {
 	const tempDirs: string[] = [];
 
+	const makeTempDir = (): string => {
+		const tempDir = path.join(os.tmpdir(), `pi-sdk-tool-activation-${Snowflake.next()}`);
+		tempDirs.push(tempDir);
+		fs.mkdirSync(tempDir, { recursive: true });
+		return tempDir;
+	};
+
+	// Shared options for every session. `rules: []` and `workspaceTree` short-circuit
+	// the two slow startup scans (rule discovery + native workspace walk, ~100ms each)
+	// that are irrelevant to tool activation: these tests assert only which tools are
+	// registered/active and that tool names appear in the system prompt. Each call
+	// returns fresh `settings`/`sessionManager` instances to keep tests isolated.
+	const baseOptions = (tempDir: string): CreateAgentSessionOptions => ({
+		cwd: tempDir,
+		agentDir: tempDir,
+		sessionManager: SessionManager.inMemory(),
+		settings: Settings.isolated(),
+		model: getBundledModel("openai", "gpt-4o-mini"),
+		disableExtensionDiscovery: true,
+		skills: [],
+		contextFiles: [],
+		promptTemplates: [],
+		slashCommands: [],
+		enableMCP: false,
+		enableLsp: false,
+		rules: [],
+		workspaceTree: { rootPath: tempDir, rendered: "", truncated: false, totalLines: 0, agentsMdFiles: [] },
+	});
+
 	afterEach(() => {
 		for (const tempDir of tempDirs.splice(0)) {
 			fs.rmSync(tempDir, { recursive: true, force: true });
@@ -43,24 +76,11 @@ describe("createAgentSession defaultInactive tool activation", () => {
 	});
 
 	it("excludes defaultInactive extension tools from the initial active set unless explicitly requested", async () => {
-		const tempDir = path.join(os.tmpdir(), `pi-sdk-tool-activation-${Snowflake.next()}`);
-		tempDirs.push(tempDir);
-		fs.mkdirSync(tempDir, { recursive: true });
+		const tempDir = makeTempDir();
 
 		const { session } = await createAgentSession({
-			cwd: tempDir,
-			agentDir: tempDir,
-			sessionManager: SessionManager.inMemory(),
-			settings: Settings.isolated(),
-			model: getBundledModel("openai", "gpt-4o-mini"),
-			disableExtensionDiscovery: true,
+			...baseOptions(tempDir),
 			extensions: [toolActivationExtension],
-			skills: [],
-			contextFiles: [],
-			promptTemplates: [],
-			slashCommands: [],
-			enableMCP: false,
-			enableLsp: false,
 		});
 
 		try {
@@ -77,24 +97,11 @@ describe("createAgentSession defaultInactive tool activation", () => {
 	});
 
 	it("allows explicitly requested defaultInactive extension tools into the initial active set", async () => {
-		const tempDir = path.join(os.tmpdir(), `pi-sdk-tool-activation-${Snowflake.next()}`);
-		tempDirs.push(tempDir);
-		fs.mkdirSync(tempDir, { recursive: true });
+		const tempDir = makeTempDir();
 
 		const { session } = await createAgentSession({
-			cwd: tempDir,
-			agentDir: tempDir,
-			sessionManager: SessionManager.inMemory(),
-			settings: Settings.isolated(),
-			model: getBundledModel("openai", "gpt-4o-mini"),
-			disableExtensionDiscovery: true,
+			...baseOptions(tempDir),
 			extensions: [toolActivationExtension],
-			skills: [],
-			contextFiles: [],
-			promptTemplates: [],
-			slashCommands: [],
-			enableMCP: false,
-			enableLsp: false,
 			toolNames: ["read", "default_inactive_tool"],
 		});
 
@@ -113,23 +120,10 @@ describe("createAgentSession defaultInactive tool activation", () => {
 		// (e.g. `["read", "search", "find", "lsp", "web_search"]`). Without this
 		// invariant, `yield` ended up registered but not active, and the model
 		// could not satisfy the idle-reminder contract that demands a `yield` call.
-		const tempDir = path.join(os.tmpdir(), `pi-sdk-tool-activation-${Snowflake.next()}`);
-		tempDirs.push(tempDir);
-		fs.mkdirSync(tempDir, { recursive: true });
+		const tempDir = makeTempDir();
 
 		const { session } = await createAgentSession({
-			cwd: tempDir,
-			agentDir: tempDir,
-			sessionManager: SessionManager.inMemory(),
-			settings: Settings.isolated(),
-			model: getBundledModel("openai", "gpt-4o-mini"),
-			disableExtensionDiscovery: true,
-			skills: [],
-			contextFiles: [],
-			promptTemplates: [],
-			slashCommands: [],
-			enableMCP: false,
-			enableLsp: false,
+			...baseOptions(tempDir),
 			requireYieldTool: true,
 			toolNames: ["read", "search", "find", "web_search"],
 		});
@@ -149,23 +143,10 @@ describe("createAgentSession defaultInactive tool activation", () => {
 		// the registry has no `deferrable` tool, so the previous gate dropped
 		// `resolve` from the registry and plan mode silently activated without
 		// it — leaving the agent stuck after drafting the plan.
-		const tempDir = path.join(os.tmpdir(), `pi-sdk-tool-activation-${Snowflake.next()}`);
-		tempDirs.push(tempDir);
-		fs.mkdirSync(tempDir, { recursive: true });
+		const tempDir = makeTempDir();
 
 		const { session } = await createAgentSession({
-			cwd: tempDir,
-			agentDir: tempDir,
-			sessionManager: SessionManager.inMemory(),
-			settings: Settings.isolated(),
-			model: getBundledModel("openai", "gpt-4o-mini"),
-			disableExtensionDiscovery: true,
-			skills: [],
-			contextFiles: [],
-			promptTemplates: [],
-			slashCommands: [],
-			enableMCP: false,
-			enableLsp: false,
+			...baseOptions(tempDir),
 			toolNames: ["read", "search", "find", "web_search"],
 		});
 
@@ -177,26 +158,14 @@ describe("createAgentSession defaultInactive tool activation", () => {
 	});
 
 	it("drops the hidden resolve tool when neither a deferrable tool nor plan mode can use it", async () => {
-		const tempDir = path.join(os.tmpdir(), `pi-sdk-tool-activation-${Snowflake.next()}`);
-		tempDirs.push(tempDir);
-		fs.mkdirSync(tempDir, { recursive: true });
+		const tempDir = makeTempDir();
 
 		const settings = Settings.isolated();
 		settings.set("plan.enabled", false);
 
 		const { session } = await createAgentSession({
-			cwd: tempDir,
-			agentDir: tempDir,
-			sessionManager: SessionManager.inMemory(),
+			...baseOptions(tempDir),
 			settings,
-			model: getBundledModel("openai", "gpt-4o-mini"),
-			disableExtensionDiscovery: true,
-			skills: [],
-			contextFiles: [],
-			promptTemplates: [],
-			slashCommands: [],
-			enableMCP: false,
-			enableLsp: false,
 			toolNames: ["read", "search", "find", "web_search"],
 		});
 
@@ -208,23 +177,10 @@ describe("createAgentSession defaultInactive tool activation", () => {
 	});
 
 	it("does not register the xAI TTS tool unless enabled", async () => {
-		const tempDir = path.join(os.tmpdir(), `pi-sdk-tool-activation-${Snowflake.next()}`);
-		tempDirs.push(tempDir);
-		fs.mkdirSync(tempDir, { recursive: true });
+		const tempDir = makeTempDir();
 
 		const { session } = await createAgentSession({
-			cwd: tempDir,
-			agentDir: tempDir,
-			sessionManager: SessionManager.inMemory(),
-			settings: Settings.isolated(),
-			model: getBundledModel("openai", "gpt-4o-mini"),
-			disableExtensionDiscovery: true,
-			skills: [],
-			contextFiles: [],
-			promptTemplates: [],
-			slashCommands: [],
-			enableMCP: false,
-			enableLsp: false,
+			...baseOptions(tempDir),
 		});
 
 		try {
@@ -237,23 +193,11 @@ describe("createAgentSession defaultInactive tool activation", () => {
 	});
 
 	it("registers the xAI TTS tool when enabled", async () => {
-		const tempDir = path.join(os.tmpdir(), `pi-sdk-tool-activation-${Snowflake.next()}`);
-		tempDirs.push(tempDir);
-		fs.mkdirSync(tempDir, { recursive: true });
+		const tempDir = makeTempDir();
 
 		const { session } = await createAgentSession({
-			cwd: tempDir,
-			agentDir: tempDir,
-			sessionManager: SessionManager.inMemory(),
+			...baseOptions(tempDir),
 			settings: Settings.isolated({ "tts.enabled": true }),
-			model: getBundledModel("openai", "gpt-4o-mini"),
-			disableExtensionDiscovery: true,
-			skills: [],
-			contextFiles: [],
-			promptTemplates: [],
-			slashCommands: [],
-			enableMCP: false,
-			enableLsp: false,
 		});
 
 		try {
diff --git a/packages/coding-agent/test/session/session-manager-fork.test.ts b/packages/coding-agent/test/session/session-manager-fork.test.ts
new file mode 100644
index 000000000..a7cf5b5c2
--- /dev/null
+++ b/packages/coding-agent/test/session/session-manager-fork.test.ts
@@ -0,0 +1,87 @@
+import { describe, expect, it } from "bun:test";
+import * as fs from "node:fs/promises";
+import * as path from "node:path";
+import {
+	CURRENT_SESSION_VERSION,
+	type SessionHeader,
+	SessionManager,
+} from "@oh-my-pi/pi-coding-agent/session/session-manager";
+import { getTerminalId } from "@oh-my-pi/pi-tui";
+import { getAgentDir, getTerminalSessionsDir, setAgentDir, TempDir } from "@oh-my-pi/pi-utils";
+
+interface JsonlMessageEntry {
+	type: "message";
+	id: string;
+	parentId: string | null;
+	timestamp: string;
+	message: {
+		role: "user";
+		content: string;
+		timestamp: number;
+	};
+}
+
+describe("SessionManager.forkFrom", () => {
+	it("suppresses terminal breadcrumbs while preserving source history under a new parented session", async () => {
+		using tempDir = TempDir.createSync("@omp-session-fork-");
+		const previousAgentDir = getAgentDir();
+		const previousTermSessionId = process.env.TERM_SESSION_ID;
+		setAgentDir(path.join(tempDir.path(), "agent"));
+		process.env.TERM_SESSION_ID = "omp-fork-test";
+		try {
+			const cwd = path.join(tempDir.path(), "project");
+			const sessionDir = path.join(tempDir.path(), "sessions");
+			await fs.mkdir(sessionDir, { recursive: true });
+			const sourceFile = path.join(sessionDir, "source.jsonl");
+			const timestamp = new Date().toISOString();
+			const sourceHeader: SessionHeader = {
+				type: "session",
+				version: CURRENT_SESSION_VERSION,
+				id: "source-session",
+				timestamp,
+				cwd,
+			};
+			const sourceMessage: JsonlMessageEntry = {
+				type: "message",
+				id: "message-1",
+				parentId: null,
+				timestamp,
+				message: { role: "user", content: "hello", timestamp: Date.now() },
+			};
+			const sourceText = `${JSON.stringify(sourceHeader)}\n${JSON.stringify(sourceMessage)}\n`;
+			await Bun.write(sourceFile, sourceText);
+
+			const terminalId = getTerminalId();
+			expect(terminalId).toBeString();
+			const breadcrumbFile = path.join(getTerminalSessionsDir(), terminalId ?? "missing");
+			await fs.rm(breadcrumbFile, { force: true });
+
+			const forked = await SessionManager.forkFrom(sourceFile, cwd, sessionDir, undefined, {
+				suppressBreadcrumb: true,
+			});
+			await Bun.sleep(10);
+			const cloneFile = forked.getSessionFile();
+			expect(cloneFile).toBeString();
+			if (!cloneFile) throw new Error("expected forked session file");
+
+			expect(await Bun.file(sourceFile).text()).toBe(sourceText);
+			expect(await Bun.file(breadcrumbFile).exists()).toBe(false);
+			expect(cloneFile).not.toBe(sourceFile);
+
+			const lines = (await Bun.file(cloneFile).text()).trim().split("\n");
+			const cloneHeader = JSON.parse(lines[0] ?? "{}") as SessionHeader;
+			const cloneMessage = JSON.parse(lines[1] ?? "{}") as JsonlMessageEntry;
+			expect(cloneHeader.id).not.toBe(sourceHeader.id);
+			expect(cloneHeader.parentSession).toBe(sourceHeader.id);
+			expect(cloneHeader.cwd).toBe(cwd);
+			expect(cloneMessage.message.content).toBe("hello");
+		} finally {
+			if (previousTermSessionId === undefined) {
+				delete process.env.TERM_SESSION_ID;
+			} else {
+				process.env.TERM_SESSION_ID = previousTermSessionId;
+			}
+			setAgentDir(previousAgentDir);
+		}
+	});
+});
diff --git a/packages/coding-agent/test/setup-wizard.test.ts b/packages/coding-agent/test/setup-wizard.test.ts
index 7f01b01fe..7be0f62f0 100644
--- a/packages/coding-agent/test/setup-wizard.test.ts
+++ b/packages/coding-agent/test/setup-wizard.test.ts
@@ -49,6 +49,15 @@ describe("setup wizard scene selection", () => {
 		expect(scenes.map(scene => scene.id)).toEqual(ALL_SCENES.map(scene => scene.id));
 	});
 
+	it("keeps CURRENT_SETUP_VERSION in sync with the highest scene minVersion", () => {
+		// main.ts's cold-launch gate sources CURRENT_SETUP_VERSION from the tiny
+		// `setup-version` module to decide whether to load the wizard at all. If a
+		// new scene raises the bar but the constant is not bumped, stale installs
+		// would never see the scene. Guard the invariant the gate relies on.
+		const highestMinVersion = Math.max(...ALL_SCENES.map(scene => scene.minVersion));
+		expect(CURRENT_SETUP_VERSION).toBe(highestMinVersion);
+	});
+
 	it("runs only scenes newer than the stored setup version", async () => {
 		const scenes = [testScene("v1-a", 1), testScene("v1-b", 1), testScene("v2", 2)];
 		const selected = await selectSetupScenes(1, scenes, fakeContextWithConfiguredModel(), { isTTY: true });
diff --git a/packages/coding-agent/test/slash-commands/btw.test.ts b/packages/coding-agent/test/slash-commands/btw.test.ts
index 1638229df..1349fa895 100644
--- a/packages/coding-agent/test/slash-commands/btw.test.ts
+++ b/packages/coding-agent/test/slash-commands/btw.test.ts
@@ -13,7 +13,6 @@ function createRuntime() {
 				editor: { setText } as unknown as InteractiveModeContext["editor"],
 				handleBtwCommand,
 			} as unknown as InteractiveModeContext,
-			handleBackgroundCommand: () => {},
 		},
 	};
 }
diff --git a/packages/coding-agent/test/slash-commands/force.test.ts b/packages/coding-agent/test/slash-commands/force.test.ts
index 28ce73f31..3c47db41b 100644
--- a/packages/coding-agent/test/slash-commands/force.test.ts
+++ b/packages/coding-agent/test/slash-commands/force.test.ts
@@ -20,7 +20,6 @@ function createRuntimeHarness(overrides?: { setForcedToolChoice?: (toolName: str
 	return {
 		runtime: {
 			ctx,
-			handleBackgroundCommand: () => {},
 		},
 		setForcedToolChoice,
 		setText,
diff --git a/packages/coding-agent/test/slash-commands/fresh.test.ts b/packages/coding-agent/test/slash-commands/fresh.test.ts
new file mode 100644
index 000000000..270c8fbd6
--- /dev/null
+++ b/packages/coding-agent/test/slash-commands/fresh.test.ts
@@ -0,0 +1,42 @@
+import { describe, expect, it, vi } from "bun:test";
+import type { InteractiveModeContext } from "@oh-my-pi/pi-coding-agent/modes/types";
+import { executeBuiltinSlashCommand } from "@oh-my-pi/pi-coding-agent/slash-commands/builtin-registry";
+
+function createRuntimeHarness(handleFreshCommand: InteractiveModeContext["handleFreshCommand"]) {
+	const setText = vi.fn();
+	return {
+		setText,
+		handleFreshCommand,
+		runtime: {
+			ctx: {
+				editor: { setText } as unknown as InteractiveModeContext["editor"],
+				handleFreshCommand,
+			} as InteractiveModeContext,
+		},
+	};
+}
+
+describe("/fresh slash command", () => {
+	it("awaits provider-state refresh before resolving", async () => {
+		const deferred = Promise.withResolvers<void>();
+		const handleFreshCommand = vi.fn(() => deferred.promise);
+		const harness = createRuntimeHarness(handleFreshCommand);
+
+		let settled = false;
+		const execution = executeBuiltinSlashCommand("/fresh", harness.runtime).then(result => {
+			settled = true;
+			return result;
+		});
+
+		await Promise.resolve();
+
+		expect(harness.setText).toHaveBeenCalledWith("");
+		expect(handleFreshCommand).toHaveBeenCalledTimes(1);
+		expect(settled).toBe(false);
+
+		deferred.resolve();
+
+		expect(await execution).toBe(true);
+		expect(settled).toBe(true);
+	});
+});
diff --git a/packages/coding-agent/test/slash-commands/login.test.ts b/packages/coding-agent/test/slash-commands/login.test.ts
index 083dc0503..7ad589d72 100644
--- a/packages/coding-agent/test/slash-commands/login.test.ts
+++ b/packages/coding-agent/test/slash-commands/login.test.ts
@@ -4,7 +4,7 @@ import type { InteractiveModeContext } from "@oh-my-pi/pi-coding-agent/modes/typ
 import { executeBuiltinSlashCommand } from "@oh-my-pi/pi-coding-agent/slash-commands/builtin-registry";
 
 type RuntimeHarness = {
-	runtime: { ctx: InteractiveModeContext; handleBackgroundCommand: () => void };
+	runtime: { ctx: InteractiveModeContext };
 	getStatus: () => string | undefined;
 	getWarning: () => string | undefined;
 	getSelectorMode: () => "login" | "logout" | undefined;
@@ -36,7 +36,6 @@ const createRuntimeHarness = (manualInput: OAuthManualInputManager): RuntimeHarn
 	return {
 		runtime: {
 			ctx,
-			handleBackgroundCommand: () => {},
 		},
 		getStatus: () => statusMessage,
 		getWarning: () => warningMessage,
diff --git a/packages/coding-agent/test/slash-commands/omfg.test.ts b/packages/coding-agent/test/slash-commands/omfg.test.ts
index 82cbec560..a95bc41f2 100644
--- a/packages/coding-agent/test/slash-commands/omfg.test.ts
+++ b/packages/coding-agent/test/slash-commands/omfg.test.ts
@@ -13,7 +13,6 @@ function createRuntime() {
 				editor: { setText } as unknown as InteractiveModeContext["editor"],
 				handleOmfgCommand,
 			} as unknown as InteractiveModeContext,
-			handleBackgroundCommand: () => {},
 		},
 	};
 }
diff --git a/packages/coding-agent/test/slash-commands/plan-history.test.ts b/packages/coding-agent/test/slash-commands/plan-history.test.ts
index e4b1dab0c..656be0978 100644
--- a/packages/coding-agent/test/slash-commands/plan-history.test.ts
+++ b/packages/coding-agent/test/slash-commands/plan-history.test.ts
@@ -30,7 +30,7 @@ function createPlanHarness(opts: { planModeEnabled: boolean; confirmExit: boolea
 	} as unknown as InteractiveModeContext;
 
 	return {
-		runtime: { ctx, handleBackgroundCommand: () => {} },
+		runtime: { ctx },
 		state,
 		addToHistory,
 		setText,
@@ -53,7 +53,7 @@ function createGoalHarness(opts: { goalModeEnabled: boolean; dropOnCall: boolean
 	} as unknown as InteractiveModeContext;
 
 	return {
-		runtime: { ctx, handleBackgroundCommand: () => {} },
+		runtime: { ctx },
 		state,
 		addToHistory,
 		setText,
diff --git a/packages/coding-agent/test/slash-commands/retry.test.ts b/packages/coding-agent/test/slash-commands/retry.test.ts
index 899031519..bc7486407 100644
--- a/packages/coding-agent/test/slash-commands/retry.test.ts
+++ b/packages/coding-agent/test/slash-commands/retry.test.ts
@@ -16,7 +16,6 @@ function createRuntime(didRetry: boolean) {
 				editor: { setText } as unknown as InteractiveModeContext["editor"],
 				showStatus,
 			} as unknown as InteractiveModeContext,
-			handleBackgroundCommand: () => {},
 		},
 	};
 }
diff --git a/packages/coding-agent/test/slash-commands/session.test.ts b/packages/coding-agent/test/slash-commands/session.test.ts
index d4662c28a..86c6821d4 100644
--- a/packages/coding-agent/test/slash-commands/session.test.ts
+++ b/packages/coding-agent/test/slash-commands/session.test.ts
@@ -28,7 +28,6 @@ function createRuntimeHarness(options?: {
 				handleSessionCommand,
 				handleSessionDeleteCommand,
 			} as InteractiveModeContext,
-			handleBackgroundCommand: () => {},
 		},
 	};
 }
diff --git a/packages/coding-agent/test/slash-commands/shake.test.ts b/packages/coding-agent/test/slash-commands/shake.test.ts
index 9edc3beb1..e7b606256 100644
--- a/packages/coding-agent/test/slash-commands/shake.test.ts
+++ b/packages/coding-agent/test/slash-commands/shake.test.ts
@@ -31,7 +31,6 @@ function tuiRuntime() {
 			handleShakeCommand,
 			showWarning,
 		} as unknown as InteractiveModeContext,
-		handleBackgroundCommand: vi.fn(),
 	};
 	return { handleShakeCommand, setText, showWarning, runtime };
 }
diff --git a/packages/coding-agent/test/slash-commands/switch.test.ts b/packages/coding-agent/test/slash-commands/switch.test.ts
index d3d4f5aae..15d0c896e 100644
--- a/packages/coding-agent/test/slash-commands/switch.test.ts
+++ b/packages/coding-agent/test/slash-commands/switch.test.ts
@@ -5,7 +5,6 @@ import { executeBuiltinSlashCommand } from "@oh-my-pi/pi-coding-agent/slash-comm
 function createRuntime() {
 	const showModelSelector = vi.fn();
 	const setText = vi.fn();
-	const handleBackgroundCommand = vi.fn();
 	return {
 		showModelSelector,
 		setText,
@@ -13,9 +12,7 @@ function createRuntime() {
 			ctx: {
 				editor: { setText } as unknown as InteractiveModeContext["editor"],
 				showModelSelector,
-				handleBackgroundCommand,
 			} as unknown as InteractiveModeContext,
-			handleBackgroundCommand,
 		},
 	};
 }
diff --git a/packages/coding-agent/test/slash-commands/tan.test.ts b/packages/coding-agent/test/slash-commands/tan.test.ts
new file mode 100644
index 000000000..348978db4
--- /dev/null
+++ b/packages/coding-agent/test/slash-commands/tan.test.ts
@@ -0,0 +1,42 @@
+import { describe, expect, it, vi } from "bun:test";
+import type { InteractiveModeContext } from "@oh-my-pi/pi-coding-agent/modes/types";
+import { executeBuiltinSlashCommand } from "@oh-my-pi/pi-coding-agent/slash-commands/builtin-registry";
+
+function createRuntime() {
+	const handleTanCommand = vi.fn(async () => {});
+	const setText = vi.fn();
+	return {
+		handleTanCommand,
+		setText,
+		runtime: {
+			ctx: {
+				editor: { setText } as unknown as InteractiveModeContext["editor"],
+				handleTanCommand,
+			} as unknown as InteractiveModeContext,
+		},
+	};
+}
+
+describe("/tan slash command", () => {
+	it("routes the full work item through the tan handler", async () => {
+		const harness = createRuntime();
+
+		const handled = await executeBuiltinSlashCommand("/tan add a changelog note", harness.runtime);
+
+		expect(handled).toBe(true);
+		expect(harness.setText).toHaveBeenCalledWith("");
+		expect(harness.handleTanCommand).toHaveBeenCalledWith("add a changelog note");
+	});
+
+	it("preserves the raw multi-word suffix after /tan", async () => {
+		const harness = createRuntime();
+
+		const handled = await executeBuiltinSlashCommand(
+			"/tan    investigate why prompt cache reuse matters here",
+			harness.runtime,
+		);
+
+		expect(handled).toBe(true);
+		expect(harness.handleTanCommand).toHaveBeenCalledWith("investigate why prompt cache reuse matters here");
+	});
+});
diff --git a/packages/coding-agent/test/startup-import-graph.test.ts b/packages/coding-agent/test/startup-import-graph.test.ts
new file mode 100644
index 000000000..2a5d6d6a6
--- /dev/null
+++ b/packages/coding-agent/test/startup-import-graph.test.ts
@@ -0,0 +1,34 @@
+import { describe, expect, it } from "bun:test";
+import * as path from "node:path";
+
+const sourceRoot = path.join(import.meta.dir, "..", "src");
+
+describe("startup import graph", () => {
+	it("keeps normal startup off the aggregate modes barrel", async () => {
+		const mainSource = await Bun.file(path.join(sourceRoot, "main.ts")).text();
+
+		expect(mainSource).toContain('import { InteractiveMode } from "./modes/interactive-mode";');
+		expect(mainSource).not.toContain('from "./modes"');
+	});
+
+	it("keeps branch-only mode runners out of the modes barrel", async () => {
+		const modesBarrelSource = await Bun.file(path.join(sourceRoot, "modes/index.ts")).text();
+
+		expect(modesBarrelSource).toContain('from "./interactive-mode"');
+		expect(modesBarrelSource).not.toContain("runAcpMode");
+		expect(modesBarrelSource).not.toContain("runPrintMode");
+		expect(modesBarrelSource).not.toContain("runRpcMode");
+		expect(modesBarrelSource).not.toContain("./rpc/rpc-mode");
+	});
+
+	it("keeps marketplace implementation behind the lightweight auto-update starter", async () => {
+		const mainSource = await Bun.file(path.join(sourceRoot, "main.ts")).text();
+		const starterSource = await Bun.file(
+			path.join(sourceRoot, "extensibility/plugins/marketplace-auto-update.ts"),
+		).text();
+
+		expect(mainSource).toContain('from "./extensibility/plugins/marketplace-auto-update"');
+		expect(mainSource).not.toContain('from "./extensibility/plugins/marketplace"');
+		expect(starterSource).toContain('await import("./marketplace")');
+	});
+});
diff --git a/packages/coding-agent/test/status-line-context-cache.test.ts b/packages/coding-agent/test/status-line-context-cache.test.ts
index 7abacdfdd..447dd825d 100644
--- a/packages/coding-agent/test/status-line-context-cache.test.ts
+++ b/packages/coding-agent/test/status-line-context-cache.test.ts
@@ -15,9 +15,11 @@
  * (messages.length shrinks) resets the cache.
  */
 import { afterAll, beforeAll, describe, expect, it } from "bun:test";
+import { countTokens } from "@oh-my-pi/pi-natives";
 import { resetSettingsForTest, Settings } from "../src/config/settings";
 import { StatusLineComponent } from "../src/modes/components/status-line";
 import { initTheme } from "../src/modes/theme/theme";
+import { computeNonMessageTokens, estimateToolSchemaTokens } from "../src/modes/utils/context-usage";
 import type { AgentSession } from "../src/session/agent-session";
 
 beforeAll(async () => {
@@ -122,6 +124,38 @@ describe("StatusLineComponent incremental context breakdown cache", () => {
 		expect(v3.usedTokens).toBeGreaterThan(v2.usedTokens);
 	});
 
+	it("non-message token shortcut matches previous category sum semantics", () => {
+		const session = makeSession({
+			messages: [],
+			systemPrompt: [
+				"You are an assistant.\n\n<skills>\n- code: Write code\n- review: Review code\n</skills>",
+				"Loaded context file",
+				"Runtime note",
+			],
+			tools: [
+				{
+					name: "bash",
+					description: "Run shell commands",
+					parameters: { type: "object", properties: { command: { type: "string" } } },
+				},
+			],
+			skills: [
+				{ name: "code", description: "Write code" },
+				{ name: "review", description: "Review code" },
+			],
+		});
+
+		const skillsTokens = countTokens(["code", "Write code", "review", "Review code"]);
+		const previousCategorySum =
+			Math.max(0, countTokens(session.systemPrompt?.[0] ?? "") - skillsTokens) +
+			countTokens((session.systemPrompt ?? []).slice(1)) +
+			estimateToolSchemaTokens(session.agent?.state?.tools ?? []) +
+			skillsTokens;
+
+		expect(new StatusLineComponent(session).getCachedContextBreakdown().usedTokens).toBe(previousCategorySum);
+		expect(computeNonMessageTokens(session)).toBe(previousCategorySum);
+	});
+
 	it("zero messages: produces only non-message tokens, no crash", () => {
 		const session = makeSession({ messages: [] });
 		const comp = new StatusLineComponent(session);
diff --git a/packages/coding-agent/test/streaming-preview-height.test.ts b/packages/coding-agent/test/streaming-preview-height.test.ts
index b5ff45709..86807a5ba 100644
--- a/packages/coding-agent/test/streaming-preview-height.test.ts
+++ b/packages/coding-agent/test/streaming-preview-height.test.ts
@@ -5,8 +5,8 @@ import * as path from "node:path";
 import type { AgentTool } from "@oh-my-pi/pi-agent-core";
 import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
 import { EDIT_MODE_STRATEGIES } from "@oh-my-pi/pi-coding-agent/edit";
-import { initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
-import { TUI } from "@oh-my-pi/pi-tui";
+import { theme as activeTheme, initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
+import { TUI, visibleWidth } from "@oh-my-pi/pi-tui";
 import { VirtualTerminal } from "../../tui/test/virtual-terminal";
 import { ToolExecutionComponent } from "../src/modes/components/tool-execution";
 
@@ -119,7 +119,7 @@ describe("streaming edit preview height (stable, full tail window)", () => {
 			tmpDir,
 		);
 		// Await the actual diff recompute rather than racing the spinner's render
-		// ticks. The streaming spinner calls requestRender every ~16ms, so on a
+		// ticks. The streaming spinner calls requestRender every ~33ms, so on a
 		// slow box a tick — not the (file-read + whole-file Myers) compute — would
 		// resolve the wait and let us sample a stale, mid-abort preview. That is the
 		// CI flake that collapsed Math.min(...steady) to 4. whenPreviewSettled()
@@ -297,3 +297,113 @@ describe("streaming edit preview height (stable, full tail window)", () => {
 		expect(hasDecrease).toBe(true);
 	});
 });
+
+describe("streaming tool call preview height (bounded across renderers)", () => {
+	let themed = false;
+
+	beforeEach(async () => {
+		if (!themed) {
+			await initTheme();
+			themed = true;
+		}
+		resetSettingsForTest();
+		await Settings.init({ inMemory: true, cwd: process.cwd() });
+	});
+
+	afterEach(() => {
+		resetSettingsForTest();
+	});
+
+	function renderPending(toolName: string, args: unknown): { lines: string[]; text: string } {
+		const term = new VirtualTerminal(80, 20);
+		const tui = new TUI(term);
+		const component = new ToolExecutionComponent(toolName, args, {}, undefined, tui, process.cwd());
+		try {
+			const lines = component.render(80);
+			return { lines, text: lines.map(line => Bun.stripANSI(line)).join("\n") };
+		} finally {
+			component.stopAnimation();
+		}
+	}
+
+	test("framed inline tool previews span the full tool width", () => {
+		const width = 80;
+		const { lines } = renderPending("bash", { command: "echo hi" });
+		const strippedLines = lines.map(line => Bun.stripANSI(line));
+		const topBorder = strippedLines.find(line => line.includes(activeTheme.boxSharp.topLeft));
+
+		expect(topBorder).toBeDefined();
+		expect(topBorder?.[0]).toBe(activeTheme.boxSharp.topLeft);
+		expect(topBorder?.endsWith(activeTheme.boxSharp.topRight)).toBe(true);
+		expect(visibleWidth(topBorder ?? "")).toBe(width);
+	});
+
+	test("bash/ssh pending previews stay short even with very long multiline args", () => {
+		const longLines = Array.from({ length: 80 }, (_, i) => `line-${i}`);
+		const cases: Array<{
+			name: string;
+			args: unknown;
+			mustContain: string[];
+			mustHide: string[];
+			marker: RegExp;
+		}> = [
+			{
+				// bash/ssh keep a bounded head+tail window: the start and the
+				// latest are both visible, the middle is elided.
+				name: "bash",
+				args: { command: longLines.join("\n") },
+				mustContain: ["line-0", "line-79"],
+				mustHide: ["line-40"],
+				marker: /more lines/,
+			},
+			{
+				name: "ssh",
+				args: { host: "example", command: longLines.join("\n") },
+				mustContain: ["line-0", "line-79"],
+				mustHide: ["line-40"],
+				marker: /more lines/,
+			},
+		];
+
+		for (const testCase of cases) {
+			const { lines, text } = renderPending(testCase.name, testCase.args);
+			expect(lines.length, `${testCase.name} preview should stay bounded`).toBeLessThan(20);
+			for (const needle of testCase.mustContain) {
+				expect(text, `${testCase.name} preview should keep ${needle}`).toContain(needle);
+			}
+			for (const needle of testCase.mustHide) {
+				expect(text, `${testCase.name} preview should elide ${needle}`).not.toContain(needle);
+			}
+			expect(text, `${testCase.name} preview should advertise truncation`).toMatch(testCase.marker);
+		}
+	});
+
+	test("task pending preview preserves full multiline context", () => {
+		const longLines = Array.from({ length: 80 }, (_, i) => `line-${i}`);
+		const { lines, text } = renderPending("task", {
+			agent: "task",
+			context: longLines.join("\n"),
+			tasks: [{ id: "alpha", description: "preview" }],
+		});
+
+		expect(lines.length, "task preview should not be capped").toBeGreaterThan(80);
+		expect(text).toContain("line-0");
+		expect(text).toContain("line-40");
+		expect(text).toContain("line-79");
+		expect(text).not.toMatch(/more lines/);
+	});
+
+	test("eval pending preview preserves full code (never collapsed)", () => {
+		const longLines = Array.from({ length: 80 }, (_, i) => `line-${i}`);
+		const { lines, text } = renderPending("eval", {
+			cells: [{ language: "js", title: "big", code: longLines.map(line => `const ${line} = 1;`).join("\n") }],
+		});
+
+		expect(lines.length, "eval code preview should not be capped").toBeGreaterThan(80);
+		expect(text).toContain("const line-0 = 1;");
+		expect(text).toContain("const line-40 = 1;");
+		expect(text).toContain("const line-79 = 1;");
+		expect(text).not.toMatch(/more lines/);
+		expect(text).not.toMatch(/earlier lines/);
+	});
+});
diff --git a/packages/coding-agent/test/streaming-reveal.test.ts b/packages/coding-agent/test/streaming-reveal.test.ts
new file mode 100644
index 000000000..0c727ad91
--- /dev/null
+++ b/packages/coding-agent/test/streaming-reveal.test.ts
@@ -0,0 +1,223 @@
+import { afterEach, describe, expect, it, vi } from "bun:test";
+import type { AssistantMessage } from "@oh-my-pi/pi-ai";
+import {
+	buildDisplayMessage,
+	CATCHUP_FRAMES,
+	MIN_STEP,
+	nextStep,
+	STREAMING_REVEAL_FRAME_MS,
+	StreamingRevealController,
+	visibleUnits,
+} from "@oh-my-pi/pi-coding-agent/modes/controllers/streaming-reveal";
+
+function makeUsage(): AssistantMessage["usage"] {
+	return {
+		input: 0,
+		output: 0,
+		cacheRead: 0,
+		cacheWrite: 0,
+		totalTokens: 0,
+		cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
+	};
+}
+
+function makeMessage(content: AssistantMessage["content"]): AssistantMessage {
+	return {
+		role: "assistant",
+		content,
+		api: "anthropic-messages",
+		provider: "anthropic",
+		model: "mock",
+		usage: makeUsage(),
+		stopReason: "stop",
+		timestamp: 0,
+	};
+}
+
+function textAt(message: AssistantMessage, index: number): string {
+	const block = message.content[index];
+	if (block?.type !== "text") {
+		throw new Error(`Expected text block at index ${index}`);
+	}
+	return block.text;
+}
+
+function thinkingAt(message: AssistantMessage, index: number): string {
+	const block = message.content[index];
+	if (block?.type !== "thinking") {
+		throw new Error(`Expected thinking block at index ${index}`);
+	}
+	return block.thinking;
+}
+
+class RecordingComponent {
+	messages: AssistantMessage[] = [];
+
+	updateContent(message: AssistantMessage): void {
+		this.messages.push(message);
+	}
+}
+
+function latestMessage(component: RecordingComponent): AssistantMessage {
+	const message = component.messages.at(-1);
+	if (!message) {
+		throw new Error("Expected at least one rendered message");
+	}
+	return message;
+}
+
+function makeController(options: { smooth?: boolean; hideThinking?: boolean; requestRender?: () => void } = {}) {
+	const component = new RecordingComponent();
+	const controller = new StreamingRevealController({
+		getSmoothStreaming: () => options.smooth ?? true,
+		getHideThinkingBlock: () => options.hideThinking ?? false,
+		requestRender: options.requestRender ?? (() => {}),
+	});
+	return { component, controller };
+}
+
+describe("streaming reveal", () => {
+	afterEach(() => {
+		vi.useRealTimers();
+	});
+
+	it("slices at grapheme boundaries without mutating the target message", () => {
+		const familyEmoji = "👨‍👩‍👧‍👦";
+		const target = makeMessage([{ type: "text", text: `${familyEmoji}B` }]);
+
+		expect(visibleUnits(target, false)).toBe(2);
+		const display = buildDisplayMessage(target, 1, false);
+
+		expect(textAt(display, 0)).toBe(familyEmoji);
+		expect(textAt(target, 0)).toBe(`${familyEmoji}B`);
+	});
+
+	it("excludes hidden thinking from the reveal budget and passes it through", () => {
+		const thinkingBlock = { type: "thinking" as const, thinking: "thought" };
+		const target = makeMessage([thinkingBlock, { type: "text", text: "answer" }]);
+
+		expect(visibleUnits(target, true)).toBe("answer".length);
+		const display = buildDisplayMessage(target, 1, true);
+
+		expect(display.content[0]).toBe(thinkingBlock);
+		expect(thinkingAt(display, 0)).toBe("thought");
+		expect(textAt(display, 1)).toBe("a");
+	});
+
+	it("smooths thinking content when thinking is shown", () => {
+		const target = makeMessage([
+			{ type: "thinking", thinking: "thought" },
+			{ type: "text", text: "answer" },
+		]);
+
+		expect(visibleUnits(target, false)).toBe("thoughtanswer".length);
+		const display = buildDisplayMessage(target, 3, false);
+
+		expect(thinkingAt(display, 0)).toBe("tho");
+		expect(textAt(display, 1)).toBe("");
+	});
+
+	it("uses an adaptive catchup step with the configured floor", () => {
+		const largeBacklog = CATCHUP_FRAMES * 101;
+		const step = nextStep(largeBacklog);
+
+		expect(step).toBe(101);
+		expect(step * CATCHUP_FRAMES).toBeGreaterThanOrEqual(largeBacklog);
+		expect(nextStep(1)).toBe(MIN_STEP);
+		expect(nextStep(MIN_STEP * CATCHUP_FRAMES)).toBe(MIN_STEP);
+	});
+
+	it("reveals cumulative targets to the exact final text with monotonic prefixes", () => {
+		vi.useFakeTimers();
+		const { component, controller } = makeController();
+		const first = makeMessage([{ type: "text", text: "Hello" }]);
+		const second = makeMessage([{ type: "text", text: "Hello world" }]);
+
+		controller.begin(component, makeMessage([{ type: "text", text: "" }]));
+		controller.setTarget(first);
+		for (let i = 0; i < 4; i++) {
+			vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS);
+		}
+		controller.setTarget(second);
+		for (let i = 0; i < 4; i++) {
+			vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS);
+		}
+
+		const renderedTexts = component.messages.map(message => textAt(message, 0));
+		expect(renderedTexts.at(-1)).toBe("Hello world");
+		for (let i = 1; i < renderedTexts.length; i++) {
+			expect(renderedTexts[i].length).toBeGreaterThanOrEqual(renderedTexts[i - 1].length);
+			expect("Hello world".startsWith(renderedTexts[i])).toBe(true);
+		}
+	});
+
+	it("renders full targets immediately when smoothing is disabled", () => {
+		vi.useFakeTimers();
+		const requestRender = vi.fn();
+		const { component, controller } = makeController({ smooth: false, requestRender });
+
+		controller.begin(component, makeMessage([{ type: "text", text: "chunk" }]));
+		controller.setTarget(makeMessage([{ type: "text", text: "chunky" }]));
+		const updates = component.messages.length;
+		vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS * 10);
+
+		expect(textAt(latestMessage(component), 0)).toBe("chunky");
+		expect(component.messages).toHaveLength(updates);
+		expect(requestRender).not.toHaveBeenCalled();
+	});
+
+	it("ticks increasing prefixes at the render cadence", () => {
+		vi.useFakeTimers();
+		const requestRender = vi.fn();
+		const { component, controller } = makeController({ requestRender });
+
+		controller.begin(component, makeMessage([{ type: "text", text: "" }]));
+		controller.setTarget(makeMessage([{ type: "text", text: "abcdefghi" }]));
+
+		vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS);
+		expect(textAt(latestMessage(component), 0)).toBe("abc");
+		vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS);
+		expect(textAt(latestMessage(component), 0)).toBe("abcdef");
+		expect(requestRender).toHaveBeenCalledTimes(2);
+	});
+
+	it("stop halts pending ticker updates", () => {
+		vi.useFakeTimers();
+		const { component, controller } = makeController();
+
+		controller.begin(component, makeMessage([{ type: "text", text: "" }]));
+		controller.setTarget(makeMessage([{ type: "text", text: "abcdefghi" }]));
+		vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS);
+		controller.stop();
+		const updates = component.messages.length;
+		const lastText = textAt(latestMessage(component), 0);
+		vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS * 10);
+
+		expect(component.messages).toHaveLength(updates);
+		expect(textAt(latestMessage(component), 0)).toBe(lastText);
+	});
+
+	it("snaps to full text when a tool call arrives", () => {
+		vi.useFakeTimers();
+		const requestRender = vi.fn();
+		const { component, controller } = makeController({ requestRender });
+
+		controller.begin(component, makeMessage([{ type: "text", text: "" }]));
+		controller.setTarget(makeMessage([{ type: "text", text: "abcdefghi" }]));
+		vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS);
+		expect(textAt(latestMessage(component), 0)).toBe("abc");
+
+		controller.setTarget(
+			makeMessage([
+				{ type: "text", text: "abcdefghi" },
+				{ type: "toolCall", id: "call-1", name: "read", arguments: { path: "README.md" } },
+			]),
+		);
+		const updates = component.messages.length;
+		vi.advanceTimersByTime(STREAMING_REVEAL_FRAME_MS * 10);
+
+		expect(textAt(latestMessage(component), 0)).toBe("abcdefghi");
+		expect(component.messages).toHaveLength(updates);
+		expect(requestRender).toHaveBeenCalledTimes(1);
+	});
+});
diff --git a/packages/coding-agent/test/task/executor-subagent-reminders.test.ts b/packages/coding-agent/test/task/executor-subagent-reminders.test.ts
index 6d346a074..347bb0002 100644
--- a/packages/coding-agent/test/task/executor-subagent-reminders.test.ts
+++ b/packages/coding-agent/test/task/executor-subagent-reminders.test.ts
@@ -474,6 +474,35 @@ describe("runSubprocess yield reminders", () => {
 		expect(result.abortReason).toBe("Cancelled before start");
 		expect(result.stderr).toBe("Cancelled before start");
 	});
+
+	it("surfaces the assistant abort message instead of 'Cancelled by caller' on an internal turn abort", async () => {
+		// No caller signal and no runtime limit: the subagent's own turn ended with
+		// stopReason "aborted" (e.g. a merged request-signal abort). abortReason is
+		// undefined, so the executor must report the assistant's real errorMessage,
+		// not the generic caller-cancellation fallback. This is also what the eval
+		// agent() bridge re-raises, so a blank/misleading reason here surfaces as an
+		// opaque "bridge call '__agent__' failed".
+		const session = createMockSession(({ emit, state }) => {
+			const aborted: AssistantMessage = {
+				...createAssistantStopMessage(""),
+				stopReason: "aborted",
+				errorMessage: "Request was aborted",
+			};
+			state.messages.push(aborted);
+			emit({ type: "message_end", message: aborted });
+		});
+
+		mockCreateAgentSession(session);
+
+		const result = await runSubprocess({ ...baseOptions, id: "subagent-internal-abort" });
+
+		expect(result.aborted).toBe(true);
+		expect(result.exitCode).toBe(1);
+		expect(result.abortReason).toBe("Request was aborted");
+		expect(result.abortReason).not.toBe("Cancelled by caller");
+		expect(result.error).toBeUndefined();
+		expect(result.stderr).toBe("");
+	});
 	it("uses modelRegistry.authStorage when only options.modelRegistry is provided", async () => {
 		const session = createMockSession(({ emit }) => {
 			emit({
diff --git a/packages/coding-agent/test/task/render-call.test.ts b/packages/coding-agent/test/task/render-call.test.ts
index 35393e19c..4ea12d880 100644
--- a/packages/coding-agent/test/task/render-call.test.ts
+++ b/packages/coding-agent/test/task/render-call.test.ts
@@ -1,7 +1,7 @@
 import { afterAll, beforeAll, describe, expect, it } from "bun:test";
 import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
 import { getThemeByName, setThemeInstance, type Theme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
-import type { TaskParams } from "@oh-my-pi/pi-coding-agent/task";
+import type { TaskParams, TaskToolDetails } from "@oh-my-pi/pi-coding-agent/task";
 import { taskToolRenderer } from "@oh-my-pi/pi-coding-agent/task/render";
 
 describe("task renderer: streaming call preview", () => {
@@ -25,6 +25,35 @@ describe("task renderer: streaming call preview", () => {
 		return Bun.stripANSI(component.render(160).join("\n"));
 	}
 
+	function renderCompleted(args: TaskParams): string {
+		const details: TaskToolDetails = {
+			projectAgentsDir: null,
+			totalDurationMs: 12,
+			results: [
+				{
+					index: 0,
+					id: "Only",
+					agent: args.agent,
+					agentSource: "bundled",
+					task: "Render the shared context",
+					exitCode: 0,
+					output: "Done.",
+					stderr: "",
+					truncated: false,
+					durationMs: 12,
+					tokens: 1,
+				},
+			],
+		};
+		const component = taskToolRenderer.renderResult(
+			{ content: [{ type: "text", text: "1 agent completed." }], details },
+			{ expanded: false, isPartial: false },
+			theme,
+			args,
+		);
+		return Bun.stripANSI(component.render(160).join("\n"));
+	}
+
 	// The preview must surface each agent's id + ui description so the user can
 	// see which agents are being dispatched, not a bare "N agents" count.
 	it("lists each task's id and description instead of only a count", () => {
@@ -41,8 +70,9 @@ describe("task renderer: streaming call preview", () => {
 		expect(out).toContain("Audit the auth module");
 		expect(out).toContain("ReviewDb");
 		expect(out).toContain("Audit the db layer");
-		// Count is kept as a compact header, but the old flat "N agents" line is gone.
-		expect(out).toContain("Tasks (2)");
+		// The per-task list stands on its own — neither the redundant "Tasks (N)"
+		// section label nor the old flat "N agents" line is drawn.
+		expect(out).not.toContain("Tasks (");
 		expect(out).not.toContain("2 agents");
 	});
 
@@ -61,7 +91,7 @@ describe("task renderer: streaming call preview", () => {
 		expect(out).toContain("Second");
 		// Missing-id slot falls back to a positional placeholder rather than crashing.
 		expect(out).toContain("#3");
-		expect(out).toContain("Tasks (3)");
+		expect(out).not.toContain("Tasks (");
 	});
 
 	it("caps the collapsed list and reports the overflow as agents", () => {
@@ -84,7 +114,7 @@ describe("task renderer: streaming call preview", () => {
 		expect(expanded).not.toContain("more agents");
 	});
 
-	it("keeps the isolation flag as the final child after the task list", () => {
+	it("surfaces the isolation flag in the header bar", () => {
 		const args: TaskParams = {
 			agent: "task",
 			isolated: true,
@@ -94,15 +124,33 @@ describe("task renderer: streaming call preview", () => {
 		const lines = out.split("\n");
 
 		expect(out).toContain("Only");
-		expect(out).toContain("Isolated");
-		// Isolation flag is rendered last, after every task entry.
-		expect(lines.at(-1)).toContain("Isolated");
+		// Isolation is surfaced as header meta in the frame's top bar (first line),
+		// not as a trailing child row under the task list.
+		expect(lines[0]).toContain("isolated");
 	});
 
-	// Once the tool produces a result, `renderResult` draws each agent as a
-	// progress/result line. The call preview must drop its own per-agent list
-	// so the non-streaming path doesn't render every task twice.
-	it("suppresses the per-task preview list once a result snapshot exists", () => {
+	it("renders shared context as markdown in call and result frames", () => {
+		const args: TaskParams = {
+			agent: "task",
+			context: ["# Goal", "Fix **rendering**.", "", "# Constraints", "- Keep `task` visible"].join("\n"),
+			tasks: [{ id: "Only", description: "Single task", assignment: "..." }],
+		};
+
+		for (const out of [render(args), renderCompleted(args)]) {
+			expect(out).toContain("Goal");
+			expect(out).toContain("Fix rendering.");
+			expect(out).toContain("Constraints");
+			expect(out).toContain("Keep task visible");
+			expect(out).not.toContain("# Goal");
+			expect(out).not.toContain("# Constraints");
+		}
+	});
+
+	// Once the tool produces a result, the container suppresses the call entirely
+	// via `mergeCallAndResult` and `renderResult` draws each agent. As a safety
+	// net, `renderCall` also drops its duplicate per-task preview when a result
+	// snapshot is present, so the two never stack.
+	it("drops the per-task preview list once a result snapshot exists", () => {
 		const args: TaskParams = {
 			agent: "reviewer",
 			tasks: [
@@ -117,9 +165,8 @@ describe("task renderer: streaming call preview", () => {
 		);
 		const out = Bun.stripANSI(component.render(160).join("\n"));
 
-		// Header stays as a section label, but the duplicated agent rows are gone.
-		expect(out).toContain("Tasks (2)");
 		expect(out).not.toContain("Audit the auth module");
 		expect(out).not.toContain("Audit the db layer");
+		expect(out).not.toContain("Tasks (");
 	});
 });
diff --git a/packages/coding-agent/test/task/render-nested-live.test.ts b/packages/coding-agent/test/task/render-nested-live.test.ts
index fac5e4431..422ce07e1 100644
--- a/packages/coding-agent/test/task/render-nested-live.test.ts
+++ b/packages/coding-agent/test/task/render-nested-live.test.ts
@@ -1,9 +1,9 @@
-import { afterAll, beforeAll, describe, expect, it } from "bun:test";
+import { afterAll, afterEach, beforeAll, describe, expect, it, vi } from "bun:test";
 import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
 import { getThemeByName, setThemeInstance } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
-import type { AgentProgress, SingleResult, TaskToolDetails } from "@oh-my-pi/pi-coding-agent/task";
+import type { AgentProgress, SingleResult, TaskParams, TaskToolDetails } from "@oh-my-pi/pi-coding-agent/task";
 import { taskToolRenderer } from "@oh-my-pi/pi-coding-agent/task/render";
-import { formatNumber } from "@oh-my-pi/pi-utils";
+import { formatDuration, formatNumber } from "@oh-my-pi/pi-utils";
 
 describe("task renderer: nested live rendering", () => {
 	beforeAll(async () => {
@@ -14,6 +14,10 @@ describe("task renderer: nested live rendering", () => {
 		setThemeInstance(theme!);
 	});
 
+	afterEach(() => {
+		vi.restoreAllMocks();
+	});
+
 	afterAll(() => {
 		resetSettingsForTest();
 	});
@@ -204,4 +208,84 @@ describe("task renderer: nested live rendering", () => {
 		expect(text).not.toContain("ctx");
 		expect(text).not.toContain("Σ");
 	});
+
+	it("keeps the shared context visible while the task is in progress", async () => {
+		const theme = (await getThemeByName("dark"))!;
+		const details: TaskToolDetails = {
+			projectAgentsDir: null,
+			results: [],
+			totalDurationMs: 0,
+			progress: [makeRunningProgress({ id: "Probe", description: "Investigate padding" })],
+		};
+		const args = {
+			agent: "task",
+			context: "# Goal\nHarden the auth stack before the cut.",
+			tasks: [],
+		} as unknown as TaskParams;
+		const component = taskToolRenderer.renderResult(
+			{ content: [{ type: "text", text: "Running 1 agents..." }], details },
+			{ expanded: false, isPartial: true, spinnerFrame: 0 },
+			theme,
+			args,
+		);
+		const text = Bun.stripANSI(component.render(160).join("\n"));
+		// The brief no longer vanishes the moment the first progress snapshot
+		// replaces the streaming call view.
+		expect(text).toContain("Goal");
+		expect(text).toContain("Harden the auth stack before the cut.");
+	});
+
+	it("renders a static result header while the body shimmers the running task name", async () => {
+		const theme = (await getThemeByName("dark"))!;
+		const details: TaskToolDetails = {
+			projectAgentsDir: null,
+			results: [],
+			totalDurationMs: 0,
+			progress: [makeRunningProgress({ id: "Probe", description: "Investigate padding" })],
+		};
+		let now = 0;
+		vi.spyOn(Date, "now").mockImplementation(() => now);
+		const renderAt = (timeMs: number): string[] => {
+			now = timeMs;
+			const component = taskToolRenderer.renderResult(
+				{ content: [{ type: "text", text: "Running 1 agents..." }], details },
+				{ expanded: false, isPartial: true, spinnerFrame: 0 },
+				theme,
+			);
+			return component.render(160).join("\n").split("\n");
+		};
+		const f0 = renderAt(0);
+		const f1 = renderAt(700);
+		const header0 = f0.find(l => Bun.stripANSI(l).includes("Task"))!;
+		const header1 = f1.find(l => Bun.stripANSI(l).includes("Task"))!;
+		const body0 = f0.find(l => Bun.stripANSI(l).includes("Probe"))!;
+		const body1 = f1.find(l => Bun.stripANSI(l).includes("Probe"))!;
+		// Header is static — no clock ticking beside "Task".
+		expect(header0).toBe(header1);
+		// The per-agent body line still animates via the shimmered task name.
+		expect(body0).not.toBe(body1);
+		expect(Bun.stripANSI(body1)).toContain("• Probe: Investigate padding");
+	});
+
+	it("wraps the completed run summary in bracket glyphs, dropping the Total: label", async () => {
+		const theme = (await getThemeByName("dark"))!;
+		const ok = makeCompletedSubResult("Alpha", "did the thing");
+		const bad: SingleResult = { ...makeCompletedSubResult("Beta", "blew up"), exitCode: 1, error: "boom" };
+		const details: TaskToolDetails = {
+			projectAgentsDir: null,
+			results: [ok, bad],
+			totalDurationMs: 39_400,
+		};
+		const component = taskToolRenderer.renderResult(
+			{ content: [{ type: "text", text: "2 agents completed" }], details },
+			{ expanded: false, isPartial: false },
+			theme,
+		);
+		const text = Bun.stripANSI(component.render(160).join("\n"));
+		// Bash-style bracketed footer: dim bracket chrome wrapping colored counts,
+		// with no "Total:" prefix between the bracket and the first count.
+		const expected = `${theme.format.bracketLeft}1 succeeded${theme.sep.dot}1 failed${theme.sep.dot}${formatDuration(39_400)}${theme.format.bracketRight}`;
+		expect(text).toContain(expected);
+		expect(text).not.toContain("Total:");
+	});
 });
diff --git a/packages/coding-agent/test/task/render-yield-shape.test.ts b/packages/coding-agent/test/task/render-yield-shape.test.ts
new file mode 100644
index 000000000..f59dd4058
--- /dev/null
+++ b/packages/coding-agent/test/task/render-yield-shape.test.ts
@@ -0,0 +1,134 @@
+import { afterAll, beforeAll, describe, expect, it } from "bun:test";
+import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import { getThemeByName, setThemeInstance } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
+import type { AgentProgress, SingleResult, TaskToolDetails } from "@oh-my-pi/pi-coding-agent/task";
+import { taskToolRenderer } from "@oh-my-pi/pi-coding-agent/task/render";
+
+// Regression for #1987: when a subagent stores a non-array value in
+// `extractedToolData.yield`, the renderer cast it to `Array<{ data }>` and
+// then called `?.map`. Optional chaining only short-circuits on null/undefined,
+// so a plain object made `.map` undefined and crashed the TUI with
+// `TypeError: completeData?.map is not a function`. The renderer must tolerate
+// both shapes (array and single object) without throwing, on both the live
+// progress branch (`renderAgentProgress`) and the final result branch
+// (`renderAgentResult`).
+describe("task renderer: malformed yield slot (#1987)", () => {
+	beforeAll(async () => {
+		resetSettingsForTest();
+		await Settings.init({ inMemory: true, cwd: process.cwd() });
+		const theme = await getThemeByName("dark");
+		expect(theme).toBeDefined();
+		setThemeInstance(theme!);
+	});
+
+	afterAll(() => {
+		resetSettingsForTest();
+	});
+
+	const reviewVerdict = {
+		overall_correctness: "correct",
+		confidence: 0.92,
+		explanation: "Looks good.",
+	};
+
+	function makeCompletedResult(extractedToolData: Record<string, unknown>): SingleResult {
+		return {
+			index: 0,
+			id: "reviewer",
+			agent: "reviewer",
+			agentSource: "bundled",
+			task: "review the patch",
+			assignment: "review the patch",
+			description: "review the patch",
+			exitCode: 0,
+			output: "",
+			stderr: "",
+			truncated: false,
+			durationMs: 250,
+			tokens: 100,
+			// Cast deliberately: production typings declare `unknown[]`, but the
+			// renderer must defend against a stray non-array value — that's
+			// exactly what this regression test exercises.
+			extractedToolData: extractedToolData as Record<string, unknown[]>,
+		};
+	}
+
+	function makeCompletedProgress(extractedToolData: Record<string, unknown>): AgentProgress {
+		return {
+			index: 0,
+			id: "reviewer",
+			agent: "reviewer",
+			agentSource: "bundled",
+			status: "completed",
+			task: "review the patch",
+			assignment: "review the patch",
+			description: "review the patch",
+			recentTools: [],
+			recentOutput: [],
+			toolCount: 1,
+			tokens: 100,
+			cost: 0,
+			durationMs: 250,
+			extractedToolData: extractedToolData as Record<string, unknown[]>,
+		};
+	}
+
+	async function renderResultText(extractedToolData: Record<string, unknown>): Promise<string> {
+		const theme = (await getThemeByName("dark"))!;
+		const details: TaskToolDetails = {
+			projectAgentsDir: null,
+			results: [makeCompletedResult(extractedToolData)],
+			totalDurationMs: 250,
+		};
+		const component = taskToolRenderer.renderResult(
+			{ content: [{ type: "text", text: "" }], details },
+			{ expanded: false, isPartial: false, spinnerFrame: 0 },
+			theme,
+		);
+		return Bun.stripANSI(component.render(160).join("\n"));
+	}
+
+	async function renderProgressText(extractedToolData: Record<string, unknown>): Promise<string> {
+		const theme = (await getThemeByName("dark"))!;
+		const details: TaskToolDetails = {
+			projectAgentsDir: null,
+			results: [],
+			totalDurationMs: 250,
+			progress: [makeCompletedProgress(extractedToolData)],
+		};
+		const component = taskToolRenderer.renderResult(
+			{ content: [{ type: "text", text: "" }], details },
+			{ expanded: false, isPartial: true, spinnerFrame: 0 },
+			theme,
+		);
+		return Bun.stripANSI(component.render(160).join("\n"));
+	}
+
+	it("does not throw and still surfaces the verdict when yield is a single object (result branch)", async () => {
+		const text = await renderResultText({
+			yield: { data: reviewVerdict, status: "success" },
+		});
+		expect(text).toContain("correct");
+	});
+
+	it("does not throw and still surfaces the verdict when yield is a single object (progress branch)", async () => {
+		const text = await renderProgressText({
+			yield: { data: reviewVerdict, status: "success" },
+		});
+		expect(text).toContain("correct");
+	});
+
+	it("does not throw when yield is a non-object primitive (both branches)", async () => {
+		// Primitives can't carry a verdict — renderer must drop them silently
+		// instead of crashing.
+		await expect(renderResultText({ yield: "not-an-array" })).resolves.toBeString();
+		await expect(renderProgressText({ yield: 42 })).resolves.toBeString();
+	});
+
+	it("still renders the canonical array shape unchanged", async () => {
+		const text = await renderResultText({
+			yield: [{ data: reviewVerdict, status: "success" }],
+		});
+		expect(text).toContain("correct");
+	});
+});
diff --git a/packages/coding-agent/test/task/task-progress-render.test.ts b/packages/coding-agent/test/task/task-progress-render.test.ts
new file mode 100644
index 000000000..2a217d19a
--- /dev/null
+++ b/packages/coding-agent/test/task/task-progress-render.test.ts
@@ -0,0 +1,100 @@
+import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
+import type { RenderResultOptions } from "@oh-my-pi/pi-agent-core";
+import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import { getThemeByName } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
+import { taskToolRenderer } from "@oh-my-pi/pi-coding-agent/task/render";
+import type { AgentProgress, TaskToolDetails } from "@oh-my-pi/pi-coding-agent/task/types";
+
+function runningProgress(overrides: Partial<AgentProgress> = {}): AgentProgress {
+	return {
+		index: 0,
+		id: "KeySettingsHotPaths",
+		agent: "task",
+		agentSource: "bundled",
+		status: "running",
+		task: "investigate hot paths",
+		recentTools: [],
+		recentOutput: [],
+		toolCount: 0,
+		tokens: 0,
+		cost: 0,
+		durationMs: 0,
+		...overrides,
+	};
+}
+
+function detailsFor(progress: AgentProgress): TaskToolDetails {
+	return { projectAgentsDir: null, results: [], totalDurationMs: 0, progress: [progress] };
+}
+
+function findRow(component: { render: (w: number) => string[] }, needle: string): string {
+	const row = component
+		.render(120)
+		.join("\n")
+		.split("\n")
+		.find(line => Bun.stripANSI(line).includes(needle));
+	expect(row).toBeDefined();
+	return row!;
+}
+
+describe("task progress rendering", () => {
+	beforeEach(async () => {
+		resetSettingsForTest();
+		await Settings.init({ inMemory: true });
+	});
+
+	afterEach(() => {
+		vi.restoreAllMocks();
+		resetSettingsForTest();
+	});
+	it("uses a static bullet and shimmers only the running subagent name", async () => {
+		const theme = (await getThemeByName("dark"))!;
+		expect(theme).toBeDefined();
+		// Pin the sweep so the shimmer crest deterministically lands on the name.
+		vi.spyOn(Date, "now").mockReturnValue(683);
+		const options: RenderResultOptions = { expanded: false, isPartial: true, spinnerFrame: 0 };
+		const progress = runningProgress({ id: "CountPackages", description: "List workspace packages" });
+
+		const rawRow = findRow(
+			taskToolRenderer.renderResult(
+				{ content: [{ type: "text", text: "" }], details: detailsFor(progress) },
+				options,
+				theme,
+			),
+			"CountPackages",
+		);
+		const strippedRow = Bun.stripANSI(rawRow);
+
+		expect(strippedRow).toContain("• CountPackages: List workspace packages");
+		expect(strippedRow).not.toContain(theme.status.running);
+		expect(strippedRow).not.toContain(theme.getSpinnerFrames("status")[0]);
+		// Bold crest only comes from the shimmer palette; the description remains
+		// one solid, non-shimmered run.
+		expect(rawRow).toContain("\x1b[1m");
+		const descriptionIndex = rawRow.indexOf(": List workspace packages");
+		expect(descriptionIndex).toBeGreaterThan(0);
+		expect(rawRow.slice(descriptionIndex)).not.toContain("\x1b[1m");
+	});
+
+	it("keeps the bullet replacement when shimmer is disabled", async () => {
+		const theme = (await getThemeByName("dark"))!;
+		resetSettingsForTest();
+		await Settings.init({ inMemory: true, overrides: { "display.shimmer": "disabled" } });
+		const options: RenderResultOptions = { expanded: false, isPartial: true, spinnerFrame: 0 };
+
+		const strippedRow = Bun.stripANSI(
+			findRow(
+				taskToolRenderer.renderResult(
+					{ content: [{ type: "text", text: "" }], details: detailsFor(runningProgress()) },
+					options,
+					theme,
+				),
+				"KeySettingsHotPaths",
+			),
+		);
+
+		expect(strippedRow).toContain("• KeySettingsHotPaths");
+		expect(strippedRow).not.toContain(theme.status.running);
+		expect(strippedRow).not.toContain(theme.getSpinnerFrames("status")[0]);
+	});
+});
diff --git a/packages/coding-agent/test/telemetry-export.test.ts b/packages/coding-agent/test/telemetry-export.test.ts
index ecdaf86e8..fa0238b04 100644
--- a/packages/coding-agent/test/telemetry-export.test.ts
+++ b/packages/coding-agent/test/telemetry-export.test.ts
@@ -33,45 +33,45 @@ afterEach(() => {
 });
 
 describe("initTelemetryExport gating", () => {
-	it("stays disabled when no OTLP endpoint is configured", () => {
-		initTelemetryExport();
+	it("stays disabled when no OTLP endpoint is configured", async () => {
+		await initTelemetryExport();
 		expect(isTelemetryExportEnabled()).toBe(false);
 	});
 
-	it("stays disabled when OTEL_SDK_DISABLED=true even with an endpoint", () => {
+	it("stays disabled when OTEL_SDK_DISABLED=true even with an endpoint", async () => {
 		process.env.OTEL_EXPORTER_OTLP_ENDPOINT = "http://localhost:4318";
 		process.env.OTEL_SDK_DISABLED = "true";
-		initTelemetryExport();
+		await initTelemetryExport();
 		expect(isTelemetryExportEnabled()).toBe(false);
 	});
 
-	it("stays disabled when OTEL_TRACES_EXPORTER=none even with an endpoint", () => {
+	it("stays disabled when OTEL_TRACES_EXPORTER=none even with an endpoint", async () => {
 		process.env.OTEL_EXPORTER_OTLP_ENDPOINT = "http://localhost:4318";
 		process.env.OTEL_TRACES_EXPORTER = "none";
-		initTelemetryExport();
+		await initTelemetryExport();
 		expect(isTelemetryExportEnabled()).toBe(false);
 	});
 
-	it("declines unsupported OTLP protocols instead of misrouting spans", () => {
+	it("declines unsupported OTLP protocols instead of misrouting spans", async () => {
 		process.env.OTEL_EXPORTER_OTLP_ENDPOINT = "http://localhost:4317";
 		process.env.OTEL_EXPORTER_OTLP_PROTOCOL = "grpc";
-		initTelemetryExport();
+		await initTelemetryExport();
 		expect(isTelemetryExportEnabled()).toBe(false);
 
 		process.env.OTEL_EXPORTER_OTLP_TRACES_PROTOCOL = "http/json";
-		initTelemetryExport();
+		await initTelemetryExport();
 		expect(isTelemetryExportEnabled()).toBe(false);
 	});
 
-	it("honors the kill-switches case-insensitively per the OTEL env contract", () => {
+	it("honors the kill-switches case-insensitively per the OTEL env contract", async () => {
 		process.env.OTEL_EXPORTER_OTLP_ENDPOINT = "http://localhost:4318";
 		process.env.OTEL_SDK_DISABLED = "TRUE";
-		initTelemetryExport();
+		await initTelemetryExport();
 		expect(isTelemetryExportEnabled()).toBe(false);
 
 		delete process.env.OTEL_SDK_DISABLED;
 		process.env.OTEL_TRACES_EXPORTER = "otlp,None";
-		initTelemetryExport();
+		await initTelemetryExport();
 		expect(isTelemetryExportEnabled()).toBe(false);
 	});
 });
diff --git a/packages/coding-agent/test/tiny-device.test.ts b/packages/coding-agent/test/tiny-device.test.ts
index f83a117ea..2c5cb09ec 100644
--- a/packages/coding-agent/test/tiny-device.test.ts
+++ b/packages/coding-agent/test/tiny-device.test.ts
@@ -49,13 +49,6 @@ describe("tiny model device setting → PI_TINY_DEVICE mapping", () => {
 		expect(tinyModelDeviceSettingToEnv("cuda")).toBe("cuda");
 	});
 
-	it("keeps every non-default setting value resolvable by the worker", () => {
-		for (const value of TINY_MODEL_DEVICE_SETTING_VALUES) {
-			if (value === TINY_MODEL_DEVICE_DEFAULT) continue;
-			expect(() => normalizeTinyModelDevice(tinyModelDeviceSettingToEnv(value))).not.toThrow();
-		}
-	});
-
 	it("keeps submenu options aligned with the accepted values", () => {
 		expect(TINY_MODEL_DEVICE_SETTING_OPTIONS.map(option => option.value)).toEqual([
 			...TINY_MODEL_DEVICE_SETTING_VALUES,
diff --git a/packages/coding-agent/test/tiny-title-generator.test.ts b/packages/coding-agent/test/tiny-title-generator.test.ts
index c744ca940..c6ce8c636 100644
--- a/packages/coding-agent/test/tiny-title-generator.test.ts
+++ b/packages/coding-agent/test/tiny-title-generator.test.ts
@@ -48,6 +48,7 @@ function createRegistry(model: Model<Api>) {
 	return {
 		getAvailable: () => [model],
 		getApiKey: async () => "test-key",
+		resolver: vi.fn(() => async () => "test-key"),
 	} as never;
 }
 
diff --git a/packages/coding-agent/test/title-generator.test.ts b/packages/coding-agent/test/title-generator.test.ts
index dc1a6b4a4..46069a57f 100644
--- a/packages/coding-agent/test/title-generator.test.ts
+++ b/packages/coding-agent/test/title-generator.test.ts
@@ -29,6 +29,9 @@ function createRegistry(model: Model<Api>) {
 	return {
 		getAvailable: () => [model],
 		getApiKey: async () => "test-key",
+		getApiKeyForProvider: async () => "test-key",
+		authStorage: { rotateSessionCredential: async () => false },
+		resolver: () => async () => "test-key",
 	} as never;
 }
 
diff --git a/packages/coding-agent/test/tool-discovery/initial-tools.test.ts b/packages/coding-agent/test/tool-discovery/initial-tools.test.ts
index 82560c87f..b629d5986 100644
--- a/packages/coding-agent/test/tool-discovery/initial-tools.test.ts
+++ b/packages/coding-agent/test/tool-discovery/initial-tools.test.ts
@@ -111,6 +111,11 @@ describe("computeEssentialBuiltinNames", () => {
 });
 
 describe("tools.discoveryMode settings schema", () => {
+	it("defaults to auto discovery mode", () => {
+		const settings = Settings.isolated({});
+		expect(settings.get("tools.discoveryMode")).toBe("auto");
+	});
+
 	it("back-compat: mcp.discoveryMode still accepted", () => {
 		const settings = Settings.isolated({ "mcp.discoveryMode": true });
 		expect(settings.get("mcp.discoveryMode")).toBe(true);
diff --git a/packages/coding-agent/test/tool-discovery/subagent.test.ts b/packages/coding-agent/test/tool-discovery/subagent.test.ts
index eb34191d9..3c2c911e1 100644
--- a/packages/coding-agent/test/tool-discovery/subagent.test.ts
+++ b/packages/coding-agent/test/tool-discovery/subagent.test.ts
@@ -1,5 +1,6 @@
 import { describe, expect, it } from "bun:test";
 import { Settings } from "../../src/config/settings";
+import { resolveEffectiveToolDiscoveryMode, TOOL_DISCOVERY_AUTO_THRESHOLD } from "../../src/tool-discovery/mode";
 
 // ─── Subagent discovery mode inheritance tests ────────────────────────────────
 // These are unit-level tests that verify the settings resolution logic
@@ -7,11 +8,8 @@ import { Settings } from "../../src/config/settings";
 // ─────────────────────────────────────────────────────────────────────────────
 
 describe("effective discovery mode resolution", () => {
-	function resolveEffectiveMode(settings: Settings): "off" | "mcp-only" | "all" {
-		const toolsMode = settings.get("tools.discoveryMode");
-		if (toolsMode !== "off") return toolsMode as "off" | "mcp-only" | "all";
-		if (settings.get("mcp.discoveryMode")) return "mcp-only";
-		return "off";
+	function resolveEffectiveMode(settings: Settings, toolCount = 0): "off" | "mcp-only" | "all" {
+		return resolveEffectiveToolDiscoveryMode(settings, toolCount);
 	}
 
 	it("tools.discoveryMode=all beats mcp.discoveryMode=false", () => {
@@ -34,8 +32,14 @@ describe("effective discovery mode resolution", () => {
 		expect(resolveEffectiveMode(s)).toBe("off");
 	});
 
-	it("default settings → off", () => {
+	it("default auto settings stay off at the threshold", () => {
 		const s = Settings.isolated({});
-		expect(resolveEffectiveMode(s)).toBe("off");
+		expect(s.get("tools.discoveryMode")).toBe("auto");
+		expect(resolveEffectiveMode(s, TOOL_DISCOVERY_AUTO_THRESHOLD)).toBe("off");
+	});
+
+	it("default auto settings enable mcp-only above the threshold", () => {
+		const s = Settings.isolated({});
+		expect(resolveEffectiveMode(s, TOOL_DISCOVERY_AUTO_THRESHOLD + 1)).toBe("mcp-only");
 	});
 });
diff --git a/packages/coding-agent/test/tool-execution-args.test.ts b/packages/coding-agent/test/tool-execution-args.test.ts
index e0983d0cf..1a4a66dc1 100644
--- a/packages/coding-agent/test/tool-execution-args.test.ts
+++ b/packages/coding-agent/test/tool-execution-args.test.ts
@@ -1,12 +1,14 @@
 import { afterEach, describe, expect, it, vi } from "bun:test";
+import type { AgentTool } from "@oh-my-pi/pi-agent-core";
 import { initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
-import type { TUI } from "@oh-my-pi/pi-tui";
+import { Text, type TUI } from "@oh-my-pi/pi-tui";
 import { ToolExecutionComponent } from "../src/modes/components/tool-execution";
 
 describe("ToolExecutionComponent.updateArgs (F8 — no clone, ref-eq fast path)", () => {
 	let initialized = false;
 
 	afterEach(() => {
+		vi.useRealTimers();
 		vi.restoreAllMocks();
 	});
 
@@ -31,23 +33,28 @@ describe("ToolExecutionComponent.updateArgs (F8 — no clone, ref-eq fast path)"
 
 		expect(cloneSpy).not.toHaveBeenCalled();
 	});
+	it("keeps bash spinner cadence when the shimmer border repaints at 30fps", async () => {
+		if (!initialized) {
+			await initTheme();
+			initialized = true;
+		}
+		vi.useFakeTimers();
+		let renderState: { spinnerFrame?: number } | undefined;
+		const uiStub = { requestRender: vi.fn() } as unknown as TUI;
+		const tool = {
+			label: "Bash",
+			renderCall: (_args: unknown, options: { spinnerFrame?: number }) => {
+				renderState = options;
+				return new Text("", 0, 0);
+			},
+			execute: async () => ({ content: [] }),
+		} as unknown as AgentTool;
+		const component = new ToolExecutionComponent("bash", { command: "echo ok" }, {}, tool, uiStub);
 
-	it("short-circuits when called with the exact same args reference", async () => {
-		const component = await makeComponent({ command: "ls" });
-		const args = { command: "ls -al" };
+		component.setArgsComplete();
+		vi.advanceTimersByTime(170);
 
-		component.updateArgs(args);
-		// Second call with the SAME object reference should be a no-op.
-		// (Render bookkeeping doesn't re-fire — assert via #args not changing.)
-		component.updateArgs(args);
-		component.updateArgs(args);
-
-		// Different object content → must NOT be short-circuited.
-		const next = { command: "echo hi" };
-		component.updateArgs(next);
-
-		// Re-issuing the prior reference is now stale but still ref-distinct.
-		// The component must accept it without crashing.
-		expect(() => component.updateArgs(args)).not.toThrow();
+		expect(renderState?.spinnerFrame).toBe(2);
+		component.stopAnimation();
 	});
 });
diff --git a/packages/coding-agent/test/tool-live-region-scrollback.test.ts b/packages/coding-agent/test/tool-live-region-scrollback.test.ts
index e30e56e82..80fa86305 100644
--- a/packages/coding-agent/test/tool-live-region-scrollback.test.ts
+++ b/packages/coding-agent/test/tool-live-region-scrollback.test.ts
@@ -73,7 +73,7 @@ describe("tool live-region scrollback", () => {
 					.join("\n");
 				expect(bufferText).not.toContain("pending [1/1]");
 				expect(bufferText).toContain("const line9 = 9;");
-				expect(bufferText).toContain("… 10 more lines");
+				expect(bufferText).toContain("const line19 = 19;");
 			} finally {
 				component.stopAnimation();
 				tui.stop();
@@ -139,6 +139,175 @@ describe("tool live-region scrollback", () => {
 			}
 		});
 	});
+
+	it("commits the scrolled-off head of an over-tall expanded streaming write to scrollback", async () => {
+		if (process.platform === "win32") return;
+
+		await withTerminalRisk(true, async () => {
+			const term = new VirtualTerminal(120, 12);
+			(term as unknown as { isNativeViewportAtBottom: () => boolean | undefined }).isNativeViewportAtBottom = () =>
+				undefined;
+			const tui = new TUI(term);
+			const chat = new TranscriptContainer();
+			const body = (n: number) => Array.from({ length: n }, (_unused, i) => `MARK-${i}`).join("\n");
+			const filePath = "packages/coding-agent/test/probe.txt";
+			// Expanded (Ctrl+O) lifts the tail-window cap, so the preview renders the
+			// whole content top-anchored — append-only growth as chunks stream in.
+			const component = new ToolExecutionComponent(
+				"write",
+				{ file_path: filePath, content: body(4) },
+				{},
+				undefined,
+				tui,
+				process.cwd(),
+			);
+			component.setExpanded(true);
+
+			try {
+				chat.addChild(component);
+				tui.addChild(chat);
+				tui.start();
+				tui.setEagerNativeScrollbackRebuild(true);
+				await term.waitForRender();
+
+				// A short preview that fits, then the full preview that alone overflows
+				// the 12-row viewport — the frame that scrolls the head above the top.
+				component.updateArgs({ file_path: filePath, content: body(4) });
+				tui.requestRender();
+				await term.waitForRender();
+
+				component.updateArgs({ file_path: filePath, content: body(40) });
+				tui.requestRender();
+				await term.waitForRender();
+
+				const strip = (rows: string[]) => rows.map(row => Bun.stripANSI(row).trimEnd()).join("\n");
+				const scrollText = strip(term.getScrollBuffer());
+				const viewportText = strip(term.getViewport());
+
+				// MARK-0 scrolled above the viewport: it must live in native scrollback
+				// (committed), not nowhere. Before the fix the tool block was not
+				// append-only, so its scrolled-off head was dropped — a yanked stream.
+				expect(viewportText).not.toContain("MARK-0");
+				expect(scrollText).toContain("MARK-0");
+				// The streaming tail stays on screen, and nothing went missing between.
+				expect(viewportText).toContain("MARK-39");
+				expect(viewportText).toContain("(streaming)");
+				expect(scrollText).toContain("MARK-20");
+			} finally {
+				component.stopAnimation();
+				tui.stop();
+				await term.flush();
+			}
+		});
+	});
+
+	it("treats a tool block as append-only only while its expanded preview streams", async () => {
+		const filePath = "packages/coding-agent/test/probe.txt";
+		const tui = new TUI(new VirtualTerminal(80, 24));
+		const args = { file_path: filePath, content: "MARK-0\nMARK-1\nMARK-2" };
+		const component = new ToolExecutionComponent("write", args, {}, undefined, tui, process.cwd());
+		type AppendOnly = { isTranscriptBlockAppendOnly(): boolean };
+		const probe = component as unknown as AppendOnly;
+		try {
+			// Collapsed: the preview slides a bounded tail window — not append-only.
+			expect(probe.isTranscriptBlockAppendOnly()).toBe(false);
+			// Expanded + streaming: append-only, eligible for head commit.
+			component.setExpanded(true);
+			expect(probe.isTranscriptBlockAppendOnly()).toBe(true);
+			// Once a final result lands the preview may collapse — boundary closes.
+			component.updateResult({ content: [{ type: "text", text: "" }], details: { path: filePath } }, false);
+			expect(probe.isTranscriptBlockAppendOnly()).toBe(false);
+		} finally {
+			component.stopAnimation();
+		}
+	});
+
+	it("commits the scrolled-off head of an expanded eval whose output streams past the viewport", async () => {
+		if (process.platform === "win32") return;
+
+		await withTerminalRisk(true, async () => {
+			const term = new VirtualTerminal(120, 12);
+			(term as unknown as { isNativeViewportAtBottom: () => boolean | undefined }).isNativeViewportAtBottom = () =>
+				undefined;
+			const tui = new TUI(term);
+			const chat = new TranscriptContainer();
+			const title = "stream lots of output";
+			const code = "for (let i = 0; i < 40; i++) console.log('MARK-' + i);";
+			const args = { cells: [{ language: "js", title, code }] };
+			const component = new ToolExecutionComponent("eval", args, {}, undefined, tui, process.cwd());
+			component.setExpanded(true);
+			const out = (n: number) => Array.from({ length: n }, (_unused, i) => `MARK-${i}`).join("\n");
+			const partial = (output: string) =>
+				component.updateResult(
+					{
+						content: [{ type: "text", text: "" }],
+						details: { cells: [{ index: 0, title, code, language: "js", output, status: "running" }] },
+					},
+					true,
+				);
+
+			try {
+				chat.addChild(component);
+				tui.addChild(chat);
+				tui.start();
+				tui.setEagerNativeScrollbackRebuild(true);
+				await term.waitForRender();
+
+				// A short output that fits, then the full stream that alone overflows the
+				// 12-row viewport — the frame that scrolls the output head above the top.
+				partial(out(4));
+				tui.requestRender();
+				await term.waitForRender();
+
+				partial(out(40));
+				tui.requestRender();
+				await term.waitForRender();
+
+				const strip = (rows: string[]) => rows.map(row => Bun.stripANSI(row).trimEnd()).join("\n");
+				const scrollText = strip(term.getScrollBuffer());
+				const viewportText = strip(term.getViewport());
+
+				// The streamed output head scrolled above the viewport: it must live in
+				// native scrollback (committed), not nowhere. The fixed code cell rides
+				// along as the stable prefix above it.
+				expect(viewportText).not.toContain("MARK-0");
+				expect(scrollText).toContain("MARK-0");
+				expect(scrollText).toContain("MARK-20");
+				// The streaming tail stays on screen, and nothing went missing between.
+				expect(viewportText).toContain("MARK-39");
+			} finally {
+				component.stopAnimation();
+				tui.stop();
+				await term.flush();
+			}
+		});
+	});
+
+	it("keeps a streaming eval append-only only while expanded and unfinalized", () => {
+		const tui = new TUI(new VirtualTerminal(80, 24));
+		const title = "t";
+		const code = "console.log('x')";
+		const args = { cells: [{ language: "js", title, code }] };
+		const component = new ToolExecutionComponent("eval", args, {}, undefined, tui, process.cwd());
+		type AppendOnly = { isTranscriptBlockAppendOnly(): boolean };
+		const probe = component as unknown as AppendOnly;
+		const details = {
+			cells: [{ index: 0, title, code, language: "js", output: "MARK-0\nMARK-1", status: "running" }],
+		};
+		try {
+			// Collapsed: bounded sliding tail windows — not append-only.
+			expect(probe.isTranscriptBlockAppendOnly()).toBe(false);
+			component.setExpanded(true);
+			// Expanded + partial (streaming output): append-only.
+			component.updateResult({ content: [{ type: "text", text: "" }], details }, true);
+			expect(probe.isTranscriptBlockAppendOnly()).toBe(true);
+			// Final result may collapse to a capped view — boundary closes.
+			component.updateResult({ content: [{ type: "text", text: "" }], details }, false);
+			expect(probe.isTranscriptBlockAppendOnly()).toBe(false);
+		} finally {
+			component.stopAnimation();
+		}
+	});
 });
 
 function makeAssistantMessage(text: string): AssistantMessage {
diff --git a/packages/coding-agent/test/tools.test.ts b/packages/coding-agent/test/tools.test.ts
index ff0825548..7493cbb1e 100644
--- a/packages/coding-agent/test/tools.test.ts
+++ b/packages/coding-agent/test/tools.test.ts
@@ -16,6 +16,7 @@ import { JobTool } from "@oh-my-pi/pi-coding-agent/tools/job";
 import { wrapToolWithMetaNotice } from "@oh-my-pi/pi-coding-agent/tools/output-meta";
 import { ReadTool } from "@oh-my-pi/pi-coding-agent/tools/read";
 import { DEFAULT_FILE_LIMIT, MULTI_FILE_PER_FILE_MATCHES, SearchTool } from "@oh-my-pi/pi-coding-agent/tools/search";
+import * as toolTimeouts from "@oh-my-pi/pi-coding-agent/tools/tool-timeouts";
 import { WriteTool } from "@oh-my-pi/pi-coding-agent/tools/write";
 import { $which, Snowflake } from "@oh-my-pi/pi-utils";
 import { unzipSync } from "fflate";
@@ -1164,7 +1165,7 @@ function b() {
 			const updates: string[] = [];
 			const result = await bashTool.execute(
 				"test-call-8-stream",
-				{ command: "for i in 1 2 3; do echo $i; sleep 0.2; done" },
+				{ command: "for i in 1 2 3; do echo $i; sleep 0.1; done" },
 				undefined,
 				update => {
 					const text = update.content?.find(c => c.type === "text")?.text ?? "";
@@ -1310,13 +1311,19 @@ function b() {
 					),
 				),
 			);
+			// Drive the effective timeout via the production clamp seam so the
+			// backgrounded job times out in ~0.5s instead of a real wall-clock
+			// second. 0.5s still renders as "1 seconds" in the executor message
+			// (Math.round), so that delivery assertion is unchanged; the
+			// auto-background-on-timeout decision path is identical.
+			vi.spyOn(toolTimeouts, "clampTimeout").mockReturnValue(0.5);
 
 			const result = await autoBackgroundBashTool.execute("test-call-9-auto-timeout-background", {
 				command: "printf 'start\\n'; sleep 1.2; printf 'done\\n'",
 				timeout: 1,
 			});
 
-			expect(result.details?.timeoutSeconds).toBe(1);
+			expect(result.details?.timeoutSeconds).toBe(0.5);
 			expect(result.details?.async?.state).toBe("running");
 			expect(getTextOutput(result)).toContain("Background job");
 			const jobId = result.details?.async?.jobId;
@@ -1344,6 +1351,9 @@ function b() {
 		});
 
 		it("should respect timeout", async () => {
+			// Reduce the effective timeout through the production clamp seam; the
+			// real subprocess kill-on-timeout path is still exercised, just faster.
+			vi.spyOn(toolTimeouts, "clampTimeout").mockReturnValue(0.1);
 			await expect(bashTool.execute("test-call-10", { command: "sleep 5", timeout: 1 })).rejects.toThrow(
 				/timed out/i,
 			);
@@ -1351,10 +1361,19 @@ function b() {
 
 		it("should abort and recover for subsequent commands", async () => {
 			const controller = new AbortController();
-			const promise = bashTool.execute("test-call-10-abort", { command: "sleep 60" }, controller.signal);
-			// Give the native shell a beat to enter `sleep`; do not depend on chunk
-			// delivery timing, which is flaky on loaded CI runners.
-			await Bun.sleep(100);
+			const started = Promise.withResolvers<void>();
+			const promise = bashTool.execute(
+				"test-call-10-abort",
+				{ command: "echo READY; sleep 60" },
+				controller.signal,
+				update => {
+					const text = update.content?.find(c => c.type === "text")?.text ?? "";
+					if (text.includes("READY")) started.resolve();
+				},
+			);
+			// Abort as soon as the command has emitted output (proving the shell is
+			// live), instead of blindly waiting a fixed beat for it to enter `sleep`.
+			await started.promise;
 			controller.abort("test abort");
 			await expect(promise).rejects.toThrow(/abort|cancel|timed out/i);
 
diff --git a/packages/coding-agent/test/tools/approval-mode.test.ts b/packages/coding-agent/test/tools/approval-mode.test.ts
index 9b0709881..f2fa648c2 100644
--- a/packages/coding-agent/test/tools/approval-mode.test.ts
+++ b/packages/coding-agent/test/tools/approval-mode.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, describe, expect, it } from "bun:test";
+import { afterAll, beforeAll, describe, expect, it } from "bun:test";
 import * as fs from "node:fs";
 import * as os from "node:os";
 import * as path from "node:path";
@@ -19,31 +19,6 @@ function emptyWorkspaceTree(cwd: string) {
 	return { rootPath: cwd, rendered: ".\n", truncated: false, totalLines: 1, agentsMdFiles: [] };
 }
 
-async function makeSession(extraSettings: Record<string, unknown> = {}) {
-	const tempDir = fs.mkdtempSync(path.join(os.tmpdir(), `pi-approval-mode-${Snowflake.next()}-`));
-	const cwd = path.join(tempDir, "cwd");
-	fs.mkdirSync(cwd, { recursive: true });
-	const sessionManager = SessionManager.create(cwd, path.join(tempDir, "sessions"));
-	const settings = Settings.isolated({ ...BASE_SETTINGS, ...extraSettings });
-	const { session } = await createAgentSession({
-		cwd,
-		agentDir: tempDir,
-		sessionManager,
-		settings,
-		model: getBundledModel("openai", "gpt-4o-mini"),
-		disableExtensionDiscovery: true,
-		skills: [],
-		contextFiles: [],
-		workspaceTree: emptyWorkspaceTree(cwd),
-		promptTemplates: [],
-		slashCommands: [],
-		enableMCP: false,
-		enableLsp: false,
-		toolNames: ["bash"],
-	});
-	return { tempDir, session, settings };
-}
-
 function textOf(result: { content?: ReadonlyArray<{ type: string; text?: string }> }): string {
 	const blocks = result.content ?? [];
 	for (const block of blocks) {
@@ -53,179 +28,155 @@ function textOf(result: { content?: ReadonlyArray<{ type: string; text?: string
 }
 
 describe("tools.approvalMode setting", () => {
-	const tempDirs: string[] = [];
+	// The per-tool approval gate (ExtensionToolWrapper) reads approvalMode / tools.approval /
+	// autoApprove exclusively from the execute-time AgentToolContext, never from the session's
+	// own settings. So a single shared session exercises every mode — we only vary the context
+	// settings per assertion. This avoids paying createAgentSession's cost (model registry,
+	// auth-storage discovery, settings init) nine times over.
+	let tempDir: string;
+	let session: Awaited<ReturnType<typeof createAgentSession>>["session"];
 
-	afterEach(async () => {
-		for (const tempDir of tempDirs.splice(0)) {
-			// Windows can briefly hold tempdir handles after session.dispose(); retry a few times.
-			for (let attempt = 0; attempt < 5; attempt++) {
-				try {
-					fs.rmSync(tempDir, { recursive: true, force: true });
-					break;
-				} catch (err) {
-					const code = (err as NodeJS.ErrnoException).code;
-					if (code !== "EBUSY" && code !== "ENOTEMPTY" && code !== "EPERM") throw err;
-					if (attempt === 4) break; // best-effort: OS will reclaim
-					await Bun.sleep(50 * (attempt + 1));
-				}
+	beforeAll(async () => {
+		tempDir = fs.mkdtempSync(path.join(os.tmpdir(), `pi-approval-mode-${Snowflake.next()}-`));
+		const cwd = path.join(tempDir, "cwd");
+		fs.mkdirSync(cwd, { recursive: true });
+		const sessionManager = SessionManager.create(cwd, path.join(tempDir, "sessions"));
+		const created = await createAgentSession({
+			cwd,
+			agentDir: tempDir,
+			sessionManager,
+			settings: Settings.isolated(BASE_SETTINGS),
+			model: getBundledModel("openai", "gpt-4o-mini"),
+			disableExtensionDiscovery: true,
+			skills: [],
+			contextFiles: [],
+			workspaceTree: emptyWorkspaceTree(cwd),
+			promptTemplates: [],
+			slashCommands: [],
+			enableMCP: false,
+			enableLsp: false,
+			toolNames: ["bash"],
+		});
+		session = created.session;
+	});
+
+	afterAll(async () => {
+		await session.dispose();
+		// Windows can briefly hold tempdir handles after session.dispose(); retry a few times.
+		for (let attempt = 0; attempt < 5; attempt++) {
+			try {
+				fs.rmSync(tempDir, { recursive: true, force: true });
+				break;
+			} catch (err) {
+				const code = (err as NodeJS.ErrnoException).code;
+				if (code !== "EBUSY" && code !== "ENOTEMPTY" && code !== "EPERM") throw err;
+				if (attempt === 4) break; // best-effort: OS will reclaim
+				await Bun.sleep(50 * (attempt + 1));
 			}
 		}
 	});
 
+	function approvalSettings(extraSettings: Record<string, unknown> = {}): Settings {
+		return Settings.isolated({ ...BASE_SETTINGS, ...extraSettings });
+	}
+
+	function bashTool() {
+		const bash = session.getToolByName("bash");
+		if (!bash) throw new Error("Expected bash tool");
+		return bash;
+	}
+
 	it("yolo mode (default) bypasses approval for non-overriding tool calls", async () => {
-		const { tempDir, session, settings } = await makeSession();
-		tempDirs.push(tempDir);
-		try {
-			const bash = session.getToolByName("bash");
-			if (!bash) throw new Error("Expected bash tool");
-			const result = await bash.execute("yolo", { command: "echo ok" }, undefined, undefined, {
-				settings,
-			} as AgentToolContext);
-			expect(textOf(result)).toContain("ok");
-		} finally {
-			await session.dispose();
-		}
+		const settings = approvalSettings();
+		const result = await bashTool().execute("yolo", { command: "echo ok" }, undefined, undefined, {
+			settings,
+		} as AgentToolContext);
+		expect(textOf(result)).toContain("ok");
 	});
 
 	it("always-ask mode rejects exec tools when no UI is available", async () => {
-		const { tempDir, session, settings } = await makeSession({
-			"tools.approvalMode": "always-ask",
-		});
-		tempDirs.push(tempDir);
-		try {
-			const bash = session.getToolByName("bash");
-			if (!bash) throw new Error("Expected bash tool");
-			await expect(
-				bash.execute("always-ask", { command: "echo blocked" }, undefined, undefined, {
-					settings,
-				} as AgentToolContext),
-			).rejects.toThrow(/requires approval but no interactive UI available/);
-		} finally {
-			await session.dispose();
-		}
+		const settings = approvalSettings({ "tools.approvalMode": "always-ask" });
+		await expect(
+			bashTool().execute("always-ask", { command: "echo blocked" }, undefined, undefined, {
+				settings,
+			} as AgentToolContext),
+		).rejects.toThrow(/requires approval but no interactive UI available/);
 	});
 
 	it("per-tool allow overrides are honored in every mode", async () => {
-		const { tempDir, session, settings } = await makeSession({
+		const settings = approvalSettings({
 			"tools.approvalMode": "always-ask",
 			"tools.approval": { bash: "allow" },
 		});
-		tempDirs.push(tempDir);
-		try {
-			const bash = session.getToolByName("bash");
-			if (!bash) throw new Error("Expected bash tool");
-			const result = await bash.execute("always-ask-allow", { command: "echo allowed" }, undefined, undefined, {
-				settings,
-			} as AgentToolContext);
-			expect(textOf(result)).toContain("allowed");
-		} finally {
-			await session.dispose();
-		}
+		const result = await bashTool().execute("always-ask-allow", { command: "echo allowed" }, undefined, undefined, {
+			settings,
+		} as AgentToolContext);
+		expect(textOf(result)).toContain("allowed");
 	});
 
 	it("per-tool prompt overrides can tighten yolo mode", async () => {
-		const { tempDir, session, settings } = await makeSession({
+		const settings = approvalSettings({
 			"tools.approvalMode": "yolo",
 			"tools.approval": { bash: "prompt" },
 		});
-		tempDirs.push(tempDir);
-		try {
-			const bash = session.getToolByName("bash");
-			if (!bash) throw new Error("Expected bash tool");
-			await expect(
-				bash.execute("yolo-prompt", { command: "echo blocked" }, undefined, undefined, {
-					settings,
-				} as AgentToolContext),
-			).rejects.toThrow(/requires approval but no interactive UI available/);
-		} finally {
-			await session.dispose();
-		}
+		await expect(
+			bashTool().execute("yolo-prompt", { command: "echo blocked" }, undefined, undefined, {
+				settings,
+			} as AgentToolContext),
+		).rejects.toThrow(/requires approval but no interactive UI available/);
 	});
 
 	it("write mode still prompts exec-tier tools", async () => {
-		const { tempDir, session, settings } = await makeSession({
+		const settings = approvalSettings({
 			"tools.approvalMode": "write",
 			"tools.approval": {},
 		});
-		tempDirs.push(tempDir);
-		try {
-			const bash = session.getToolByName("bash");
-			if (!bash) throw new Error("Expected bash tool");
-			await expect(
-				bash.execute("write-mode", { command: "echo unconfigured" }, undefined, undefined, {
-					settings,
-				} as AgentToolContext),
-			).rejects.toThrow(/requires approval but no interactive UI available/);
-		} finally {
-			await session.dispose();
-		}
+		await expect(
+			bashTool().execute("write-mode", { command: "echo unconfigured" }, undefined, undefined, {
+				settings,
+			} as AgentToolContext),
+		).rejects.toThrow(/requires approval but no interactive UI available/);
 	});
 
 	it("critical bash patterns do not prompt in yolo mode with bash allowed", async () => {
-		const { tempDir, session, settings } = await makeSession({
+		const settings = approvalSettings({
 			"tools.approvalMode": "yolo",
 			"tools.approval": { bash: "allow" },
 		});
-		tempDirs.push(tempDir);
-		try {
-			const bash = session.getToolByName("bash");
-			if (!bash) throw new Error("Expected bash tool");
-
-			const result = await bash.execute(
-				"critical",
-				{ command: "rm -f /tmp/bun-fake-timer-probe.test.ts" },
-				undefined,
-				undefined,
-				{
-					settings,
-				} as AgentToolContext,
-			);
-			expect(textOf(result)).toContain("(no output)");
-		} finally {
-			await session.dispose();
-		}
+		const result = await bashTool().execute(
+			"critical",
+			{ command: "rm -f /tmp/bun-fake-timer-probe.test.ts" },
+			undefined,
+			undefined,
+			{
+				settings,
+			} as AgentToolContext,
+		);
+		expect(textOf(result)).toContain("(no output)");
 	});
 
 	it("CLI --auto-approve forces yolo mode for non-overriding tool calls", async () => {
-		const { tempDir, session, settings } = await makeSession({
-			"tools.approvalMode": "always-ask",
-		});
-		tempDirs.push(tempDir);
-		try {
-			const bash = session.getToolByName("bash");
-			if (!bash) throw new Error("Expected bash tool");
-			const result = await bash.execute("cli-override", { command: "echo override" }, undefined, undefined, {
-				settings,
-				autoApprove: true,
-			} as AgentToolContext);
-			expect(textOf(result)).toContain("override");
-		} finally {
-			await session.dispose();
-		}
+		const settings = approvalSettings({ "tools.approvalMode": "always-ask" });
+		const result = await bashTool().execute("cli-override", { command: "echo override" }, undefined, undefined, {
+			settings,
+			autoApprove: true,
+		} as AgentToolContext);
+		expect(textOf(result)).toContain("override");
 	});
 
 	it("CLI --auto-approve also bypasses safety-override patterns", async () => {
-		const { tempDir, session, settings } = await makeSession({
-			"tools.approvalMode": "always-ask",
-		});
-		tempDirs.push(tempDir);
-		try {
-			const bash = session.getToolByName("bash");
-			if (!bash) throw new Error("Expected bash tool");
-			const result = await bash.execute(
-				"cli-critical",
-				{ command: "rm -f /tmp/bun-fake-timer-probe.test.ts" },
-				undefined,
-				undefined,
-				{
-					settings,
-					autoApprove: true,
-				} as AgentToolContext,
-			);
-			expect(textOf(result)).toContain("(no output)");
-		} finally {
-			await session.dispose();
-		}
+		const settings = approvalSettings({ "tools.approvalMode": "always-ask" });
+		const result = await bashTool().execute(
+			"cli-critical",
+			{ command: "rm -f /tmp/bun-fake-timer-probe.test.ts" },
+			undefined,
+			undefined,
+			{
+				settings,
+				autoApprove: true,
+			} as AgentToolContext,
+		);
+		expect(textOf(result)).toContain("(no output)");
 	});
 
 	it("constructs an extensionRunner unconditionally so the approval gate is always installed", async () => {
@@ -236,12 +187,6 @@ describe("tools.approvalMode setting", () => {
 		// any non-yolo approval mode setting would be a no-op without feedback. The
 		// fix is to construct the runner unconditionally; this test makes that contract explicit so
 		// a future change to make the runner optional again cannot silently re-open the hole.
-		const { tempDir, session } = await makeSession();
-		tempDirs.push(tempDir);
-		try {
-			expect(session.extensionRunner).toBeDefined();
-		} finally {
-			await session.dispose();
-		}
+		expect(session.extensionRunner).toBeDefined();
 	});
 });
diff --git a/packages/coding-agent/test/tools/ask.test.ts b/packages/coding-agent/test/tools/ask.test.ts
index e189676d6..560839b5e 100644
--- a/packages/coding-agent/test/tools/ask.test.ts
+++ b/packages/coding-agent/test/tools/ask.test.ts
@@ -1153,6 +1153,57 @@ describe("AskTool option markers", () => {
 		expect(text).not.toContain(theme!.radio.unselected);
 	});
 
+	it("keeps option rows stable across repeated renders", async () => {
+		const theme = await getThemeByName("dark");
+		expect(theme).toBeDefined();
+		const options = [
+			{ label: "TypeScript" },
+			{ label: "Rust" },
+			{ label: "Python" },
+			{ label: "Go" },
+			{ label: "C" },
+			{ label: "C++" },
+			{ label: "Zig" },
+			{ label: "Java" },
+			{ label: "Swift" },
+			{ label: "Haskell" },
+		];
+		const renderedCall = askToolRenderer.renderCall(
+			{ questions: [{ id: "fav_lang", question: "Which programming language?", options }] },
+			{ expanded: true, isPartial: true },
+			theme!,
+		);
+
+		const firstCall = stripAnsi(renderedCall.render(120).join("\n"));
+		const secondCall = stripAnsi(renderedCall.render(120).join("\n"));
+		expect(secondCall).toBe(firstCall);
+		expect(secondCall.match(/TypeScript/g)?.length).toBe(1);
+		expect(secondCall.match(/Haskell/g)?.length).toBe(1);
+
+		const renderedResult = askToolRenderer.renderResult(
+			{
+				content: [{ type: "text", text: "" }],
+				details: {
+					results: [
+						{
+							id: "fav_lang",
+							question: "Which programming language?",
+							options: options.map(option => option.label),
+							multi: false,
+							selectedOptions: ["Python"],
+						},
+					],
+				},
+			},
+			{ expanded: true, isPartial: false },
+			theme!,
+		);
+		const firstResult = stripAnsi(renderedResult.render(120).join("\n"));
+		const secondResult = stripAnsi(renderedResult.render(120).join("\n"));
+		expect(secondResult).toBe(firstResult);
+		expect(secondResult.match(/TypeScript/g)?.length).toBe(1);
+		expect(secondResult.match(/Haskell/g)?.length).toBe(1);
+	});
 	it("renders single-choice result selection with a filled radio marker", async () => {
 		const theme = await getThemeByName("dark");
 		expect(theme).toBeDefined();
diff --git a/packages/coding-agent/test/tools/ast-edit.test.ts b/packages/coding-agent/test/tools/ast-edit.test.ts
index 5262909e1..64bf976a6 100644
--- a/packages/coding-agent/test/tools/ast-edit.test.ts
+++ b/packages/coding-agent/test/tools/ast-edit.test.ts
@@ -211,9 +211,10 @@ describe("ast_edit tool schema", () => {
 				| { totalReplacements?: number; fileReplacements?: Array<{ path: string; count: number }> }
 				| undefined;
 
-			// Tree-grouped output: `# packages/pkg-…/src/` then `## root.ts#<hash> (1 replacement)`.
+			// Multi-level tree output: `# packages/pkg-…/src/`, `## root.ts#<hash>`, then a
+			// nested `## nested/` directory with `### child.ts#<hash>` under it.
 			expect(text).toMatch(/^## root\.ts#[0-9A-F]{4} \(\d+ replacement[s]?\)$/m);
-			expect(text).toMatch(/^## child\.ts#[0-9A-F]{4} \(\d+ replacement[s]?\)$/m);
+			expect(text).toMatch(/^### child\.ts#[0-9A-F]{4} \(\d+ replacement[s]?\)$/m);
 			expect(text).not.toContain("ignore.js");
 			expect(text).not.toContain("outside.ts");
 			expect(details?.totalReplacements).toBe(2);
diff --git a/packages/coding-agent/test/tools/ast-grep.test.ts b/packages/coding-agent/test/tools/ast-grep.test.ts
index b70a139de..f584859dd 100644
--- a/packages/coding-agent/test/tools/ast-grep.test.ts
+++ b/packages/coding-agent/test/tools/ast-grep.test.ts
@@ -100,9 +100,10 @@ describe("ast_grep parse errors", () => {
 			const text = result.content.find(content => content.type === "text")?.text ?? "";
 			const details = result.details as { matchCount?: number; fileCount?: number } | undefined;
 
-			// Directory mode uses tree-grouped `# dir/` + `## name#hash` headers.
-			expect(text).toMatch(/## root\.ts#[0-9A-F]{4}/);
-			expect(text).toMatch(/## child\.ts#[0-9A-F]{4}/);
+			// Multi-level tree output: `# packages/pkg-…/src/`, `## root.ts#<hash>`, then a
+			// nested `## nested/` directory with `### child.ts#<hash>` under it.
+			expect(text).toMatch(/^## root\.ts#[0-9A-F]{4}/m);
+			expect(text).toMatch(/^### child\.ts#[0-9A-F]{4}/m);
 			expect(text).not.toContain("ignore.js");
 			expect(text).not.toContain("outside.ts");
 			expect(details?.matchCount).toBe(2);
diff --git a/packages/coding-agent/test/tools/auto-generated-guard.test.ts b/packages/coding-agent/test/tools/auto-generated-guard.test.ts
index 50a053878..0360a8ab1 100644
--- a/packages/coding-agent/test/tools/auto-generated-guard.test.ts
+++ b/packages/coding-agent/test/tools/auto-generated-guard.test.ts
@@ -45,38 +45,6 @@ describe("assertEditableFileContent", () => {
 			"/**\n * This file was generated by kysely-codegen.\n * Please do not edit it manually.\n */\n\nexport interface Database {}";
 		expect(() => assertEditableFileContent(content, "db.ts")).toThrow(ToolError);
 	});
-
-	it("does not block broad prose comment markers", () => {
-		const content = "// auto generated dont edit bla bla\n// this is a hand-written file note\nexport const foo = 1;";
-		expect(() => assertEditableFileContent(content, "test.ts")).not.toThrow();
-	});
-
-	it("does not match generated markers after code starts", () => {
-		const content = "export const foo = 1;\n\n// Code generated by sqlc. DO NOT EDIT.";
-		expect(() => assertEditableFileContent(content, "test.ts")).not.toThrow();
-	});
-
-	it("uses language-specific comment styles", () => {
-		const tsContent = "# Code generated by sqlc. DO NOT EDIT.\nexport const foo = 1;";
-		expect(() => assertEditableFileContent(tsContent, "test.ts")).not.toThrow();
-
-		const pyContent = "// Code generated by sqlc. DO NOT EDIT.\nvalue = 1";
-		expect(() => assertEditableFileContent(pyContent, "test.py")).not.toThrow();
-	});
-
-	it("does not block editing the guard file itself", async () => {
-		const guardPath = path.join(import.meta.dir, "../../src/tools/auto-generated-guard.ts");
-		const content = await Bun.file(guardPath).text();
-		expect(() =>
-			assertEditableFileContent(content, "packages/coding-agent/src/tools/auto-generated-guard.ts"),
-		).not.toThrow();
-	});
-
-	it("checks only first 1024 bytes of content", () => {
-		const prefix = "A".repeat(1024);
-		const content = `${prefix}\n// Code generated by sqlc. DO NOT EDIT.`;
-		expect(() => assertEditableFileContent(content, "test.ts")).not.toThrow();
-	});
 });
 
 describe("assertEditableFile", () => {
diff --git a/packages/coding-agent/test/tools/bash-sixel-render.test.ts b/packages/coding-agent/test/tools/bash-sixel-render.test.ts
index 90e47d5cb..28db6c233 100644
--- a/packages/coding-agent/test/tools/bash-sixel-render.test.ts
+++ b/packages/coding-agent/test/tools/bash-sixel-render.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, describe, expect, it } from "bun:test";
+import { afterEach, describe, expect, it, vi } from "bun:test";
 import * as os from "node:os";
 import * as path from "node:path";
 import type { RenderResultOptions } from "@oh-my-pi/pi-agent-core";
@@ -83,9 +83,9 @@ describe("bashToolRenderer", () => {
 		expect(lines.length).toBeGreaterThanOrEqual(3);
 		const header = lines[0]!;
 		const body = lines.slice(1, -1).join("\n");
-		// The header carries the title only; the command lives inside the framed body
-		// (not inline on the status line as the old one-liner preview rendered it).
-		expect(header).toContain("Bash");
+		// Bash commands already carry a `$` prompt in the body, so the frame header
+		// stays a plain rule instead of repeating "Bash" in the title bar.
+		expect(header).not.toContain("Bash");
 		expect(header).not.toContain("sleep 30");
 		expect(body).toContain("$ sleep 30");
 	});
@@ -126,6 +126,29 @@ describe("bashToolRenderer", () => {
 		// only place wall time is shown so users don't read it twice.
 		expect(rendered).not.toContain("Wall time: 1.23 seconds");
 	});
+
+	it("folds raw output artifact notices into the status footer", async () => {
+		const theme = await getThemeByName("dark");
+		expect(theme).toBeDefined();
+		const uiTheme = theme!;
+		const component = bashToolRenderer.renderResult(
+			{
+				content: [{ type: "text", text: "filtered\n[raw output: artifact://13]\n\nWall time: 0.08 seconds" }],
+				details: { timeoutSeconds: 300, wallTimeMs: 80 },
+				isError: false,
+			},
+			{ expanded: false, isPartial: false },
+			uiTheme,
+			{ command: "bun run check:types" },
+		);
+		const rendered = sanitizeText(component.render(120).join("\n"));
+		expect(rendered).toContain("filtered");
+		expect(rendered).toContain("Wall: 0.08s");
+		expect(rendered).toContain("Timeout: 300s");
+		expect(rendered).toContain("Artifact: 13");
+		expect(rendered).not.toContain("[raw output: artifact://13]");
+		expect(rendered).not.toContain("artifact://13");
+	});
 	it("renders the exit status in the footer and strips the textual exit notice for failed commands", async () => {
 		const theme = await getThemeByName("dark");
 		expect(theme).toBeDefined();
@@ -234,4 +257,36 @@ describe("bashToolRenderer", () => {
 			expect(rendered[idx]).toMatch(/\u001b\[38;(?:2|5);/);
 		}
 	});
+
+	it("keeps a backgrounded command's border static while a foreground one still shimmers", async () => {
+		const theme = await getThemeByName("dark");
+		expect(theme).toBeDefined();
+		const uiTheme = theme!;
+		// Render a still-partial bash result at two wall-clock instants a quarter of
+		// a shimmer cycle apart. The animated bottom-edge segment lives at the far
+		// left at t=0 and near center at t=750ms, so an animating border yields
+		// different bytes across the two frames while a static one is identical.
+		const renderAt = (details: Record<string, unknown>, now: number): string => {
+			const spy = vi.spyOn(Date, "now").mockReturnValue(now);
+			try {
+				const component = bashToolRenderer.renderResult(
+					{ content: [{ type: "text", text: "Background job bg_1 started: sleep 30" }], details, isError: false },
+					{ expanded: false, isPartial: true },
+					uiTheme,
+					{ command: "sleep 30" },
+				);
+				return component.render(60).join("\n");
+			} finally {
+				spy.mockRestore();
+			}
+		};
+
+		// Backgrounded (finalizes later via the async update path): no shimmer, so
+		// the committed frame can't freeze a stray dark "bar" into the border.
+		const backgrounded = { async: { state: "running", jobId: "bg_1", type: "bash" } };
+		expect(renderAt(backgrounded, 0)).toBe(renderAt(backgrounded, 750));
+
+		// Foreground pending: the border still sweeps, so frames differ over time.
+		expect(renderAt({}, 0)).not.toBe(renderAt({}, 750));
+	});
 });
diff --git a/packages/coding-agent/test/tools/conflict-integration.test.ts b/packages/coding-agent/test/tools/conflict-integration.test.ts
index d1b5b318e..5d1a92a8b 100644
--- a/packages/coding-agent/test/tools/conflict-integration.test.ts
+++ b/packages/coding-agent/test/tools/conflict-integration.test.ts
@@ -26,7 +26,11 @@ function getText(result: { content: Array<{ type: string; text?: string }> }): s
 }
 
 async function getTool(session: ToolSession, name: "read" | "write") {
-	const tools = await createTools(session);
+	// Request only the tool under test: createTools(session) with no toolNames
+	// builds every builtin factory (LSP, MCP discovery, browser, eval preflight,
+	// …) on each call, which is pure overhead here. The conflict contract lives
+	// entirely in the read/write tools + session.conflictHistory.
+	const tools = await createTools(session, [name]);
 	const tool = tools.find(entry => entry.name === name);
 	if (!tool) throw new Error(`Missing ${name} tool`);
 	return tool;
@@ -511,7 +515,7 @@ describe("write resolves conflicts via conflict://N", () => {
 		await read.execute("read-hashed", { path: "hashed.ts" });
 		const result = await write.execute("write-hashed", {
 			path: "conflict://1",
-			content: "¶hashed.ts#1a2b\n42:cleanline\n",
+			content: "[hashed.ts#1a2b]\n42:cleanline\n",
 		});
 		expect(getText(result)).toContain("auto-stripped hashline display prefixes");
 		const after = await Bun.file(filePath).text();
diff --git a/packages/coding-agent/test/tools/edit-renderer.test.ts b/packages/coding-agent/test/tools/edit-renderer.test.ts
index 75c8d1e16..01e8abd97 100644
--- a/packages/coding-agent/test/tools/edit-renderer.test.ts
+++ b/packages/coding-agent/test/tools/edit-renderer.test.ts
@@ -8,7 +8,7 @@ import { resetSettingsForTest, Settings } from "@oh-my-pi/pi-coding-agent/config
 import { editToolRenderer } from "@oh-my-pi/pi-coding-agent/edit/renderer";
 import { ToolExecutionComponent } from "@oh-my-pi/pi-coding-agent/modes/components/tool-execution";
 import * as themeModule from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
-import { Text, type TUI } from "@oh-my-pi/pi-tui";
+import { Text, type TUI, visibleWidth } from "@oh-my-pi/pi-tui";
 
 beforeAll(async () => {
 	resetSettingsForTest();
@@ -53,11 +53,41 @@ describe("editToolRenderer", () => {
 		expect(rendered).toContain("packages/coding-agent/src/edit/renderer.ts");
 	});
 
+	it("lifts the streaming diff tail window when expanded", async () => {
+		const uiTheme = await getUiTheme();
+		const diff = Array.from({ length: 20 }, (_, index) =>
+			index === 0 ? "-head-line-1" : `+tail-line-${index + 1}`,
+		).join("\n");
+		const renderPreview = (expanded: boolean): string =>
+			Bun.stripANSI(
+				editToolRenderer
+					.renderCall(
+						{ file_path: "/tmp/preview.ts", previewDiff: diff },
+						{ expanded, isPartial: true, spinnerFrame: 0, renderContext: { editMode: "replace" } },
+						uiTheme,
+					)
+					.render(200)
+					.join("\n"),
+			);
+
+		const collapsed = renderPreview(false);
+		expect(collapsed).toContain("tail-line-20");
+		expect(collapsed).not.toContain("head-line-1");
+		expect(collapsed).toContain("more lines above");
+		expect(collapsed).toContain("(preview)");
+
+		const expanded = renderPreview(true);
+		expect(expanded).toContain("head-line-1");
+		expect(expanded).toContain("tail-line-20");
+		expect(expanded).not.toContain("more lines above");
+		expect(expanded).not.toContain("(preview)");
+	});
+
 	it("uses hashline input headers for streaming call path without apply_patch errors", async () => {
 		const uiTheme = await getUiTheme();
 		const component = editToolRenderer.renderCall(
 			{
-				input: "¶packages/coding-agent/src/edit/renderer.ts\nEOF:\n|// preview",
+				input: "[packages/coding-agent/src/edit/renderer.ts]\ninsert tail:\n+// preview",
 			},
 			{ expanded: false, isPartial: true, spinnerFrame: 0, renderContext: { editMode: "hashline" } },
 			uiTheme,
@@ -75,9 +105,12 @@ describe("editToolRenderer", () => {
 		const component = new ToolExecutionComponent(
 			"edit",
 			{
-				input: ["*** Begin Patch", "¶crates/pi-natives/src/shell.rs", "EOF:", "|pub fn streaming_preview() {"].join(
-					"\n",
-				),
+				input: [
+					"*** Begin Patch",
+					"[crates/pi-natives/src/shell.rs]",
+					"insert tail:",
+					"+pub fn streaming_preview() {",
+				].join("\n"),
 			},
 			{},
 			hashlineTool,
@@ -86,8 +119,8 @@ describe("editToolRenderer", () => {
 
 		const rendered = Bun.stripANSI(component.render(160).join("\n"));
 		expect(rendered).toContain("crates/pi-natives/src/shell.rs");
-		expect(rendered).not.toContain("EOF:");
-		expect(rendered).not.toContain("|pub fn streaming_preview() {");
+		expect(rendered).not.toContain("insert tail:");
+		expect(rendered).not.toContain("+pub fn streaming_preview() {");
 		expect(rendered).not.toContain("*** Begin Patch");
 	});
 
@@ -95,7 +128,7 @@ describe("editToolRenderer", () => {
 		const uiTheme = await getUiTheme();
 		const compactComponent = editToolRenderer.renderCall(
 			{
-				input: "¶foo bar.ts\nBOF:\n|// preview",
+				input: "[foo bar.ts]\ninsert head:\n+// preview",
 			},
 			{ expanded: true, isPartial: true, spinnerFrame: 0, renderContext: { editMode: "hashline" } },
 			uiTheme,
@@ -103,7 +136,7 @@ describe("editToolRenderer", () => {
 
 		const quotedComponent = editToolRenderer.renderCall(
 			{
-				input: "¶'baz qux.ts'\nBOF:\n|// preview",
+				input: "['baz qux.ts']\ninsert head:\n+// preview",
 			},
 			{ expanded: false, isPartial: true, spinnerFrame: 0, renderContext: { editMode: "hashline" } },
 			uiTheme,
@@ -115,33 +148,33 @@ describe("editToolRenderer", () => {
 		expect(quotedRendered).toContain("baz qux.ts");
 	});
 
-	it("strips canonical `¶` and longer `¶` runs from hashline input headers", async () => {
+	it("strips bracket delimiters from hashline input headers", async () => {
 		const uiTheme = await getUiTheme();
 
-		// Canonical `¶PATH` form — the parser strips the marker and the
+		// Canonical `[PATH]` form — the parser strips the delimiters and the
 		// renderer keeps the title clean.
 		const canonical = editToolRenderer.renderCall(
 			{
-				input: "¶packages/coding-agent/src/slash-commands/builtin-registry.ts\nBOF:\n|// preview",
+				input: "[packages/coding-agent/src/slash-commands/builtin-registry.ts]\ninsert head:\n+// preview",
 			},
 			{ expanded: true, isPartial: true, spinnerFrame: 0, renderContext: { editMode: "hashline" } },
 			uiTheme,
 		);
 
-		// Even longer runs should still produce the clean path.
-		const triple = editToolRenderer.renderCall(
-			{ input: "¶¶¶a/b/c.ts\nBOF:\n|// preview" },
+		// While streaming, the closing bracket may not have arrived yet.
+		const partial = editToolRenderer.renderCall(
+			{ input: "[a/b/c.ts\ninsert head:\n+// preview" },
 			{ expanded: true, isPartial: true, spinnerFrame: 0, renderContext: { editMode: "hashline" } },
 			uiTheme,
 		);
 
 		const canonicalRendered = Bun.stripANSI(canonical.render(160).join("\n"));
-		const tripleRendered = Bun.stripANSI(triple.render(160).join("\n"));
+		const partialRendered = Bun.stripANSI(partial.render(160).join("\n"));
 
 		expect(canonicalRendered).toContain("packages/coding-agent/src/slash-commands/builtin-registry.ts");
-		expect(canonicalRendered).not.toMatch(/¶packages\/coding-agent/);
-		expect(tripleRendered).toContain("a/b/c.ts");
-		expect(tripleRendered).not.toMatch(/¶+a\/b\/c\.ts/);
+		expect(canonicalRendered).not.toMatch(/\[packages\/coding-agent/);
+		expect(partialRendered).toContain("a/b/c.ts");
+		expect(partialRendered).not.toMatch(/\[a\/b\/c\.ts/);
 	});
 
 	it("uses hashline input headers for completed single-file result path", async () => {
@@ -157,7 +190,7 @@ describe("editToolRenderer", () => {
 			{ expanded: false, isPartial: false, renderContext: { editMode: "hashline" } },
 			uiTheme,
 			{
-				input: "¶packages/coding-agent/src/edit/renderer.ts\nEOF:\n|// preview",
+				input: "[packages/coding-agent/src/edit/renderer.ts]\ninsert tail:\n+// preview",
 			},
 		);
 
@@ -182,7 +215,7 @@ describe("editToolRenderer", () => {
 			// The trailing payload line carries no newline — the common shape for a
 			// single-line edit. The streaming pass trims that in-flight line, so the
 			// preview only becomes computable once args are marked complete.
-			const input = `¶memory.ts#${tag}\nreplace 2..2:\n+export const b = 22;`;
+			const input = `[memory.ts#${tag}]\nreplace 2..2:\n+export const b = 22;`;
 			const component = new ToolExecutionComponent("edit", { input }, { snapshots }, hashlineTool, uiStub, tmpDir);
 
 			component.setArgsComplete();
@@ -208,7 +241,7 @@ describe("editToolRenderer", () => {
 
 			const snapshots = new InMemorySnapshotStore();
 			const tag = snapshots.record(filePath, content);
-			const input = `¶memory.ts#${tag}\nreplace 2..2:\n+export const b = 22;\n`;
+			const input = `[memory.ts#${tag}]\nreplace 2..2:\n+export const b = 22;\n`;
 			const component = new ToolExecutionComponent(
 				"edit",
 				{ __partialJson: input },
@@ -271,4 +304,50 @@ describe("editToolRenderer", () => {
 		const rendered = Bun.stripANSI(component.render(160).join("\n"));
 		expect(rendered).toContain("plain streamed text");
 	});
+
+	it("renders change stats inline on the result header with no separate metadata or stats row", async () => {
+		const uiTheme = await getUiTheme();
+		const diff = [" 115│ ctx", "-116│ old", "+117│ new one", "+118│ new two"].join("\n");
+		const component = editToolRenderer.renderResult(
+			{
+				content: [{ type: "text", text: "Updated demo.go" }],
+				details: { diff, op: "update" },
+			},
+			{ expanded: false, isPartial: false, renderContext: { editMode: "hashline" } },
+			uiTheme,
+			{ file_path: "demo.go" },
+		);
+
+		const lines = Bun.stripANSI(component.render(160).join("\n")).split("\n");
+		// Stats ride on the header line next to the path…
+		expect(lines[0]).toContain("demo.go");
+		expect(lines[0]).toContain("+2");
+		expect(lines[0]).toContain("-1");
+		expect(lines[0]).toContain("1 hunk");
+		// …only there (no standalone stats row), and the diff starts immediately
+		// below the header (no blank line, no lone lang-icon metadata row).
+		expect(lines[1]).toContain("115│ ctx");
+		expect(lines.filter(line => line.includes("hunk"))).toHaveLength(1);
+	});
+
+	it("renders completed edit gutters without inherited frame padding", async () => {
+		const uiTheme = await getUiTheme();
+		const component = editToolRenderer.renderResult(
+			{
+				content: [{ type: "text", text: "Updated demo.ts" }],
+				details: {
+					diff: "+1│const renamedIdentifier = computeValueFromSomeVeryLongInputName();",
+					op: "update",
+				},
+			},
+			{ expanded: false, isPartial: false, renderContext: { editMode: "hashline" } },
+			uiTheme,
+			{ file_path: "demo.ts" },
+		);
+
+		const lines = component.render(48).map(line => Bun.stripANSI(line));
+		expect(lines.every(line => visibleWidth(line) === 48)).toBe(true);
+		expect(lines[1]).toStartWith("│+1│");
+		expect(lines[1]).not.toStartWith("│ +1│");
+	});
 });
diff --git a/packages/coding-agent/test/tools/eval-fallback.test.ts b/packages/coding-agent/test/tools/eval-fallback.test.ts
index 112fd43f8..083b5faa6 100644
--- a/packages/coding-agent/test/tools/eval-fallback.test.ts
+++ b/packages/coding-agent/test/tools/eval-fallback.test.ts
@@ -1,10 +1,21 @@
-import { afterEach, describe, expect, it, vi } from "bun:test";
+import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
 import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
 import * as evalIndex from "@oh-my-pi/pi-coding-agent/eval";
 import * as pyKernel from "@oh-my-pi/pi-coding-agent/eval/py/kernel";
 import type { ToolSession } from "@oh-my-pi/pi-coding-agent/tools";
 import { EvalTool } from "@oh-my-pi/pi-coding-agent/tools/eval";
+import { resolveEvalBackends } from "@oh-my-pi/pi-coding-agent/tools/eval-backends";
 
+let originalPiPy: string | undefined;
+let originalPiJs: string | undefined;
+
+function restoreEnv(name: "PI_PY" | "PI_JS", value: string | undefined): void {
+	if (value === undefined) {
+		delete Bun.env[name];
+		return;
+	}
+	Bun.env[name] = value;
+}
 function makeSession(settings = Settings.isolated()): ToolSession {
 	return {
 		cwd: "/tmp/eval-test",
@@ -29,8 +40,17 @@ const mockResult = {
 };
 
 describe("EvalTool language dispatch", () => {
+	beforeEach(() => {
+		originalPiPy = Bun.env.PI_PY;
+		originalPiJs = Bun.env.PI_JS;
+		delete Bun.env.PI_PY;
+		delete Bun.env.PI_JS;
+	});
+
 	afterEach(() => {
 		vi.restoreAllMocks();
+		restoreEnv("PI_PY", originalPiPy);
+		restoreEnv("PI_JS", originalPiJs);
 	});
 
 	it('dispatches to the JS backend when cell.language === "js"', async () => {
@@ -100,4 +120,26 @@ describe("EvalTool language dispatch", () => {
 			}),
 		).rejects.toThrow(/eval\.js = false/);
 	});
+
+	it("uses settings for eval backends whose env flag is unset", () => {
+		Bun.env.PI_PY = "1";
+		const settings = Settings.isolated();
+		settings.set("eval.py", false);
+		settings.set("eval.js", false);
+
+		expect(resolveEvalBackends(makeSession(settings))).toEqual({ python: true, js: false });
+	});
+
+	it("lets PI_JS disable js execution even when eval.js is enabled", async () => {
+		Bun.env.PI_JS = "0";
+		const settings = Settings.isolated();
+		settings.set("eval.js", true);
+		const tool = new EvalTool(makeSession(settings));
+
+		await expect(
+			tool.execute("call-js-env-disabled", {
+				cells: [{ language: "js", code: "const x = 1;" }],
+			}),
+		).rejects.toThrow(/PI_JS=0/);
+	});
 });
diff --git a/packages/coding-agent/test/tools/fetch-binary-dispatch.test.ts b/packages/coding-agent/test/tools/fetch-binary-dispatch.test.ts
new file mode 100644
index 000000000..cab602c17
--- /dev/null
+++ b/packages/coding-agent/test/tools/fetch-binary-dispatch.test.ts
@@ -0,0 +1,217 @@
+import { Database } from "bun:sqlite";
+import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
+import * as fs from "node:fs";
+import * as os from "node:os";
+import * as path from "node:path";
+import type { ImageContent, TextContent } from "@oh-my-pi/pi-ai";
+import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import type { ToolSession } from "@oh-my-pi/pi-coding-agent/tools";
+import { ReadTool } from "@oh-my-pi/pi-coding-agent/tools/read";
+import * as scrapers from "@oh-my-pi/pi-coding-agent/web/scrapers/types";
+import * as scraperUtils from "@oh-my-pi/pi-coding-agent/web/scrapers/utils";
+import { Snowflake } from "@oh-my-pi/pi-utils";
+import { zipSync } from "fflate";
+
+function makeSession(testDir: string): ToolSession {
+	const sessionFile = path.join(testDir, "session.jsonl");
+	const artifactsDir = sessionFile.slice(0, -6);
+	let nextArtifactId = 0;
+	return {
+		cwd: testDir,
+		hasUI: false,
+		getSessionFile: () => sessionFile,
+		getArtifactsDir: () => artifactsDir,
+		getSessionSpawns: () => null,
+		allocateOutputArtifact: async toolType => {
+			const id = String(nextArtifactId++);
+			return { id, path: path.join(artifactsDir, `${id}.${toolType}.log`) };
+		},
+		settings: Settings.isolated({ "fetch.enabled": true }),
+	};
+}
+
+function stubUrlBytes(bytes: Uint8Array, contentType: string, contentDisposition?: string) {
+	const decoded = Buffer.from(bytes).toString("utf-8");
+	vi.spyOn(scrapers, "loadPage").mockImplementation(async requestedUrl => ({
+		ok: true,
+		status: 200,
+		finalUrl: requestedUrl,
+		contentType,
+		content: decoded,
+	}));
+	return vi.spyOn(scraperUtils, "fetchBinary").mockImplementation(async () => ({
+		ok: true,
+		buffer: bytes,
+		contentDisposition,
+	}));
+}
+
+function stubUrlText(body: string, contentType: string) {
+	vi.spyOn(scrapers, "loadPage").mockImplementation(async requestedUrl => ({
+		ok: true,
+		status: 200,
+		finalUrl: requestedUrl,
+		contentType,
+		content: body,
+	}));
+	return vi.spyOn(scraperUtils, "fetchBinary").mockImplementation(async () => {
+		throw new Error("unexpected binary fetch");
+	});
+}
+
+function textOutput(result: { content: Array<TextContent | ImageContent> }): string {
+	return result.content
+		.filter((content): content is TextContent => content.type === "text")
+		.map(content => content.text)
+		.join("\n");
+}
+
+async function createSqliteFixtureBytes(testDir: string): Promise<Uint8Array> {
+	const dbPath = path.join(testDir, `url-fixture-${Snowflake.next()}.sqlite`);
+	const db = new Database(dbPath);
+	try {
+		db.exec(`
+			CREATE TABLE notes (id INTEGER PRIMARY KEY, title TEXT NOT NULL);
+			INSERT INTO notes (title) VALUES ('alpha'), ('beta');
+		`);
+	} finally {
+		db.close();
+	}
+	return Bun.file(dbPath).bytes();
+}
+
+function createNotebookFixtureBytes(): Uint8Array {
+	return Buffer.from(
+		JSON.stringify({
+			cells: [
+				{ cell_type: "markdown", metadata: {}, source: ["# Remote notebook\n", "Rendered as editable cells"] },
+				{ cell_type: "code", execution_count: null, metadata: {}, outputs: [], source: ["answer = 42\n"] },
+			],
+			metadata: {},
+			nbformat: 4,
+			nbformat_minor: 5,
+		}),
+	);
+}
+
+function uniqueUrl(name: string, extension: string): string {
+	return `https://example.com/${name}-${Snowflake.next()}${extension}`;
+}
+
+describe("read URL binary dispatch", () => {
+	let testDir: string;
+
+	beforeEach(() => {
+		testDir = path.join(os.tmpdir(), `fetch-binary-dispatch-${Snowflake.next()}`);
+		fs.mkdirSync(testDir, { recursive: true });
+	});
+
+	afterEach(() => {
+		vi.restoreAllMocks();
+		fs.rmSync(testDir, { recursive: true, force: true });
+	});
+
+	it("lists a remote zip instead of dumping decoded bytes", async () => {
+		const zipBytes = zipSync({
+			"root.txt": Buffer.from("root file\n"),
+			"nested/data.txt": Buffer.from("nested file\n"),
+		});
+		const url = uniqueUrl("archive", ".zip");
+		stubUrlBytes(zipBytes, "application/octet-stream");
+
+		const tool = new ReadTool(makeSession(testDir));
+		const result = await tool.execute("read-url-zip", { path: url });
+		const text = textOutput(result);
+
+		expect(result.details?.method).toBe("archive");
+		expect(text).toContain("Method: archive");
+		expect(text).toContain("root.txt");
+		expect(text).toContain("nested/");
+		expect(text).not.toContain("PK\u0003\u0004");
+		expect(text).not.toContain("�");
+	});
+
+	it("returns a metadata notice when a hinted binary refetch fails", async () => {
+		const url = uniqueUrl("oversized", ".zip");
+		vi.spyOn(scrapers, "loadPage").mockImplementation(async requestedUrl => ({
+			ok: true,
+			status: 200,
+			finalUrl: requestedUrl,
+			contentType: "application/octet-stream",
+			content: "PK\u0003\u0004\u0000\u0001",
+		}));
+		vi.spyOn(scraperUtils, "fetchBinary").mockResolvedValue({
+			ok: false,
+			error: "content-length 52428801 exceeds 52428800",
+		});
+
+		const tool = new ReadTool(makeSession(testDir));
+		const result = await tool.execute("read-url-zip-too-large", { path: url });
+		const text = textOutput(result);
+
+		expect(result.details?.method).toBe("binary");
+		expect(result.details?.notes).toContain("Binary fetch failed: content-length 52428801 exceeds 52428800");
+		expect(text).toContain("[Binary content: application/octet-stream");
+		expect(text).not.toContain("PK\u0003\u0004");
+	});
+
+	it("renders a remote sqlite database through the sqlite reader", async () => {
+		const sqliteBytes = await createSqliteFixtureBytes(testDir);
+		const url = uniqueUrl("data", ".db");
+		stubUrlBytes(sqliteBytes, "application/octet-stream");
+
+		const tool = new ReadTool(makeSession(testDir));
+		const result = await tool.execute("read-url-sqlite", { path: url });
+		const text = textOutput(result);
+
+		expect(result.details?.method).toBe("sqlite");
+		expect(text).toContain("Method: sqlite");
+		expect(text).toContain("notes (2 rows)");
+	});
+
+	it("renders a remote notebook as editable cells", async () => {
+		const notebookBytes = createNotebookFixtureBytes();
+		const url = uniqueUrl("notebook", ".ipynb");
+		stubUrlBytes(notebookBytes, "application/octet-stream");
+
+		const tool = new ReadTool(makeSession(testDir));
+		const result = await tool.execute("read-url-notebook", { path: url });
+		const text = textOutput(result);
+
+		expect(result.details?.method).toBe("notebook");
+		expect(text).toContain("Method: notebook");
+		expect(text).toContain("# %% [markdown] cell:0");
+		expect(text).toContain("# %% [code] cell:1");
+		expect(text).toContain("answer = 42");
+	});
+
+	it("returns a metadata notice for unrenderable binary bytes", async () => {
+		const binaryBytes = new Uint8Array([0, 1, 2, 3, 255, 254, 253, 0, 7, 8]);
+		const url = uniqueUrl("payload", ".bin");
+		stubUrlBytes(binaryBytes, "application/octet-stream");
+
+		const tool = new ReadTool(makeSession(testDir));
+		const result = await tool.execute("read-url-binary", { path: url });
+		const text = textOutput(result);
+
+		expect(result.details?.method).toBe("binary");
+		expect(text).toContain("Method: binary");
+		expect(text).toContain("[Binary content: application/octet-stream");
+		expect(text).not.toContain("\u0000");
+		expect(text).not.toContain("�");
+	});
+
+	it("leaves valid UTF-8 octet-stream payloads on the text path", async () => {
+		const url = uniqueUrl("plain", ".txt");
+		const fetchBinarySpy = stubUrlText("plain UTF-8 text\nsecond line", "application/octet-stream");
+
+		const tool = new ReadTool(makeSession(testDir));
+		const result = await tool.execute("read-url-text-octet", { path: url });
+		const text = textOutput(result);
+
+		expect(result.details?.method).toBe("raw");
+		expect(text).toContain("plain UTF-8 text");
+		expect(text).toContain("second line");
+		expect(fetchBinarySpy).not.toHaveBeenCalled();
+	});
+});
diff --git a/packages/coding-agent/test/tools/fetch-jina-stall.test.ts b/packages/coding-agent/test/tools/fetch-jina-stall.test.ts
index 03f33b1dc..4b150f16a 100644
--- a/packages/coding-agent/test/tools/fetch-jina-stall.test.ts
+++ b/packages/coding-agent/test/tools/fetch-jina-stall.test.ts
@@ -14,7 +14,9 @@ describe("renderHtmlToText: jina stall does not starve local fallbacks (#1449)",
 	});
 
 	it("falls back to native renderer when jina hangs until aborted", async () => {
-		const settings = Settings.isolated({ "providers.parallelFetch": false });
+		// Force jina first so the stall path is actually exercised before the
+		// native fallback runs.
+		const settings = Settings.isolated({ "providers.fetch": "jina" });
 		// Substantive HTML so the native converter produces >100 chars and
 		// `isLowQualityOutput` does not reject it.
 		const paragraphs = Array.from(
@@ -45,9 +47,12 @@ describe("renderHtmlToText: jina stall does not starve local fallbacks (#1449)",
 		});
 
 		const started = Date.now();
-		// `timeout: 2` keeps the overall budget tight — the test must complete
-		// within ~2s even though Jina would otherwise hang for the full budget.
-		const result = await renderHtmlToText("https://example.com/article", html, 2, settings, undefined, null);
+		// Tight 300ms reader-mode budget. Jina would otherwise hang forever, but
+		// the remote sub-budget (min(timeout*1000, REMOTE_READER_MAX_MS)) aborts
+		// the stalled request so the local native renderer still runs. Kept small
+		// so the test exercises the same abort path without burning real
+		// wall-clock time waiting out the stall.
+		const result = await renderHtmlToText("https://example.com/article", html, 0.3, settings, undefined, null);
 		const elapsedMs = Date.now() - started;
 
 		expect(result.ok).toBe(true);
@@ -56,14 +61,14 @@ describe("renderHtmlToText: jina stall does not starve local fallbacks (#1449)",
 		// If trafilatura or lynx happened to succeed first, that's also a valid
 		// non-aborted outcome.
 		expect(["native", "trafilatura", "lynx"]).toContain(result.method);
-		// Must finish well before the overall budget elapses: the remote
-		// sub-budget caps Jina at min(timeout, REMOTE_READER_MAX_MS), so the
-		// remaining ~1s of the 2s budget is enough for the native renderer.
-		expect(elapsedMs).toBeLessThan(2_500);
+		// Must finish shortly after the 300ms budget aborts the stalled Jina
+		// request — never anywhere near an unbounded hang. The generous bound
+		// absorbs scheduler jitter under full-suite parallelism.
+		expect(elapsedMs).toBeLessThan(1_500);
 	});
 
 	it("re-throws when the user signal is aborted, not when Jina sub-budget expires", async () => {
-		const settings = Settings.isolated({ "providers.parallelFetch": false });
+		const settings = Settings.isolated({ "providers.fetch": "jina" });
 		const html = "<html><body><p>short</p></body></html>";
 
 		using _hook = hookFetch((_input, init, _next) => {
diff --git a/packages/coding-agent/test/tools/fetch-kagi-toggle.test.ts b/packages/coding-agent/test/tools/fetch-kagi-toggle.test.ts
index 0b3dfea14..facfef759 100644
--- a/packages/coding-agent/test/tools/fetch-kagi-toggle.test.ts
+++ b/packages/coding-agent/test/tools/fetch-kagi-toggle.test.ts
@@ -34,7 +34,7 @@ describe("read tool URL selector shorthands", () => {
 		fs.rmSync(testDir, { recursive: true, force: true });
 	});
 
-	const createSession = (): ToolSession => {
+	const createSession = (settingsOverrides: Partial<Record<SettingPath, unknown>> = {}): ToolSession => {
 		const sessionFile = path.join(testDir, "session.jsonl");
 		const artifactsDir = sessionFile.slice(0, -6);
 		let nextArtifactId = 0;
@@ -53,6 +53,7 @@ describe("read tool URL selector shorthands", () => {
 			},
 			settings: Settings.isolated({
 				"fetch.enabled": true,
+				...settingsOverrides,
 			}),
 		};
 	};
@@ -575,9 +576,9 @@ describe("read tool URL handling", () => {
 		void missingSystemPython;
 		void hook;
 	});
-	it("prefers Parallel extract before other HTML renderers when configured", async () => {
+	it("prefers Parallel extract first when providers.fetch is set to parallel", async () => {
 		process.env.PARALLEL_API_KEY = "test-parallel-key";
-		const session = createSession();
+		const session = createSession({ "providers.fetch": "parallel" });
 		const tool = new ReadTool(session);
 		const pageUrl = "https://example.com/parallel-page";
 		const pageHtml = "<html><body><main><h1>Parallel Page</h1></main></body></html>";
diff --git a/packages/coding-agent/test/tools/find-renderer.test.ts b/packages/coding-agent/test/tools/find-renderer.test.ts
new file mode 100644
index 000000000..c0b960b92
--- /dev/null
+++ b/packages/coding-agent/test/tools/find-renderer.test.ts
@@ -0,0 +1,28 @@
+import { describe, expect, it } from "bun:test";
+import { sanitizeText } from "@oh-my-pi/pi-utils";
+import { getThemeByName } from "../../src/modes/theme/theme";
+import { findToolRenderer } from "../../src/tools/find";
+
+describe("findToolRenderer", () => {
+	it("indents inline find output and avoids accent-colored success headers", async () => {
+		const theme = await getThemeByName("dark");
+		expect(theme).toBeDefined();
+		const uiTheme = theme!;
+		const result = {
+			content: [{ type: "text", text: "" }],
+			details: {
+				fileCount: 2,
+				files: ["src/a.ts", "src/b.ts"],
+			},
+		};
+
+		const renderedLines = findToolRenderer
+			.renderResult(result as never, { expanded: true, isPartial: false }, uiTheme, { paths: "src/**/*.ts" })
+			.render(240);
+		const plainLines = sanitizeText(renderedLines.join("\n")).split("\n");
+
+		expect(plainLines.every(line => line.startsWith(" "))).toBe(true);
+		expect(renderedLines[0]).not.toContain(uiTheme.fg("accent", uiTheme.symbol("icon.search")));
+		expect(renderedLines[0]).not.toContain(uiTheme.fg("accent", "Find"));
+	});
+});
diff --git a/packages/coding-agent/test/tools/gh.test.ts b/packages/coding-agent/test/tools/gh.test.ts
index 2deb69989..b1776f807 100644
--- a/packages/coding-agent/test/tools/gh.test.ts
+++ b/packages/coding-agent/test/tools/gh.test.ts
@@ -1,4 +1,4 @@
-import { afterEach, describe, expect, it, vi } from "bun:test";
+import { afterAll, afterEach, beforeAll, describe, expect, it, vi } from "bun:test";
 import * as fs from "node:fs/promises";
 import * as os from "node:os";
 import * as path from "node:path";
@@ -85,15 +85,24 @@ function runGit(cwd: string, args: string[]): string {
 	return new TextDecoder().decode(result.stdout).trim();
 }
 
-async function createPrFixture(): Promise<{
+interface PrFixture {
 	baseDir: string;
 	repoRoot: string;
 	originBare: string;
 	forkBare: string;
 	headRefName: string;
 	headRefOid: string;
-}> {
-	const baseDir = await fs.mkdtemp(path.join(os.tmpdir(), "gh-pr-tool-"));
+}
+
+// Building the fixture costs ~16 real `git` subprocess spawns (~200ms). Six
+// tests need it, so we build it ONCE as an immutable template in `beforeAll`
+// and materialize per-test copies via `fs.cp` (~12ms). Each copy is a fully
+// independent repo tree, so the mutating tests (worktree checkout, config
+// writes, extra branches) can't contaminate each other.
+let prFixtureTemplate: PrFixture | null = null;
+
+async function buildPrFixtureTemplate(): Promise<PrFixture> {
+	const baseDir = await fs.mkdtemp(path.join(os.tmpdir(), "gh-pr-tool-template-"));
 	const repoRoot = path.join(baseDir, "repo");
 	const originBare = path.join(baseDir, "origin.git");
 	const forkBare = path.join(baseDir, "fork.git");
@@ -119,13 +128,32 @@ async function createPrFixture(): Promise<{
 	runGit(repoRoot, ["push", "-u", "forksrc", headRefName]);
 	runGit(repoRoot, ["checkout", "main"]);
 
+	return { baseDir, repoRoot, originBare, forkBare, headRefName, headRefOid };
+}
+
+async function createPrFixture(): Promise<PrFixture> {
+	const template = prFixtureTemplate;
+	if (!template) throw new Error("PR fixture template was not built (missing beforeAll)");
+
+	const baseDir = await fs.mkdtemp(path.join(os.tmpdir(), "gh-pr-tool-"));
+	const repoRoot = path.join(baseDir, "repo");
+	const originBare = path.join(baseDir, "origin.git");
+	const forkBare = path.join(baseDir, "fork.git");
+
+	await fs.cp(template.baseDir, baseDir, { recursive: true });
+	// Remote URLs in the copied repo still point at the template's absolute
+	// `origin.git`/`fork.git`. Repoint them at this copy so pushes/fetches stay
+	// isolated and `remote get-url` assertions match the returned paths.
+	runGit(repoRoot, ["remote", "set-url", "origin", originBare]);
+	runGit(repoRoot, ["remote", "set-url", "forksrc", forkBare]);
+
 	return {
 		baseDir,
 		repoRoot,
 		originBare,
 		forkBare,
-		headRefName,
-		headRefOid,
+		headRefName: template.headRefName,
+		headRefOid: template.headRefOid,
 	};
 }
 
@@ -210,6 +238,17 @@ describe("parsePrUnifiedDiff", () => {
 });
 
 describe("github tool", () => {
+	beforeAll(async () => {
+		prFixtureTemplate = await buildPrFixtureTemplate();
+	});
+
+	afterAll(async () => {
+		if (prFixtureTemplate) {
+			await fs.rm(prFixtureTemplate.baseDir, { recursive: true, force: true });
+			prFixtureTemplate = null;
+		}
+	});
+
 	afterEach(() => {
 		vi.useRealTimers();
 		vi.restoreAllMocks();
diff --git a/packages/coding-agent/test/tools/grouped-file-output.test.ts b/packages/coding-agent/test/tools/grouped-file-output.test.ts
new file mode 100644
index 000000000..d9cf45db2
--- /dev/null
+++ b/packages/coding-agent/test/tools/grouped-file-output.test.ts
@@ -0,0 +1,118 @@
+import { describe, expect, it } from "bun:test";
+import { formatFindGroupedOutput } from "../../src/tools/find";
+import { classifyGroupedLines, formatGroupedFiles, groupLineIndicesByBlank } from "../../src/tools/grouped-file-output";
+
+describe("formatFindGroupedOutput", () => {
+	it("folds a shared absolute prefix into one heading and nests the rest", () => {
+		const output = formatFindGroupedOutput([
+			"/Users/me/proj/shared/wasm/llvm.hpp",
+			"/Users/me/proj/shared/wasm/vm.hpp",
+			"/Users/me/proj/shared/xstd.hpp",
+			"/Users/me/proj/shared/apollo/details/hash.hpp",
+			"/Users/me/proj/flash/main.cpp",
+		]);
+
+		expect(output).toBe(
+			[
+				"# /Users/me/proj/",
+				"## shared/",
+				"xstd.hpp",
+				"### wasm/",
+				"llvm.hpp",
+				"vm.hpp",
+				"### apollo/details/",
+				"hash.hpp",
+				"## flash/",
+				"main.cpp",
+			].join("\n"),
+		);
+	});
+
+	it("lists a directory's own files before its subdirectories", () => {
+		const output = formatFindGroupedOutput(["pkg/sub/deep.txt", "pkg/top.txt"]);
+		// `top.txt` is a direct child of pkg; `sub/` is a subdirectory. Files first.
+		expect(output).toBe(["# pkg/", "top.txt", "## sub/", "deep.txt"].join("\n"));
+	});
+
+	it("emits a single root-level file with no directory heading", () => {
+		expect(formatFindGroupedOutput(["single.txt"])).toBe("single.txt");
+	});
+
+	it("keeps matched directories (trailing slash) as headings", () => {
+		expect(formatFindGroupedOutput(["alpha/tests/", "beta/tests/"])).toBe(
+			["# alpha/tests/", "# beta/tests/"].join("\n"),
+		);
+	});
+});
+
+describe("formatGroupedFiles", () => {
+	it("nests subdirectories with deeper headings and blank-separates top groups", () => {
+		const { model } = formatGroupedFiles(["pkg/ai/CHANGELOG.md", "pkg/ai/src/util/x.ts", "README.md"], file => ({
+			modelLines: [`  ${file}`],
+			headerSuffix: " (1)",
+		}));
+
+		expect(model).toEqual([
+			"# README.md (1)",
+			"  README.md",
+			"",
+			"# pkg/ai/",
+			"## CHANGELOG.md (1)",
+			"  pkg/ai/CHANGELOG.md",
+			"",
+			"## src/util/",
+			"### x.ts (1)",
+			"  pkg/ai/src/util/x.ts",
+		]);
+	});
+
+	it("omits skipped files and their now-empty directories", () => {
+		const { model } = formatGroupedFiles(["a/keep.ts", "a/drop.ts"], file => ({
+			modelLines: [`  ${file}`],
+			skip: file.endsWith("drop.ts"),
+		}));
+		expect(model).toEqual(["# a/", "## keep.ts", "  a/keep.ts"]);
+	});
+});
+
+describe("classifyGroupedLines", () => {
+	it("reconstructs absolute paths across a nested directory stack", () => {
+		const lines = ["# pkg/ai/", "## CHANGELOG.md", "  match", "## src/util/", "### x.ts", "*12│const y = 1;"];
+		const ctx = classifyGroupedLines(lines, "/repo");
+
+		expect(ctx[0]).toMatchObject({ kind: "dir", headerPath: "/repo/pkg/ai" });
+		expect(ctx[1]).toMatchObject({ kind: "file", headerPath: "/repo/pkg/ai/CHANGELOG.md" });
+		expect(ctx[2]).toMatchObject({ kind: "content", filePath: "/repo/pkg/ai/CHANGELOG.md" });
+		// `src/util/` is a folded subdirectory chain under `pkg/ai/`, not the root.
+		expect(ctx[3]).toMatchObject({ kind: "dir", headerPath: "/repo/pkg/ai/src/util" });
+		expect(ctx[4]).toMatchObject({ kind: "file", headerPath: "/repo/pkg/ai/src/util/x.ts" });
+		expect(ctx[5]).toMatchObject({ kind: "content", filePath: "/repo/pkg/ai/src/util/x.ts" });
+	});
+
+	it("keeps an absolute folded prefix instead of joining it onto the search base", () => {
+		const ctx = classifyGroupedLines(["# /outside/dir/", "## file.txt"], "/repo");
+		expect(ctx[0]).toMatchObject({ kind: "dir", headerPath: "/outside/dir" });
+		expect(ctx[1]).toMatchObject({ kind: "file", headerPath: "/outside/dir/file.txt" });
+	});
+
+	it("links body lines before any header to the single-file search base", () => {
+		const ctx = classifyGroupedLines(["*7│needle();"], "/repo/file.ts");
+		expect(ctx[0]).toMatchObject({ kind: "content", filePath: "/repo/file.ts" });
+	});
+
+	it("flags url-like headers for caller-side resolution without a filesystem path", () => {
+		const ctx = classifyGroupedLines(["# omp://docs/", "  body"], "/repo");
+		expect(ctx[0]).toMatchObject({ kind: "file", isUrl: true });
+		expect(ctx[0]?.headerPath).toBeUndefined();
+	});
+});
+
+describe("groupLineIndicesByBlank", () => {
+	it("breaks on blank-line runs and keeps original indices", () => {
+		expect(groupLineIndicesByBlank(["a", "b", "", "", "c"])).toEqual([[0, 1], [4]]);
+	});
+
+	it("returns a single group of non-empty lines when no blanks are present", () => {
+		expect(groupLineIndicesByBlank(["a", "b", "c"])).toEqual([[0, 1, 2]]);
+	});
+});
diff --git a/packages/coding-agent/test/tools/image-gen.test.ts b/packages/coding-agent/test/tools/image-gen.test.ts
index 74b5182db..2ecdaa8bc 100644
--- a/packages/coding-agent/test/tools/image-gen.test.ts
+++ b/packages/coding-agent/test/tools/image-gen.test.ts
@@ -62,6 +62,8 @@ describe("imageGenTool", () => {
 			modelRegistry: {
 				getApiKey: async () => "test-openai-key",
 				getApiKeyForProvider: async () => undefined,
+				authStorage: { rotateSessionCredential: async () => false },
+				resolver: () => async () => "test-openai-key",
 			} as unknown as ModelRegistry,
 			model,
 			isIdle: () => true,
@@ -126,7 +128,9 @@ describe("imageGenTool", () => {
 				getAll: () => [],
 				authStorage: {
 					hasNonEnvCredential: (provider: string) => provider === "xai-oauth",
+					rotateSessionCredential: async () => false,
 				},
+				resolver: () => async () => "test-xai-token",
 			} as unknown as ModelRegistry,
 			model: undefined,
 			isIdle: () => true,
diff --git a/packages/coding-agent/test/tools/inspect-image.test.ts b/packages/coding-agent/test/tools/inspect-image.test.ts
index fe27a38cb..8365ce9c2 100644
--- a/packages/coding-agent/test/tools/inspect-image.test.ts
+++ b/packages/coding-agent/test/tools/inspect-image.test.ts
@@ -68,6 +68,9 @@ function createSession(
 		modelRegistry: {
 			getAvailable: () => availableModels,
 			getApiKey: async () => apiKey,
+			getApiKeyForProvider: async () => apiKey,
+			authStorage: { rotateSessionCredential: async () => false },
+			resolver: () => async () => apiKey,
 		} as unknown as NonNullable<ToolSession["modelRegistry"]>,
 	};
 }
@@ -170,7 +173,7 @@ describe("InspectImageTool", () => {
 			uiTheme,
 		);
 		const callOutput = sanitizeText(callComponent.render(100).join("\n"));
-		expect(callOutput).toContain("Inspect Image");
+		expect(callOutput).toContain("Inspect");
 		expect(callOutput).toContain("Question:");
 		expect(callOutput).toContain("What error text is visible?");
 
@@ -188,7 +191,7 @@ describe("InspectImageTool", () => {
 			{ path: "/tmp/screenshot.png", question: "What error text is visible?" },
 		);
 		const resultOutput = sanitizeText(resultComponent.render(100).join("\n"));
-		expect(resultOutput).toContain("Inspect Image");
+		expect(resultOutput).toContain("Inspect");
 		expect(resultOutput).toContain("image/png");
 		expect(resultOutput).toContain("Question:");
 		expect(resultOutput).toContain("What error text is visible?");
diff --git a/packages/coding-agent/test/tools/lsp-regressions.test.ts b/packages/coding-agent/test/tools/lsp-regressions.test.ts
index b15bdbba4..f246f5545 100644
--- a/packages/coding-agent/test/tools/lsp-regressions.test.ts
+++ b/packages/coding-agent/test/tools/lsp-regressions.test.ts
@@ -145,6 +145,376 @@ process.abort();
 		}
 	});
 
+	it("advertises workspace folder support during LSP initialization", async () => {
+		const tempDir = TempDir.createSync("@omp-lsp-workspace-folders-");
+		try {
+			const initPath = path.join(tempDir.path(), "initialize.json");
+			const serverPath = path.join(tempDir.path(), "server.ts");
+			await Bun.write(
+				serverPath,
+				`
+const initPath = process.argv[2];
+const decoder = new TextDecoder();
+let buffer = "";
+
+function send(message) {
+	const content = JSON.stringify(message);
+	process.stdout.write(\`Content-Length: \${Buffer.byteLength(content, "utf8")}\\r\\n\\r\\n\${content}\`);
+}
+
+for await (const chunk of Bun.stdin.stream()) {
+	buffer += decoder.decode(chunk, { stream: true });
+	while (true) {
+		const headerEnd = buffer.indexOf("\\r\\n\\r\\n");
+		if (headerEnd === -1) break;
+
+		const header = buffer.slice(0, headerEnd);
+		const match = /Content-Length: (\\d+)/i.exec(header);
+		if (!match) process.exit(2);
+
+		const contentLength = Number(match[1]);
+		const contentStart = headerEnd + 4;
+		const contentEnd = contentStart + contentLength;
+		if (buffer.length < contentEnd) break;
+
+		const message = JSON.parse(buffer.slice(contentStart, contentEnd));
+		buffer = buffer.slice(contentEnd);
+
+		if (message.method === "initialize") {
+			await Bun.write(initPath, JSON.stringify(message.params));
+			send({ jsonrpc: "2.0", id: message.id, result: { capabilities: {} } });
+		} else if (message.method === "shutdown") {
+			send({ jsonrpc: "2.0", id: message.id, result: null });
+		} else if (message.method === "exit") {
+			process.exit(0);
+		}
+	}
+}
+`,
+			);
+
+			const server: ServerConfig = {
+				command: process.execPath,
+				args: [serverPath, initPath],
+				fileTypes: ["rs"],
+				rootMarkers: [],
+			};
+
+			await lspClient.getOrCreateClient(server, tempDir.path(), 1_000);
+			const params = (await Bun.file(initPath).json()) as {
+				capabilities?: { workspace?: { workspaceFolders?: unknown } };
+				workspaceFolders?: unknown;
+			};
+
+			expect(params.capabilities?.workspace?.workspaceFolders).toBe(true);
+			expect(params.workspaceFolders).toEqual([
+				{ uri: fileToUri(tempDir.path()), name: path.basename(tempDir.path()) },
+			]);
+		} finally {
+			await lspClient.shutdownAll();
+			tempDir.removeSync();
+		}
+	});
+
+	it("answers workspace/workspaceFolders requests with the current folder set", async () => {
+		const tempDir = TempDir.createSync("@omp-lsp-workspace-folders-request-");
+		try {
+			const responsePath = path.join(tempDir.path(), "folders-response.json");
+			const serverPath = path.join(tempDir.path(), "server.ts");
+			await Bun.write(
+				serverPath,
+				`
+const responsePath = process.argv[2];
+const decoder = new TextDecoder();
+let buffer = "";
+
+function send(message) {
+	const content = JSON.stringify(message);
+	process.stdout.write(\`Content-Length: \${Buffer.byteLength(content, "utf8")}\\r\\n\\r\\n\${content}\`);
+}
+
+for await (const chunk of Bun.stdin.stream()) {
+	buffer += decoder.decode(chunk, { stream: true });
+	while (true) {
+		const headerEnd = buffer.indexOf("\\r\\n\\r\\n");
+		if (headerEnd === -1) break;
+
+		const header = buffer.slice(0, headerEnd);
+		const match = /Content-Length: (\\d+)/i.exec(header);
+		if (!match) process.exit(2);
+
+		const contentLength = Number(match[1]);
+		const contentStart = headerEnd + 4;
+		const contentEnd = contentStart + contentLength;
+		if (buffer.length < contentEnd) break;
+
+		const message = JSON.parse(buffer.slice(contentStart, contentEnd));
+		buffer = buffer.slice(contentEnd);
+
+		if (message.method === "initialize") {
+			send({ jsonrpc: "2.0", id: message.id, result: { capabilities: {} } });
+			send({ jsonrpc: "2.0", id: 9001, method: "workspace/workspaceFolders" });
+		} else if (message.id === 9001) {
+			await Bun.write(responsePath, JSON.stringify(message));
+		} else if (message.method === "shutdown") {
+			send({ jsonrpc: "2.0", id: message.id, result: null });
+		} else if (message.method === "exit") {
+			process.exit(0);
+		}
+	}
+}
+`,
+			);
+
+			const server: ServerConfig = {
+				command: process.execPath,
+				args: [serverPath, responsePath],
+				fileTypes: ["rs"],
+				rootMarkers: [],
+			};
+
+			await lspClient.getOrCreateClient(server, tempDir.path(), 1_000);
+			const deadline = Date.now() + 1_000;
+			while (!fs.existsSync(responsePath) && Date.now() < deadline) {
+				await Bun.sleep(20);
+			}
+			const response = (await Bun.file(responsePath).json()) as {
+				error?: { code: number };
+				result?: unknown;
+			};
+
+			expect(response.error).toBeUndefined();
+			expect(response.result).toEqual([{ uri: fileToUri(tempDir.path()), name: path.basename(tempDir.path()) }]);
+		} finally {
+			await lspClient.shutdownAll();
+			tempDir.removeSync();
+		}
+	});
+
+	it("opens rust-analyzer Cargo workspace files before polling workspace readiness", async () => {
+		const tempDir = TempDir.createSync("@omp-lsp-rust-workspace-");
+		try {
+			const sourcePath = path.join(tempDir.path(), "src", "main.rs");
+			const serverPath = path.join(tempDir.path(), "server.ts");
+			const eventLogPath = path.join(tempDir.path(), "events.log");
+			const statusCountPath = path.join(tempDir.path(), "status-count.txt");
+			await Bun.write(path.join(tempDir.path(), "Cargo.toml"), '[package]\nname = "fixture"\nversion = "0.0.0"\n');
+			await Bun.write(sourcePath, "fn greet() {}\nfn main() { greet(); }\n");
+			await Bun.write(
+				serverPath,
+				`
+const eventLogPath = process.argv[2];
+const statusCountPath = process.argv[3];
+const definitionUri = process.argv[4];
+const decoder = new TextDecoder();
+let buffer = "";
+let statusRequests = 0;
+let eventLog = "";
+
+function send(message) {
+	const content = JSON.stringify(message);
+	process.stdout.write(\`Content-Length: \${Buffer.byteLength(content, "utf8")}\\r\\n\\r\\n\${content}\`);
+}
+
+for await (const chunk of Bun.stdin.stream()) {
+	buffer += decoder.decode(chunk, { stream: true });
+	while (true) {
+		const headerEnd = buffer.indexOf("\\r\\n\\r\\n");
+		if (headerEnd === -1) break;
+
+		const header = buffer.slice(0, headerEnd);
+		const match = /Content-Length: (\\d+)/i.exec(header);
+		if (!match) process.exit(2);
+
+		const contentLength = Number(match[1]);
+		const contentStart = headerEnd + 4;
+		const contentEnd = contentStart + contentLength;
+		if (buffer.length < contentEnd) break;
+
+		const message = JSON.parse(buffer.slice(contentStart, contentEnd));
+		buffer = buffer.slice(contentEnd);
+
+		if (message.method === "initialize") {
+			send({ jsonrpc: "2.0", id: message.id, result: { capabilities: { definitionProvider: true } } });
+			send({ jsonrpc: "2.0", method: "$/progress", params: { token: "workspace", value: { kind: "begin" } } });
+			send({ jsonrpc: "2.0", method: "$/progress", params: { token: "workspace", value: { kind: "end" } } });
+		} else if (message.method === "rust-analyzer/analyzerStatus") {
+			statusRequests++;
+			eventLog += "status\\n";
+			await Bun.write(eventLogPath, eventLog);
+			await Bun.write(statusCountPath, String(statusRequests));
+			if (statusRequests === 1) {
+				continue;
+			}
+			const result = statusRequests < 3 ? "No workspaces" : "Workspaces:\\nLoaded 1 package across 1 workspace.";
+			send({ jsonrpc: "2.0", id: message.id, result });
+		} else if (message.method === "textDocument/didOpen") {
+			eventLog += "open\\n";
+			await Bun.write(eventLogPath, eventLog);
+		} else if (message.method === "textDocument/definition") {
+			send({
+				jsonrpc: "2.0",
+				id: message.id,
+				result: [{ uri: definitionUri, range: { start: { line: 0, character: 3 }, end: { line: 0, character: 8 } } }],
+			});
+		} else if (message.method === "shutdown") {
+			send({ jsonrpc: "2.0", id: message.id, result: null });
+		} else if (message.method === "exit") {
+			process.exit(0);
+		}
+	}
+}
+`,
+			);
+
+			const server: ServerConfig = {
+				command: "rust-analyzer",
+				resolvedCommand: process.execPath,
+				args: [serverPath, eventLogPath, statusCountPath, fileToUri(sourcePath)],
+				fileTypes: ["rs"],
+				rootMarkers: [],
+				// Shrink the workspace-ready polling window so the test exercises the
+				// timeout→retry→ready sequence without waiting out the 2s production settle.
+				// The status-request timeout stays generous to avoid racing the subprocess.
+				workspaceReadyTimings: { timeoutMs: 5_000, pollMs: 10, settleMs: 20, statusRequestTimeoutMs: 150 },
+			};
+
+			vi.spyOn(lspConfig, "loadConfig").mockReturnValue({
+				servers: { "rust-analyzer": server },
+				idleTimeoutMs: undefined,
+			});
+			vi.spyOn(lspConfig, "getServersForFile").mockReturnValue([["rust-analyzer", server]]);
+
+			const tool = new LspTool({ cwd: tempDir.path() } as ToolSession);
+			const result = await tool.execute("rust-wait-test", {
+				action: "definition",
+				file: sourcePath,
+				line: 2,
+				symbol: "greet",
+				timeout: 10,
+			});
+			const output = result.content
+				.filter(block => block.type === "text")
+				.map(block => block.text)
+				.join("\n");
+
+			const eventLog = (await Bun.file(eventLogPath).text()).trim().split("\n");
+			expect(output).toContain("Found 1 definition(s)");
+			expect(eventLog[0]).toBe("open");
+			expect(eventLog.filter(line => line === "status").length).toBeGreaterThanOrEqual(3);
+			expect(Number(await Bun.file(statusCountPath).text())).toBeGreaterThanOrEqual(3);
+		} finally {
+			vi.restoreAllMocks();
+			await lspClient.shutdownAll();
+			tempDir.removeSync();
+		}
+	});
+
+	it("skips rust-analyzer workspace polling for standalone Rust files", async () => {
+		const tempDir = TempDir.createSync("@omp-lsp-rust-standalone-");
+		try {
+			const sourcePath = path.join(tempDir.path(), "foo.rs");
+			const serverPath = path.join(tempDir.path(), "server.ts");
+			const eventLogPath = path.join(tempDir.path(), "events.log");
+			await Bun.write(sourcePath, 'fn greet() -> &\'static str { "hi" }\n');
+			await Bun.write(
+				serverPath,
+				`
+const eventLogPath = process.argv[2];
+const definitionUri = process.argv[3];
+const decoder = new TextDecoder();
+let buffer = "";
+let eventLog = "";
+
+function send(message) {
+	const content = JSON.stringify(message);
+	process.stdout.write(\`Content-Length: \${Buffer.byteLength(content, "utf8")}\\r\\n\\r\\n\${content}\`);
+}
+
+for await (const chunk of Bun.stdin.stream()) {
+	buffer += decoder.decode(chunk, { stream: true });
+	while (true) {
+		const headerEnd = buffer.indexOf("\\r\\n\\r\\n");
+		if (headerEnd === -1) break;
+
+		const header = buffer.slice(0, headerEnd);
+		const match = /Content-Length: (\\d+)/i.exec(header);
+		if (!match) process.exit(2);
+
+		const contentLength = Number(match[1]);
+		const contentStart = headerEnd + 4;
+		const contentEnd = contentStart + contentLength;
+		if (buffer.length < contentEnd) break;
+
+		const message = JSON.parse(buffer.slice(contentStart, contentEnd));
+		buffer = buffer.slice(contentEnd);
+
+		if (message.method === "initialize") {
+			send({ jsonrpc: "2.0", id: message.id, result: { capabilities: { definitionProvider: true } } });
+		} else if (message.method === "rust-analyzer/analyzerStatus") {
+			eventLog += "status\\n";
+			await Bun.write(eventLogPath, eventLog);
+			send({ jsonrpc: "2.0", id: message.id, result: "No workspaces" });
+		} else if (message.method === "textDocument/didOpen") {
+			eventLog += "open\\n";
+			await Bun.write(eventLogPath, eventLog);
+		} else if (message.method === "textDocument/definition") {
+			send({
+				jsonrpc: "2.0",
+				id: message.id,
+				result: [{ uri: definitionUri, range: { start: { line: 0, character: 3 }, end: { line: 0, character: 8 } } }],
+			});
+		} else if (message.method === "shutdown") {
+			send({ jsonrpc: "2.0", id: message.id, result: null });
+		} else if (message.method === "exit") {
+			process.exit(0);
+		}
+	}
+}
+`,
+			);
+
+			const server: ServerConfig = {
+				command: "rust-analyzer",
+				resolvedCommand: process.execPath,
+				args: [serverPath, eventLogPath, fileToUri(sourcePath)],
+				fileTypes: ["rs"],
+				rootMarkers: [],
+			};
+
+			vi.spyOn(lspConfig, "loadConfig").mockReturnValue({
+				servers: { "rust-analyzer": server },
+				idleTimeoutMs: undefined,
+			});
+			vi.spyOn(lspConfig, "getServersForFile").mockReturnValue([["rust-analyzer", server]]);
+
+			const tool = new LspTool({ cwd: tempDir.path() } as ToolSession);
+			const started = Date.now();
+			const result = await tool.execute("rust-standalone-test", {
+				action: "definition",
+				file: sourcePath,
+				line: 1,
+				symbol: "greet",
+				timeout: 10,
+			});
+			const elapsed = Date.now() - started;
+			const output = result.content
+				.filter(block => block.type === "text")
+				.map(block => block.text)
+				.join("\n");
+			const eventLog = (await Bun.file(eventLogPath).text()).trim().split("\n");
+
+			expect(output).toContain("Found 1 definition(s)");
+			expect(eventLog).toContain("open");
+			expect(eventLog).not.toContain("status");
+			expect(elapsed).toBeLessThan(2_000);
+		} finally {
+			vi.restoreAllMocks();
+			await lspClient.shutdownAll();
+			tempDir.removeSync();
+		}
+	});
+
 	it("limits glob collection to avoid large diagnostic stalls", async () => {
 		const tempDir = TempDir.createSync("@omp-lsp-glob-");
 		try {
diff --git a/packages/coding-agent/test/tools/memory-renderer.test.ts b/packages/coding-agent/test/tools/memory-renderer.test.ts
index d44f2296b..5ab918bfa 100644
--- a/packages/coding-agent/test/tools/memory-renderer.test.ts
+++ b/packages/coding-agent/test/tools/memory-renderer.test.ts
@@ -52,7 +52,6 @@ describe("retainToolRenderer", () => {
 			80,
 		);
 		const item = rendered.find(line => line.includes(bullet));
-		expect(item).toBeDefined();
 		expect(item!.length).toBeLessThanOrEqual(80);
 		expect(item).toContain("…");
 	});
diff --git a/packages/coding-agent/test/tools/plan-mode-guard-local.test.ts b/packages/coding-agent/test/tools/plan-mode-guard-local.test.ts
index 070abe496..6ac46329e 100644
--- a/packages/coding-agent/test/tools/plan-mode-guard-local.test.ts
+++ b/packages/coding-agent/test/tools/plan-mode-guard-local.test.ts
@@ -43,72 +43,50 @@ describe("resolvePlanPath local:// support", () => {
 	});
 });
 
-describe("resolvePlanPath plan-mode redirect", () => {
-	const planMode: PlanModeState = { enabled: true, planFilePath: "local://PLAN.md" };
-	const approvedPlanMode: PlanModeState = { enabled: true, planFilePath: "local://APPROVED.md" };
+describe("resolvePlanPath resolves literally (no plan-mode redirect)", () => {
+	const planMode: PlanModeState = { enabled: true, planFilePath: "local://some-plan.md" };
 
-	it("redirects bare PLAN.md to the session plan path when plan mode is active", () => {
-		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", planMode });
-		expect(resolvePlanPath(session, "PLAN.md")).toBe(path.join("/tmp/agent-artifacts", "local", "PLAN.md"));
-	});
-
-	it("redirects ./PLAN.md and absolute cwd-PLAN.md alike", () => {
+	it("resolves a bare path against cwd regardless of plan mode", () => {
 		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", cwd: "/repo", planMode });
-		const expected = path.join("/tmp/agent-artifacts", "local", "PLAN.md");
-		expect(resolvePlanPath(session, "./PLAN.md")).toBe(expected);
-		expect(resolvePlanPath(session, "/repo/PLAN.md")).toBe(expected);
-	});
-
-	it("does not redirect when plan mode is disabled", () => {
-		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", cwd: "/repo" });
 		expect(resolvePlanPath(session, "PLAN.md")).toBe(path.join("/repo", "PLAN.md"));
-	});
-
-	it("does not redirect paths whose basename differs from the plan basename", () => {
-		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", cwd: "/repo", planMode });
 		expect(resolvePlanPath(session, "src/foo.ts")).toBe(path.join("/repo", "src/foo.ts"));
 	});
 
-	it("leaves an explicit local://PLAN.md unchanged", () => {
+	it("resolves a local:// plan file to the session local root", () => {
 		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", planMode });
-		expect(resolvePlanPath(session, "local://PLAN.md")).toBe(path.join("/tmp/agent-artifacts", "local", "PLAN.md"));
-	});
-
-	it("redirects PLAN.md aliases to the active titled plan artifact", () => {
-		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", cwd: "/repo", planMode: approvedPlanMode });
-		const expected = path.join("/tmp/agent-artifacts", "local", "APPROVED.md");
-		expect(resolvePlanPath(session, "PLAN.md")).toBe(expected);
-		expect(resolvePlanPath(session, "./PLAN.md")).toBe(expected);
-		expect(resolvePlanPath(session, "/repo/PLAN.md")).toBe(expected);
-		expect(resolvePlanPath(session, "local://PLAN.md")).toBe(expected);
-	});
-});
-
-describe("enforcePlanModeWrite plan-mode redirect", () => {
-	const planMode: PlanModeState = { enabled: true, planFilePath: "local://PLAN.md" };
-	const approvedPlanMode: PlanModeState = { enabled: true, planFilePath: "local://APPROVED.md" };
-
-	it("accepts bare PLAN.md as a write to the plan file", () => {
-		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", planMode });
-		expect(() => enforcePlanModeWrite(session, "PLAN.md", { op: "update" })).not.toThrow();
-	});
-
-	it("still rejects non-plan paths in plan mode", () => {
-		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", planMode });
-		expect(() => enforcePlanModeWrite(session, "src/foo.ts", { op: "update" })).toThrow(
-			/only the plan file may be modified/,
-		);
-	});
-
-	it("accepts bare PLAN.md as a write to a titled active plan file", () => {
-		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", planMode: approvedPlanMode });
-		expect(() => enforcePlanModeWrite(session, "PLAN.md", { op: "update" })).not.toThrow();
-	});
-
-	it("rejects deletes of PLAN.md even when basename matches", () => {
-		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", planMode });
-		expect(() => enforcePlanModeWrite(session, "PLAN.md", { op: "delete" })).toThrow(
-			/Plan mode: deleting files is not allowed/,
+		expect(resolvePlanPath(session, "local://some-plan.md")).toBe(
+			path.join("/tmp/agent-artifacts", "local", "some-plan.md"),
 		);
 	});
 });
+
+describe("enforcePlanModeWrite (working tree read-only, local:// sandbox writable)", () => {
+	const planMode: PlanModeState = { enabled: true, planFilePath: "local://some-plan.md" };
+
+	it("accepts writes to any local:// file", () => {
+		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", planMode });
+		expect(() => enforcePlanModeWrite(session, "local://auth-refactor-plan.md", { op: "create" })).not.toThrow();
+		expect(() => enforcePlanModeWrite(session, "local://scratch/notes.md", { op: "update" })).not.toThrow();
+	});
+
+	it("rejects writes to the working tree", () => {
+		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", cwd: "/repo", planMode });
+		expect(() => enforcePlanModeWrite(session, "src/foo.ts", { op: "update" })).toThrow(/working tree is read-only/);
+		expect(() => enforcePlanModeWrite(session, "PLAN.md", { op: "create" })).toThrow(/working tree is read-only/);
+	});
+
+	it("rejects deletes and renames outright", () => {
+		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", planMode });
+		expect(() => enforcePlanModeWrite(session, "local://some-plan.md", { op: "delete" })).toThrow(
+			/deleting files is not allowed/,
+		);
+		expect(() => enforcePlanModeWrite(session, "local://some-plan.md", { move: "local://renamed.md" })).toThrow(
+			/renaming files is not allowed/,
+		);
+	});
+
+	it("is a no-op when plan mode is disabled", () => {
+		const session = makeSession({ artifactsDir: "/tmp/agent-artifacts", cwd: "/repo" });
+		expect(() => enforcePlanModeWrite(session, "src/foo.ts", { op: "update" })).not.toThrow();
+	});
+});
diff --git a/packages/coding-agent/test/tools/read-renderer.test.ts b/packages/coding-agent/test/tools/read-renderer.test.ts
index 001799496..0c7991235 100644
--- a/packages/coding-agent/test/tools/read-renderer.test.ts
+++ b/packages/coding-agent/test/tools/read-renderer.test.ts
@@ -1,6 +1,8 @@
 import { afterAll, afterEach, beforeAll, describe, expect, it } from "bun:test";
 import { resetSettingsForTest, Settings, settings } from "@oh-my-pi/pi-coding-agent/config/settings";
-import { getThemeByName } from "../../src/modes/theme/theme";
+import { ToolExecutionComponent } from "@oh-my-pi/pi-coding-agent/modes/components/tool-execution";
+import type { TUI } from "@oh-my-pi/pi-tui";
+import { theme as activeTheme, getThemeByName, initTheme } from "../../src/modes/theme/theme";
 import { readToolRenderer } from "../../src/tools/read";
 
 function extractLinkUris(text: string): string[] {
@@ -8,6 +10,7 @@ function extractLinkUris(text: string): string[] {
 }
 
 beforeAll(async () => {
+	await initTheme();
 	resetSettingsForTest();
 	await Settings.init({ inMemory: true });
 });
@@ -74,3 +77,36 @@ describe("readToolRenderer hyperlinks", () => {
 		expect(extractLinkUris(rendered)).toContain("http://example.com/final");
 	});
 });
+
+describe("read ToolExecutionComponent framing", () => {
+	it("renders framed read results inside the standard tool container padding", () => {
+		const uiStub = { requestRender() {} } as unknown as TUI;
+		const component = new ToolExecutionComponent("read", { path: "src/example.ts" }, {}, undefined, uiStub);
+		component.updateResult(
+			{
+				content: [{ type: "text", text: "export const x = 1;" }],
+				details: {
+					displayContent: { text: "export const x = 1;", startLine: 1 },
+					contentType: "text/plain",
+				},
+			},
+			false,
+		);
+
+		try {
+			const lines = component.render(80).map(line => Bun.stripANSI(line));
+			const topBorderIndex = lines.findIndex(
+				line => line.includes(activeTheme.boxSharp.topLeft) && line.includes("Read"),
+			);
+			const bottomBorderIndex = lines.findIndex(
+				(line, index) => index > topBorderIndex && line.includes(activeTheme.boxSharp.bottomLeft),
+			);
+
+			expect(topBorderIndex).toBeGreaterThanOrEqual(0);
+			expect(lines[topBorderIndex + 1]).toContain("export const x = 1;");
+			expect(bottomBorderIndex).toBeGreaterThan(topBorderIndex);
+		} finally {
+			component.stopAnimation();
+		}
+	});
+});
diff --git a/packages/coding-agent/test/tools/render-utils.test.ts b/packages/coding-agent/test/tools/render-utils.test.ts
index e4a7909a6..f1c32d99c 100644
--- a/packages/coding-agent/test/tools/render-utils.test.ts
+++ b/packages/coding-agent/test/tools/render-utils.test.ts
@@ -1,16 +1,20 @@
-import { beforeAll, describe, expect, it } from "bun:test";
+import { afterEach, beforeAll, beforeEach, describe, expect, it } from "bun:test";
 import * as os from "node:os";
 import * as path from "node:path";
-import { getThemeByName, initTheme, theme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
+import { KeybindingsManager } from "@oh-my-pi/pi-coding-agent/config/keybindings";
+import { getThemeByName, initTheme, type Theme, theme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
 import {
 	dedupeParseErrors,
+	expandKeyHint,
 	formatCodeFrameLine,
 	formatDiagnostics,
 	formatErrorMessage,
+	formatExpandHint,
 	formatParseErrors,
 	formatScreenshot,
 	truncateDiffByHunk,
 } from "@oh-my-pi/pi-coding-agent/tools/render-utils";
+import { getKeybindings, setKeybindings, type KeybindingsManager as TuiKeybindingsManager } from "@oh-my-pi/pi-tui";
 
 describe("parse error formatting", () => {
 	it("deduplicates parse errors while preserving order", () => {
@@ -282,3 +286,39 @@ describe("formatErrorMessage (F4 sanitization)", () => {
 		expect(out).toContain("Unknown error");
 	});
 });
+
+describe("formatExpandHint / expandKeyHint", () => {
+	// Plain stub: `fg` is a passthrough and brackets are literal `[`/`]`, so the
+	// rendered hint is deterministic regardless of the active theme's bracket glyphs.
+	const plainTheme = {
+		fg: (_color: unknown, text: string) => text,
+		format: { bracketLeft: "[", bracketRight: "]" },
+	} as unknown as Theme;
+
+	let previous: TuiKeybindingsManager;
+	beforeEach(() => {
+		previous = getKeybindings();
+	});
+	afterEach(() => {
+		setKeybindings(previous);
+	});
+
+	it("reports the default tool-output expand key", () => {
+		setKeybindings(KeybindingsManager.inMemory());
+		expect(expandKeyHint()).toBe("Ctrl+O");
+		// Single bracket pair from the theme, no double-wrapping around the key.
+		expect(formatExpandHint(plainTheme, false, true)).toBe("[Ctrl+O: Expand]");
+	});
+
+	it("tracks a user remap of the expand binding", () => {
+		setKeybindings(KeybindingsManager.inMemory({ "app.tools.expand": "alt+e" }));
+		expect(expandKeyHint()).toBe("Alt+E");
+		expect(formatExpandHint(plainTheme, false, true)).toBe("[Alt+E: Expand]");
+	});
+
+	it("renders nothing when expanded or there is no more content", () => {
+		setKeybindings(KeybindingsManager.inMemory());
+		expect(formatExpandHint(plainTheme, true, true)).toBe("");
+		expect(formatExpandHint(plainTheme, false, false)).toBe("");
+	});
+});
diff --git a/packages/coding-agent/test/tools/report-tool-issue.test.ts b/packages/coding-agent/test/tools/report-tool-issue.test.ts
index 60d41632c..dcbee11be 100644
--- a/packages/coding-agent/test/tools/report-tool-issue.test.ts
+++ b/packages/coding-agent/test/tools/report-tool-issue.test.ts
@@ -1,7 +1,11 @@
 import { Database } from "bun:sqlite";
 import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
 import { Settings } from "@oh-my-pi/pi-coding-agent/config/settings";
-import { __resetAutoQaFlushStateForTests, flushGrievances } from "@oh-my-pi/pi-coding-agent/tools/report-tool-issue";
+import {
+	__resetAutoQaFlushStateForTests,
+	flushGrievances,
+	isAutoQaEnabled,
+} from "@oh-my-pi/pi-coding-agent/tools/report-tool-issue";
 import * as piUtils from "@oh-my-pi/pi-utils";
 import { hookFetch } from "@oh-my-pi/pi-utils";
 
@@ -57,11 +61,23 @@ function pushSettings(overrides: Record<string, unknown> = {}): Settings {
 	});
 }
 
+let originalPiAutoQa: string | undefined;
+
+function restoreAutoQaEnv(): void {
+	if (originalPiAutoQa === undefined) {
+		delete Bun.env.PI_AUTO_QA;
+		return;
+	}
+	Bun.env.PI_AUTO_QA = originalPiAutoQa;
+}
+
 describe("flushGrievances", () => {
 	let db: Database;
 
 	beforeEach(() => {
 		__resetAutoQaFlushStateForTests();
+		originalPiAutoQa = Bun.env.PI_AUTO_QA;
+		delete Bun.env.PI_AUTO_QA;
 		vi.spyOn(piUtils, "getInstallId").mockReturnValue("11111111-2222-3333-4444-555555555555");
 		db = openTempDb();
 	});
@@ -69,9 +85,22 @@ describe("flushGrievances", () => {
 	afterEach(() => {
 		vi.restoreAllMocks();
 		__resetAutoQaFlushStateForTests();
+		restoreAutoQaEnv();
 		db.close();
 	});
 
+	it("lets PI_AUTO_QA=false disable auto QA when the setting is enabled", () => {
+		Bun.env.PI_AUTO_QA = "0";
+
+		expect(isAutoQaEnabled(Settings.isolated({ "dev.autoqa": true }))).toBe(false);
+	});
+
+	it("lets PI_AUTO_QA=true enable auto QA when the setting is disabled", () => {
+		Bun.env.PI_AUTO_QA = "1";
+
+		expect(isAutoQaEnabled(Settings.isolated({ "dev.autoqa": false }))).toBe(true);
+	});
+
 	it("skips network when consent is missing and leaves rows intact", async () => {
 		insertGrievance(db, "find", "weird ordering");
 		const fetchSpy = vi.fn(() => new Response("unexpected", { status: 200 }));
diff --git a/packages/coding-agent/test/tools/resolve.test.ts b/packages/coding-agent/test/tools/resolve.test.ts
index 14573934c..c9db93a56 100644
--- a/packages/coding-agent/test/tools/resolve.test.ts
+++ b/packages/coding-agent/test/tools/resolve.test.ts
@@ -35,6 +35,17 @@ describe("ResolveTool", () => {
 		);
 	});
 
+	it("treats discard with no pending action as a successful cancellation, not an error", async () => {
+		const tool = new ResolveTool(createSession());
+		const result = await tool.execute("call-discard-none", {
+			action: "discard",
+			reason: "Abandoning the staged edit.",
+		});
+		expect(result.isError ?? false).toBe(false);
+		expect(getText(result)).toContain("Nothing to discard");
+		expect(result.details).toMatchObject({ action: "discard", reason: "Abandoning the staged edit." });
+	});
+
 	it("discards pending action and clears store", async () => {
 		let discardedReason: string | undefined;
 		const handler = async (input: unknown) => {
diff --git a/packages/coding-agent/test/tools/search-internal-urls.test.ts b/packages/coding-agent/test/tools/search-internal-urls.test.ts
index 52edda3b5..c541ab4a8 100644
--- a/packages/coding-agent/test/tools/search-internal-urls.test.ts
+++ b/packages/coding-agent/test/tools/search-internal-urls.test.ts
@@ -251,7 +251,7 @@ describe("SearchTool internal URL resolution", () => {
 		const text = getResultText(result);
 		expect(text).toContain("needle");
 		// No hashline section headers or numbered editable lines for immutable sources.
-		expect(text).not.toMatch(/^¶.*#[0-9A-F]{4}$/m);
+		expect(text).not.toMatch(/^\[[^#\r\n]+#[0-9A-F]{4}\]$/m);
 		expect(text).not.toMatch(/^\*?\s*\d+:/m);
 	});
 
@@ -291,7 +291,7 @@ describe("SearchTool internal URL resolution", () => {
 		const text = getResultText(result);
 		expect(text).toContain("needle");
 		// Mutable local:// sources keep a hashline section header plus numbered match lines.
-		expect(text).toMatch(/^¶.*#[0-9A-F]{4}$/m);
+		expect(text).toMatch(/^\[[^#\r\n]+#[0-9A-F]{4}\]$/m);
 		expect(text).toMatch(/^\*\d+:.*needle/m);
 	});
 
diff --git a/packages/coding-agent/test/tools/search-path-lists.test.ts b/packages/coding-agent/test/tools/search-path-lists.test.ts
index d23a66c4e..cb7b695dd 100644
--- a/packages/coding-agent/test/tools/search-path-lists.test.ts
+++ b/packages/coding-agent/test/tools/search-path-lists.test.ts
@@ -194,7 +194,7 @@ describe("tool path arrays", () => {
 		const text = getText(result);
 		const details = result.details as { fileCount?: number; missingPaths?: string[] } | undefined;
 
-		expect(text).toMatch(/^¶packages\/grep\.txt#[0-9A-F]{4}/m);
+		expect(text).toMatch(/^\[packages\/grep\.txt#[0-9A-F]{4}\]/m);
 		expect(text).toContain("Skipped missing paths: missing.txt");
 		expect(text).not.toContain("apps");
 		expect(details?.fileCount).toBe(1);
diff --git a/packages/coding-agent/test/tools/search-renderer.test.ts b/packages/coding-agent/test/tools/search-renderer.test.ts
index bbd101e70..502de64de 100644
--- a/packages/coding-agent/test/tools/search-renderer.test.ts
+++ b/packages/coding-agent/test/tools/search-renderer.test.ts
@@ -22,7 +22,30 @@ afterAll(() => {
 });
 
 describe("searchToolRenderer", () => {
-	it("keeps summary and truncation rows inside the collapsed line budget", async () => {
+	it("indents inline search output and avoids accent-colored success headers", async () => {
+		const theme = await getThemeByName("dark");
+		expect(theme).toBeDefined();
+		const uiTheme = theme!;
+		const result = {
+			content: [{ type: "text", text: "" }],
+			details: {
+				matchCount: 1,
+				fileCount: 1,
+				displayContent: ["# src/", "## file.ts#abcd", "*12│const needle = true;"].join("\n"),
+			},
+		};
+
+		const renderedLines = searchToolRenderer
+			.renderResult(result as never, { expanded: true, isPartial: false }, uiTheme, { pattern: "needle" })
+			.render(240);
+		const plainLines = sanitizeText(renderedLines.join("\n")).split("\n");
+
+		expect(plainLines.every(line => line.startsWith(" "))).toBe(true);
+		expect(renderedLines[0]).not.toContain(uiTheme.fg("accent", uiTheme.symbol("icon.search")));
+		expect(renderedLines[0]).not.toContain(uiTheme.fg("accent", "Search"));
+	});
+
+	it("keeps truncation status in the header without a bottom notice", async () => {
 		const theme = await getThemeByName("dark");
 		expect(theme).toBeDefined();
 		const uiTheme = theme!;
@@ -38,6 +61,8 @@ describe("searchToolRenderer", () => {
 				matchCount: 6,
 				fileCount: 3,
 				fileLimitReached: 3,
+				perFileLimitReached: 20,
+				truncated: true,
 			},
 		};
 
@@ -52,10 +77,53 @@ describe("searchToolRenderer", () => {
 		const renderedLines = sanitizeText(collapsed.render(200).join("\n")).split("\n");
 		const bodyLines = renderedLines.slice(1);
 
+		expect(renderedLines[0]).toContain("truncated");
 		expect(bodyLines).toHaveLength(6);
-		expect(bodyLines.at(-1)).toContain("truncated: first 3 files (skip to paginate)");
+		expect(renderedLines.join("\n")).not.toContain("truncated:");
+		expect(renderedLines.join("\n")).not.toContain("skip to paginate");
+		expect(renderedLines.join("\n")).not.toContain("matches per file");
+		expect(bodyLines.some(line => line.includes("gamma:1"))).toBe(true);
+	});
+
+	it("shows actual matches when one grouped search section is larger than the collapsed budget", async () => {
+		const theme = await getThemeByName("dark");
+		expect(theme).toBeDefined();
+		const uiTheme = theme!;
+
+		const result = {
+			content: [{ type: "text", text: "" }],
+			details: {
+				matchCount: 3,
+				fileCount: 3,
+				displayContent: [
+					"# src/",
+					"## first.ts#aaaa",
+					" 1│before",
+					"*2│const firstFlag = true;",
+					" 3│after",
+					"## second.ts#bbbb",
+					"*4│const secondFlag = true;",
+					"## third.ts#cccc",
+					"*5│const thirdFlag = true;",
+				].join("\n"),
+			},
+		};
+
+		const collapsed = searchToolRenderer.renderResult(
+			result as never,
+			{ expanded: false, isPartial: false },
+			uiTheme,
+			{ pattern: "Flag" },
+		);
+		const renderedLines = sanitizeText(collapsed.render(240).join("\n")).split("\n");
+		const bodyLines = renderedLines.slice(1);
+
+		expect(bodyLines).toHaveLength(6);
+		expect(bodyLines.some(line => line.includes("const firstFlag = true;"))).toBe(true);
+		expect(bodyLines.some(line => line.includes("const secondFlag = true;"))).toBe(true);
 		expect(bodyLines.some(line => line.includes("1 more match"))).toBe(true);
-		expect(bodyLines.some(line => line.includes("gamma:1"))).toBe(false);
+		expect(bodyLines.some(line => line.includes("before"))).toBe(false);
+		expect(bodyLines.some(line => line.includes("thirdFlag"))).toBe(false);
 	});
 
 	it("links grouped file headers and code-frame lines to filesystem targets", async () => {
@@ -109,4 +177,54 @@ describe("searchToolRenderer", () => {
 
 		expect(extractLinkUris(rendered)).toContain("file:///tmp/omp-project/file.ts?line=7");
 	});
+
+	it("bounds the expanded single-file view instead of dumping every match", async () => {
+		const theme = await getThemeByName("dark");
+		expect(theme).toBeDefined();
+		const uiTheme = theme!;
+
+		// One file's matches collapse into a single blank-line group (no `#`/`##`
+		// headers, `│...` gap separators). Before the fix the expanded renderer
+		// dumped the entire span because the tree list ignored the line budget.
+		const clusters = Array.from({ length: 12 }, (_, i) => i * 100 + 1);
+		const displayContent = clusters
+			.map((line, idx) => {
+				const cluster = [` ${line}│ context before`, `*${line + 1}│ MATCH ${idx}`, ` ${line + 2}│ context after`];
+				return idx === 0 ? cluster.join("\n") : ["    │...", ...cluster].join("\n");
+			})
+			.join("\n");
+
+		const result = {
+			content: [{ type: "text", text: "" }],
+			details: {
+				matchCount: clusters.length,
+				fileCount: 1,
+				searchPath: "/tmp/omp-project/renderer.ts",
+				scopePath: "renderer.ts",
+				displayContent,
+			},
+		};
+
+		const render = (expanded: boolean) =>
+			sanitizeText(
+				searchToolRenderer
+					.renderResult(result as never, { expanded, isPartial: false }, uiTheme, { pattern: "needle" })
+					.render(200)
+					.join("\n"),
+			).split("\n");
+
+		const expanded = render(true);
+		const expandedBody = expanded.slice(1);
+		// Bounded: must not render all 12 clusters (36+ lines).
+		expect(expandedBody.length).toBeLessThan(clusters.length * 3);
+		expect(expandedBody.some(line => line.includes("more matches"))).toBe(true);
+		// Expanded keeps surrounding context lines (unlike the compact collapsed view).
+		expect(expandedBody.some(line => line.includes("context before"))).toBe(true);
+
+		const collapsedBody = render(false).slice(1);
+		expect(collapsedBody.length).toBeLessThan(expandedBody.length);
+		// Collapsed compacts to match lines only — no context.
+		expect(collapsedBody.some(line => line.includes("context before"))).toBe(false);
+		expect(collapsedBody.some(line => line.includes("more matches"))).toBe(true);
+	});
 });
diff --git a/packages/coding-agent/test/tools/task-repair-args.test.ts b/packages/coding-agent/test/tools/task-repair-args.test.ts
index 7979c367c..9dfedf3f0 100644
--- a/packages/coding-agent/test/tools/task-repair-args.test.ts
+++ b/packages/coding-agent/test/tools/task-repair-args.test.ts
@@ -72,9 +72,4 @@ describe("repairTaskParams", () => {
 		} as unknown as TaskParams;
 		expect(repairTaskParams(params)).toBe(params);
 	});
-
-	it("tolerates partially-streamed args without throwing", () => {
-		const partial = { agent: "task", tasks: [{ id: "A" }, undefined] } as unknown as TaskParams;
-		expect(() => repairTaskParams(partial)).not.toThrow();
-	});
 });
diff --git a/packages/coding-agent/test/tools/todo.test.ts b/packages/coding-agent/test/tools/todo.test.ts
index 73d0d6a61..17f54e0b2 100644
--- a/packages/coding-agent/test/tools/todo.test.ts
+++ b/packages/coding-agent/test/tools/todo.test.ts
@@ -345,3 +345,124 @@ describe("todoMatchesAnyDescription", () => {
 		expect(todoMatchesAnyDescription("Audit AGENTS.md compliance", ["Audit AGENTS md compliance"])).toBe(true);
 	});
 });
+describe("todoToolRenderer.renderResult phase collapsing", () => {
+	async function buildThreePhaseAfterDone() {
+		const tool = new TodoTool(createSession());
+		await tool.execute("init", {
+			ops: [
+				{
+					op: "init",
+					list: [
+						{ phase: "Alpha", items: ["a1", "a2"] },
+						{ phase: "Beta", items: ["b1", "b2"] },
+						{ phase: "Gamma", items: ["c1", "c2"] },
+					],
+				},
+			],
+		});
+		// `done a1` keeps the active task inside Alpha (auto-promotes a2), leaving
+		// Beta and Gamma untouched by this update.
+		return tool.execute("done", { ops: [{ op: "done", task: "a1" }] });
+	}
+	function innerLines(component: ReturnType<typeof todoToolRenderer.renderResult>): string[] {
+		const lines = Bun.stripANSI(component.render(100).join("\n")).split("\n");
+		return lines.slice(1, -1).map(line => line.replace(/^│/, "").replace(/│\s*$/, "").trim());
+	}
+	it("collapses untouched phases to a one-line summary while expanding the active phase", async () => {
+		const result = await buildThreePhaseAfterDone();
+		const component = todoToolRenderer.renderResult(result, { expanded: false, isPartial: false }, theme, {
+			ops: [{ op: "done", task: "a1" }],
+		});
+		const rendered = Bun.stripANSI(component.render(100).join("\n"));
+		// Active phase renders its full task list.
+		expect(rendered).toContain("a1");
+		expect(rendered).toContain("a2");
+		// Untouched phases collapse: headers + progress counts, no task contents.
+		expect(rendered).toContain("II. Beta");
+		expect(rendered).toContain("III. Gamma");
+		expect(rendered).toContain("0/2");
+		expect(rendered).not.toContain("b1");
+		expect(rendered).not.toContain("b2");
+		expect(rendered).not.toContain("c1");
+		expect(rendered).not.toContain("c2");
+	});
+	it("falls back to in_progress / completed signals when call args are unavailable", async () => {
+		const result = await buildThreePhaseAfterDone();
+		// Transcript rebuilds may not carry call args; the active (Alpha) phase is
+		// still derived from the in_progress task and the completion transition.
+		const component = todoToolRenderer.renderResult(result, { expanded: false, isPartial: false }, theme);
+		const rendered = Bun.stripANSI(component.render(100).join("\n"));
+		expect(rendered).toContain("a2");
+		expect(rendered).not.toContain("b1");
+		expect(rendered).not.toContain("c1");
+	});
+	it("shows every phase fully when manually expanded", async () => {
+		const result = await buildThreePhaseAfterDone();
+		const component = todoToolRenderer.renderResult(result, { expanded: true, isPartial: false }, theme, {
+			ops: [{ op: "done", task: "a1" }],
+		});
+		const rendered = Bun.stripANSI(component.render(100).join("\n"));
+		expect(rendered).toContain("b1");
+		expect(rendered).toContain("b2");
+		expect(rendered).toContain("c1");
+		expect(rendered).toContain("c2");
+	});
+	it("drops blank separator lines between phases", async () => {
+		const result = await buildThreePhaseAfterDone();
+		const component = todoToolRenderer.renderResult(result, { expanded: true, isPartial: false }, theme, {
+			ops: [{ op: "done", task: "a1" }],
+		});
+		// No empty body line survives between phases.
+		expect(innerLines(component).every(line => line.length > 0)).toBe(true);
+	});
+});
+
+describe("todoToolRenderer.renderCall malformed-args regression (#2005)", () => {
+	// Reporter saw `TypeError: args?.ops?.map is not a function` against
+	// Xiaomi Token Plan's Anthropic protocol because `parseStreamingJson`
+	// surfaced `{ ops: "[..." }` shapes mid-stream. The renderer is invoked
+	// on every streaming delta, so any non-array `ops` (string, object,
+	// number) must NOT crash the TUI render loop and trigger the spam-warn /
+	// retry cascade.
+	const renderOptions = { expanded: false, isPartial: true } as const;
+
+	it("does not throw when ops is a streaming-truncated string", () => {
+		// Mid-stream `partialJson === '{"ops":"[{'` parses into `{ops: "[{"}`.
+		const args = { ops: '[{"op":"init"' } as unknown as Parameters<typeof todoToolRenderer.renderCall>[0];
+		expect(() => todoToolRenderer.renderCall(args, renderOptions, theme)).not.toThrow();
+	});
+
+	it("does not throw when ops entries are null", () => {
+		// `partialParse` of `'{"ops":[null'` can hand back `{ops: [null]}` in
+		// intermediate states before the entry object opens.
+		const args = { ops: [null] } as unknown as Parameters<typeof todoToolRenderer.renderCall>[0];
+		expect(() => todoToolRenderer.renderCall(args, renderOptions, theme)).not.toThrow();
+	});
+
+	it("does not throw when an entry's items field is a non-array", () => {
+		const args = {
+			ops: [{ op: "append", phase: "Work", items: "Second" as unknown as string[] }],
+		} as unknown as Parameters<typeof todoToolRenderer.renderCall>[0];
+		expect(() => todoToolRenderer.renderCall(args, renderOptions, theme)).not.toThrow();
+	});
+
+	it("still renders ops summary metadata for well-formed args", () => {
+		const args = {
+			ops: [
+				{ op: "init", items: ["a", "b", "c"] },
+				{ op: "done", task: "a" },
+				{ op: "append", phase: "Cleanup", items: ["d"] },
+			],
+		};
+		const component = todoToolRenderer.renderCall(args, renderOptions, theme);
+		// `Text(text, 0, 0)` from `@oh-my-pi/pi-tui` exposes the content via .render().
+		const rendered = Bun.stripANSI(component.render(120).join("\n"));
+		expect(rendered).toContain("init");
+		expect(rendered).toContain("3 items");
+		expect(rendered).toContain("done");
+		expect(rendered).toContain("a");
+		expect(rendered).toContain("append");
+		expect(rendered).toContain("Cleanup");
+		expect(rendered).toContain("1 item");
+	});
+});
diff --git a/packages/coding-agent/test/tools/tool-output-hyperlinks.test.ts b/packages/coding-agent/test/tools/tool-output-hyperlinks.test.ts
new file mode 100644
index 000000000..e7da44dfc
--- /dev/null
+++ b/packages/coding-agent/test/tools/tool-output-hyperlinks.test.ts
@@ -0,0 +1,182 @@
+import { afterAll, afterEach, beforeAll, describe, expect, it } from "bun:test";
+import * as fs from "node:fs";
+import * as os from "node:os";
+import * as path from "node:path";
+import { resetSettingsForTest, Settings, settings } from "@oh-my-pi/pi-coding-agent/config/settings";
+import { editToolRenderer } from "@oh-my-pi/pi-coding-agent/edit/renderer";
+import type { ToolSession } from "@oh-my-pi/pi-coding-agent/tools";
+import { astGrepToolRenderer } from "@oh-my-pi/pi-coding-agent/tools/ast-grep";
+import { ReadTool, readToolRenderer } from "@oh-my-pi/pi-coding-agent/tools/read";
+import { searchToolRenderer } from "@oh-my-pi/pi-coding-agent/tools/search";
+import { WriteTool, writeToolRenderer } from "@oh-my-pi/pi-coding-agent/tools/write";
+import { getThemeByName, initTheme } from "../../src/modes/theme/theme";
+
+// 1x1 PNG so the read tool takes its image branch.
+const TINY_PNG_BASE64 =
+	"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8z8DwHwAFBQIAX8jx0gAAAABJRU5ErkJggg==";
+
+function extractLinkUris(text: string): string[] {
+	return [...text.matchAll(/\x1b\]8;[^;]*;([^\x1b]+)\x1b\\/g)].map(match => match[1]!);
+}
+
+function createTestToolSession(cwd: string): ToolSession {
+	return {
+		cwd,
+		hasUI: false,
+		getSessionFile: () => null,
+		getSessionSpawns: () => "*",
+		settings: Settings.isolated(),
+	};
+}
+
+beforeAll(async () => {
+	await initTheme(false);
+	resetSettingsForTest();
+	await Settings.init({ inMemory: true });
+});
+
+afterEach(() => {
+	settings.clearOverride("tui.hyperlinks");
+});
+
+afterAll(() => {
+	resetSettingsForTest();
+});
+
+describe("tool output OSC 8 file:// hyperlinks", () => {
+	it("links plain text and image read titles to the resolved filesystem path", async () => {
+		settings.override("tui.hyperlinks", "always");
+		const theme = (await getThemeByName("dark"))!;
+		const dir = fs.mkdtempSync(path.join(os.tmpdir(), "omp-link-read-"));
+		try {
+			const textPath = path.join(dir, "task.txt");
+			fs.writeFileSync(textPath, "hello\nworld\n");
+			const imgPath = path.join(dir, "task.png");
+			fs.writeFileSync(imgPath, Buffer.from(TINY_PNG_BASE64, "base64"));
+
+			const tool = new ReadTool(createTestToolSession(dir));
+			const textRes = await tool.execute("t", { path: textPath });
+			const imgRes = await tool.execute("i", { path: imgPath });
+
+			const textRender = readToolRenderer
+				.renderResult(
+					{ content: textRes.content, details: textRes.details, isError: textRes.isError },
+					{ expanded: false, isPartial: false },
+					theme,
+					{ path: textPath },
+				)
+				.render(200)
+				.join("\n");
+			const imgRender = readToolRenderer
+				.renderResult(
+					{ content: imgRes.content, details: imgRes.details, isError: imgRes.isError },
+					{ expanded: false, isPartial: false },
+					theme,
+					{ path: imgPath },
+				)
+				.render(200)
+				.join("\n");
+
+			expect(extractLinkUris(textRender)).toContain(`file://${textPath}`);
+			expect(extractLinkUris(imgRender)).toContain(`file://${imgPath}`);
+		} finally {
+			fs.rmSync(dir, { recursive: true, force: true });
+		}
+	});
+
+	it("links the write header to the absolute path it wrote", async () => {
+		settings.override("tui.hyperlinks", "always");
+		const theme = (await getThemeByName("dark"))!;
+		const dir = fs.mkdtempSync(path.join(os.tmpdir(), "omp-link-write-"));
+		try {
+			const filePath = path.join(dir, "out.ts");
+			const tool = new WriteTool(createTestToolSession(dir));
+			const res = await tool.execute("w", { path: filePath, content: "export const x = 1;\n" });
+			const rendered = writeToolRenderer
+				.renderResult(
+					{ content: res.content, details: res.details, isError: res.isError },
+					{ expanded: false, isPartial: false },
+					theme,
+					{ path: filePath },
+				)
+				.render(200)
+				.join("\n");
+			expect(extractLinkUris(rendered)).toContain(`file://${filePath}`);
+		} finally {
+			fs.rmSync(dir, { recursive: true, force: true });
+		}
+	});
+
+	it("resolves scoped search links against cwd, not the (sub)scope path", async () => {
+		settings.override("tui.hyperlinks", "always");
+		const theme = (await getThemeByName("dark"))!;
+		// Scoped search: scope dir (`searchPath`) is below cwd, and the grouped
+		// display paths are cwd-relative. Resolving against searchPath would double
+		// the `src` prefix (`/proj/src/src/...`).
+		const result = {
+			content: [{ type: "text", text: "" }],
+			details: {
+				matchCount: 1,
+				fileCount: 1,
+				cwd: "/tmp/omp-project",
+				searchPath: "/tmp/omp-project/src",
+				scopePath: "src",
+				displayContent: ["# src/", "## interactive-mode.ts#abcd", "*12│const needle = true;"].join("\n"),
+			},
+		};
+		const rendered = searchToolRenderer
+			.renderResult(result as never, { expanded: true, isPartial: false }, theme, { pattern: "needle" })
+			.render(240)
+			.join("\n");
+		const uris = extractLinkUris(rendered);
+		expect(uris).toContain("file:///tmp/omp-project/src/interactive-mode.ts");
+		expect(uris).toContain("file:///tmp/omp-project/src/interactive-mode.ts?line=12");
+		expect(uris.some(uri => uri.includes("/src/src/"))).toBe(false);
+	});
+
+	it("resolves scoped ast-grep links against cwd, not the (sub)scope path", async () => {
+		settings.override("tui.hyperlinks", "always");
+		const theme = (await getThemeByName("dark"))!;
+		const result = {
+			content: [{ type: "text", text: "" }],
+			details: {
+				matchCount: 1,
+				fileCount: 1,
+				filesSearched: 1,
+				limitReached: false,
+				cwd: "/tmp/omp-project",
+				searchPath: "/tmp/omp-project/src",
+				scopePath: "src",
+				displayContent: ["# src/", "## interactive-mode.ts", "  *12│const needle = true;"].join("\n"),
+			},
+		};
+		const rendered = astGrepToolRenderer
+			.renderResult(result as never, { expanded: true, isPartial: false }, theme, { pat: "needle" })
+			.render(240)
+			.join("\n");
+		const uris = extractLinkUris(rendered);
+		expect(uris).toContain("file:///tmp/omp-project/src/interactive-mode.ts");
+		expect(uris.some(uri => uri.includes("/src/src/"))).toBe(false);
+	});
+
+	it("links the edit header to the absolute details.path even when the arg path is relative", async () => {
+		settings.override("tui.hyperlinks", "always");
+		const theme = (await getThemeByName("dark"))!;
+		const rendered = editToolRenderer
+			.renderResult(
+				{
+					content: [{ type: "text", text: "Updated src/a.ts" }],
+					details: { diff: "+1|// x", op: "update", path: "/tmp/omp-project/src/a.ts" },
+				},
+				{ expanded: false, isPartial: false, renderContext: { editMode: "hashline" } },
+				theme,
+				{ path: "src/a.ts" },
+			)
+			.render(200)
+			.join("\n");
+		const uris = extractLinkUris(rendered);
+		expect(uris).toContain("file:///tmp/omp-project/src/a.ts");
+		// A relative arg path must not leak into a root-anchored `file:///src/a.ts`.
+		expect(uris).not.toContain("file:///src/a.ts");
+	});
+});
diff --git a/packages/coding-agent/test/tools/web-scrapers/academic.test.ts b/packages/coding-agent/test/tools/web-scrapers/academic.test.ts
index 724faef00..7cbd53b79 100644
--- a/packages/coding-agent/test/tools/web-scrapers/academic.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/academic.test.ts
@@ -164,15 +164,6 @@ describe.skipIf(SKIP)("handleArxiv", () => {
 		}
 	});
 
-	it("handles arxiv.org/abs/ format", async () => {
-		const result = await handleArxiv("https://arxiv.org/abs/1706.03762", 30000);
-		expect(result).not.toBeNull();
-		expect(result?.method).toBe("arxiv");
-		if (!result?.content.includes("Too Many Requests") && !result?.content.includes("Failed to fetch")) {
-			expect(result?.content).toContain("1706.03762");
-		}
-	});
-
 	it("includes paper metadata", async () => {
 		const result = await handleArxiv("https://arxiv.org/abs/1706.03762", 30000);
 		expect(result).not.toBeNull();
@@ -183,14 +174,6 @@ describe.skipIf(SKIP)("handleArxiv", () => {
 			expect(result?.content).toMatch(/Published:/);
 		}
 	});
-
-	it("handles rate limiting gracefully", async () => {
-		const result = await handleArxiv("https://arxiv.org/abs/1706.03762", 5000);
-		expect(result).not.toBeNull();
-		expect(result?.method).toBe("arxiv");
-		// Should return something, even if rate limited
-		expect(result?.content).toBeTruthy();
-	});
 });
 
 describe.skipIf(SKIP)("handleIacr", () => {
@@ -216,14 +199,6 @@ describe.skipIf(SKIP)("handleIacr", () => {
 		}
 	});
 
-	it("handles rate limiting gracefully", async () => {
-		const result = await handleIacr("https://eprint.iacr.org/2023/123", 5000);
-		expect(result).not.toBeNull();
-		expect(result?.method).toBe("iacr");
-		// Should return something, even if rate limited
-		expect(result?.content).toBeTruthy();
-	});
-
 	it("handles PDF URLs", async () => {
 		const result = await handleIacr("https://eprint.iacr.org/2023/123.pdf", 30000);
 		expect(result).not.toBeNull();
diff --git a/packages/coding-agent/test/tools/web-scrapers/business.test.ts b/packages/coding-agent/test/tools/web-scrapers/business.test.ts
index 78fa468f7..efc2ac435 100644
--- a/packages/coding-agent/test/tools/web-scrapers/business.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/business.test.ts
@@ -26,8 +26,6 @@ describe.skipIf(SKIP)("handleSecEdgar", () => {
 		expect(result?.content).toContain("0000320193");
 		expect(result?.content).toContain("10-K"); // Apple files 10-K annually
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches via data.sec.gov submissions URL", async () => {
@@ -67,8 +65,6 @@ describe.skipIf(SKIP)("handleOpenCorporates", () => {
 		expect(result?.content).toContain("2927442");
 		expect(result?.content).toContain("US_DE");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches Microsoft Corporation", async () => {
diff --git a/packages/coding-agent/test/tools/web-scrapers/documentation.test.ts b/packages/coding-agent/test/tools/web-scrapers/documentation.test.ts
index be89cc16a..f61d12341 100644
--- a/packages/coding-agent/test/tools/web-scrapers/documentation.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/documentation.test.ts
@@ -29,7 +29,6 @@ describe.skipIf(SKIP)("handleMDN", () => {
 		expect(result?.method).toBe("mdn");
 		expect(result?.content).toContain("map");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
 	});
 
 	it("fetches Promise documentation", async () => {
@@ -40,7 +39,6 @@ describe.skipIf(SKIP)("handleMDN", () => {
 		expect(result).not.toBeNull();
 		expect(result?.method).toBe("mdn");
 		expect(result?.content).toContain("Promise");
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches CSS documentation", async () => {
@@ -66,8 +64,6 @@ describe.skipIf(SKIP)("handleReadTheDocs", () => {
 		const result = await handleReadTheDocs("https://requests.readthedocs.io/en/latest/", 20);
 		expect(result).not.toBeNull();
 		expect(result?.method).toBe("readthedocs");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("returns null for non-readthedocs sites", async () => {
diff --git a/packages/coding-agent/test/tools/web-scrapers/finance-media.test.ts b/packages/coding-agent/test/tools/web-scrapers/finance-media.test.ts
index 24b36dfc0..18d60da1d 100644
--- a/packages/coding-agent/test/tools/web-scrapers/finance-media.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/finance-media.test.ts
@@ -31,8 +31,6 @@ describe.skipIf(SKIP)("handleCoinGecko", () => {
 			expect(result?.content).toContain("Price");
 		}
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches Ethereum data", async () => {
@@ -44,7 +42,6 @@ describe.skipIf(SKIP)("handleCoinGecko", () => {
 			expect(result?.content).toContain("ETH");
 			expect(result?.content).toContain("Market Cap");
 		}
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("handles URL without locale prefix", async () => {
@@ -77,8 +74,6 @@ describe.skipIf(SKIP)("handleDiscogs", () => {
 		expect(result?.method).toBe("discogs");
 		expect(result?.content).toContain("Tracklist");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches master release", async () => {
@@ -87,7 +82,6 @@ describe.skipIf(SKIP)("handleDiscogs", () => {
 		expect(result).not.toBeNull();
 		expect(result?.method).toBe("discogs");
 		expect(result?.content).toContain("Master Release");
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("handles release URL with just ID", async () => {
@@ -121,8 +115,6 @@ describe.skipIf(SKIP)("handleArtifactHub", () => {
 		expect(result?.content).toContain("Helm Chart");
 		expect(result?.content).toContain("Version");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches prometheus-community/prometheus helm chart", async () => {
@@ -134,7 +126,6 @@ describe.skipIf(SKIP)("handleArtifactHub", () => {
 		expect(result?.method).toBe("artifacthub");
 		expect(result?.content).toContain("prometheus");
 		expect(result?.content).toContain("Repository");
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("handles www subdomain", async () => {
diff --git a/packages/coding-agent/test/tools/web-scrapers/git-hosting.test.ts b/packages/coding-agent/test/tools/web-scrapers/git-hosting.test.ts
index d54a9ee47..e219e5167 100644
--- a/packages/coding-agent/test/tools/web-scrapers/git-hosting.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/git-hosting.test.ts
@@ -1,5 +1,5 @@
 import { describe, expect, it } from "bun:test";
-import { handleGitHub } from "@oh-my-pi/pi-coding-agent/web/scrapers/github";
+import { handleGitHub, parseGitHubUrl, stripActionsLogTimestamps } from "@oh-my-pi/pi-coding-agent/web/scrapers/github";
 import { handleGitHubGist } from "@oh-my-pi/pi-coding-agent/web/scrapers/github-gist";
 
 const SKIP = !Bun.env.WEB_FETCH_INTEGRATION;
@@ -203,3 +203,57 @@ describe.skipIf(SKIP)("handleGitHubGist", () => {
 		expect(result).toBeDefined();
 	});
 });
+
+// =============================================================================
+// GitHub Actions URL parsing (pure, network-free)
+// =============================================================================
+
+describe("parseGitHubUrl — Actions", () => {
+	it("classifies a workflow run URL", () => {
+		const gh = parseGitHubUrl("https://github.com/can1357/oh-my-pi/actions/runs/27070071296");
+		expect(gh).toEqual({ type: "actions-run", owner: "can1357", repo: "oh-my-pi", runId: 27070071296 });
+	});
+
+	it("classifies a job URL using the web-form singular `job` segment", () => {
+		const gh = parseGitHubUrl("https://github.com/can1357/oh-my-pi/actions/runs/27070071296/job/79897931171");
+		expect(gh).toEqual({
+			type: "actions-job",
+			owner: "can1357",
+			repo: "oh-my-pi",
+			runId: 27070071296,
+			jobId: 79897931171,
+		});
+	});
+
+	it("classifies a job URL using the API-form plural `jobs` segment", () => {
+		const gh = parseGitHubUrl("https://github.com/can1357/oh-my-pi/actions/runs/27070071296/jobs/79897931171");
+		expect(gh?.type).toBe("actions-job");
+		expect(gh?.jobId).toBe(79897931171);
+	});
+
+	it("does not treat non-run Actions URLs (e.g. workflow files) as runs/jobs", () => {
+		expect(parseGitHubUrl("https://github.com/can1357/oh-my-pi/actions/workflows/ci.yml")?.type).toBe("other");
+		expect(parseGitHubUrl("https://github.com/can1357/oh-my-pi/actions")?.type).toBe("other");
+	});
+
+	it("does not misparse a run URL with a non-numeric id", () => {
+		expect(parseGitHubUrl("https://github.com/can1357/oh-my-pi/actions/runs/latest")?.type).toBe("other");
+	});
+
+	it("returns null for non-github hosts", () => {
+		expect(parseGitHubUrl("https://gitlab.com/o/r/actions/runs/1")).toBeNull();
+	});
+});
+
+describe("stripActionsLogTimestamps", () => {
+	it("removes the per-line ISO timestamp prefix and a leading BOM", () => {
+		const raw =
+			"\uFEFF2026-06-06T18:14:12.8793443Z Current runner version: '2.334.0'\n2026-06-06T18:14:13.0000000Z done\n";
+		expect(stripActionsLogTimestamps(raw)).toBe("Current runner version: '2.334.0'\ndone\n");
+	});
+
+	it("leaves grouped/non-timestamped lines untouched", () => {
+		const raw = "2026-06-06T18:14:12.0000000Z ##[group]Operating System\nUbuntu\n##[endgroup]\n";
+		expect(stripActionsLogTimestamps(raw)).toBe("##[group]Operating System\nUbuntu\n##[endgroup]\n");
+	});
+});
diff --git a/packages/coding-agent/test/tools/web-scrapers/package-managers-2.test.ts b/packages/coding-agent/test/tools/web-scrapers/package-managers-2.test.ts
index 32bc64b62..011ab7e6e 100644
--- a/packages/coding-agent/test/tools/web-scrapers/package-managers-2.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/package-managers-2.test.ts
@@ -25,8 +25,6 @@ describe.skipIf(SKIP)("handleMetaCPAN", () => {
 		expect(result?.method).toBe("metacpan");
 		expect(result?.content).toContain("Moose");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches release by distribution name", async () => {
@@ -55,8 +53,6 @@ describe.skipIf(SKIP)("handleHackage", () => {
 		expect(result?.content).toContain("aeson");
 		expect(result?.content).toContain("JSON");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	}, 20000);
 
 	it("fetches text package", async () => {
@@ -85,8 +81,6 @@ describe.skipIf(SKIP)("handleDockerHub", () => {
 		expect(result?.content).toContain("nginx");
 		expect(result?.content).toContain("docker pull");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches grafana/grafana image", async () => {
@@ -115,8 +109,6 @@ describe.skipIf(SKIP)("handleChocolatey", () => {
 		expect(result?.method).toBe("chocolatey");
 		expect(result?.content).toContain("choco install");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches nodejs package", async () => {
@@ -145,8 +137,6 @@ describe.skipIf(SKIP)("handleRepology", () => {
 		expect(result?.content).toContain("firefox");
 		expect(result?.content).toContain("Repositories");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches vim project", async () => {
@@ -176,8 +166,6 @@ describe.skipIf(SKIP)("handleTerraform", () => {
 		expect(result?.content).toContain("hashicorp");
 		expect(result?.content).toContain("required_providers");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches terraform-aws-modules/vpc/aws module", async () => {
diff --git a/packages/coding-agent/test/tools/web-scrapers/package-managers.test.ts b/packages/coding-agent/test/tools/web-scrapers/package-managers.test.ts
index 4fedcf761..a6720fa74 100644
--- a/packages/coding-agent/test/tools/web-scrapers/package-managers.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/package-managers.test.ts
@@ -16,8 +16,6 @@ describe.skipIf(SKIP)("handleBrew", () => {
 		expect(result?.content).toContain("wget");
 		expect(result?.content).toContain("brew install wget");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches firefox cask", async () => {
@@ -27,8 +25,6 @@ describe.skipIf(SKIP)("handleBrew", () => {
 		expect(result?.content).toContain("Firefox");
 		expect(result?.content).toContain("brew install --cask firefox");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 });
 
@@ -41,8 +37,6 @@ describe.skipIf(SKIP)("handleAur", () => {
 		expect(result?.content).toContain("AUR helper");
 		expect(result?.content).toContain("yay -S yay");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 });
 
@@ -54,8 +48,6 @@ describe.skipIf(SKIP)("handleRubyGems", () => {
 		expect(result?.content).toContain("rails");
 		expect(result?.content).toContain("Total Downloads");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 });
 
@@ -67,8 +59,6 @@ describe.skipIf(SKIP)("handleNuGet", () => {
 		expect(result?.content).toContain("Newtonsoft.Json");
 		expect(result?.content).toContain("JSON");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 });
 
@@ -80,8 +70,6 @@ describe.skipIf(SKIP)("handlePackagist", () => {
 		expect(result?.content).toContain("laravel/framework");
 		expect(result?.content).toContain("Downloads");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 });
 
@@ -95,8 +83,6 @@ describe.skipIf(SKIP)("handleMaven", () => {
 		expect(result?.content).toContain("<groupId>");
 		expect(result?.content).toContain("implementation");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches commons-lang3 artifact from mvnrepository.com", async () => {
diff --git a/packages/coding-agent/test/tools/web-scrapers/research.test.ts b/packages/coding-agent/test/tools/web-scrapers/research.test.ts
index a2dcb96d5..e56b28f80 100644
--- a/packages/coding-agent/test/tools/web-scrapers/research.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/research.test.ts
@@ -23,8 +23,6 @@ describe.skipIf(SKIP)("handleWikidata", () => {
 		expect(result?.content).toContain("Apple");
 		expect(result?.content).toContain("Q312");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches Q5 - human (entity)", async () => {
@@ -34,8 +32,6 @@ describe.skipIf(SKIP)("handleWikidata", () => {
 		expect(result?.content).toContain("human");
 		expect(result?.content).toContain("Q5");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 });
 
@@ -55,8 +51,6 @@ describe.skipIf(SKIP)("handleOpenLibrary", () => {
 		expect(result).not.toBeNull();
 		expect(result?.method).toBe("openlibrary");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches work OL45804W - The Lord of the Rings", async () => {
@@ -65,8 +59,6 @@ describe.skipIf(SKIP)("handleOpenLibrary", () => {
 		expect(result?.method).toBe("openlibrary");
 		expect(result?.content).toContain("OL45804W");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 });
 
@@ -89,8 +81,6 @@ describe.skipIf(SKIP)("handleBiorxiv", () => {
 		expect(result?.content).toContain("AlphaFold");
 		expect(result?.content).toContain("Abstract");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	// Testing with version suffix handling
diff --git a/packages/coding-agent/test/tools/web-scrapers/security.test.ts b/packages/coding-agent/test/tools/web-scrapers/security.test.ts
index ad7998d3c..9e705c775 100644
--- a/packages/coding-agent/test/tools/web-scrapers/security.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/security.test.ts
@@ -28,8 +28,6 @@ describe.skipIf(SKIP)("handleNvd", () => {
 		expect(result?.content).toContain("Log4j");
 		expect(result?.content).toContain("CVSS");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches CVE-2014-0160 (Heartbleed)", async () => {
@@ -38,7 +36,6 @@ describe.skipIf(SKIP)("handleNvd", () => {
 		expect(result?.method).toBe("nvd");
 		expect(result?.content).toContain("CVE-2014-0160");
 		expect(result?.content).toContain("OpenSSL");
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("handles lowercase CVE IDs", async () => {
@@ -72,8 +69,6 @@ describe.skipIf(SKIP)("handleOsv", () => {
 		expect(result?.content).toContain("GHSA-jfh8-c2jp-5v3q");
 		expect(result?.content).toContain("log4j");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches CVE-2021-44228 via OSV", async () => {
@@ -81,7 +76,6 @@ describe.skipIf(SKIP)("handleOsv", () => {
 		expect(result).not.toBeNull();
 		expect(result?.method).toBe("osv");
 		expect(result?.content).toContain("CVE-2021-44228");
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches PYSEC vulnerability", async () => {
diff --git a/packages/coding-agent/test/tools/web-scrapers/social-extended.test.ts b/packages/coding-agent/test/tools/web-scrapers/social-extended.test.ts
index 278aaf529..4f5c5cb0f 100644
--- a/packages/coding-agent/test/tools/web-scrapers/social-extended.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/social-extended.test.ts
@@ -28,8 +28,6 @@ describe.skipIf(SKIP)("handleMastodon", () => {
 			expect(result?.content).toContain("**Followers:**");
 			expect(result?.content).toContain("**Following:**");
 			expect(result?.content).toContain("**Posts:**");
-			expect(result?.fetchedAt).toBeTruthy();
-			expect(result?.truncated).toBeDefined();
 			expect(result?.notes?.[0]).toContain("Mastodon API");
 		},
 		{ timeout: 30000 },
@@ -67,8 +65,6 @@ describe.skipIf(SKIP)("handleBluesky", () => {
 			expect(result?.content).toContain("**Following:**");
 			expect(result?.content).toContain("**Posts:**");
 			expect(result?.content).toContain("**DID:**");
-			expect(result?.fetchedAt).toBeTruthy();
-			expect(result?.truncated).toBeDefined();
 			expect(result?.notes).toContain("Fetched via AT Protocol API");
 		},
 		{ timeout: 30000 },
@@ -84,8 +80,6 @@ describe.skipIf(SKIP)("handleBluesky", () => {
 			expect(result?.contentType).toBe("text/markdown");
 			expect(result?.content).toContain("@jay.bsky.team");
 			expect(result?.content).toContain("**Followers:**");
-			expect(result?.fetchedAt).toBeTruthy();
-			expect(result?.truncated).toBeDefined();
 		},
 		{ timeout: 30000 },
 	);
diff --git a/packages/coding-agent/test/tools/web-scrapers/stackexchange.test.ts b/packages/coding-agent/test/tools/web-scrapers/stackexchange.test.ts
index ae4c9a4fe..9efd13f8c 100644
--- a/packages/coding-agent/test/tools/web-scrapers/stackexchange.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/stackexchange.test.ts
@@ -29,8 +29,6 @@ describe.skipIf(SKIP)("handleStackOverflow", () => {
 		expect(result?.method).toBe("stackexchange");
 		expect(result?.content).toContain("NullPointerException");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 		expect(result?.notes?.[0]).toContain("site=stackoverflow");
 	});
 
@@ -44,7 +42,6 @@ describe.skipIf(SKIP)("handleStackOverflow", () => {
 		expect(result?.method).toBe("stackexchange");
 		expect(result?.content).toContain("whitespace");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
 		expect(result?.notes?.[0]).toContain("site=unix");
 	});
 
@@ -58,7 +55,6 @@ describe.skipIf(SKIP)("handleStackOverflow", () => {
 		expect(result?.method).toBe("stackexchange");
 		expect(result?.content).toContain("PATH");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
 		expect(result?.notes?.[0]).toContain("site=superuser");
 	});
 
@@ -72,7 +68,6 @@ describe.skipIf(SKIP)("handleStackOverflow", () => {
 		expect(result?.method).toBe("stackexchange");
 		expect(result?.content).toContain("apt");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
 		expect(result?.notes?.[0]).toContain("site=askubuntu");
 	});
 
@@ -86,7 +81,6 @@ describe.skipIf(SKIP)("handleStackOverflow", () => {
 		expect(result?.method).toBe("stackexchange");
 		expect(result?.content).toMatch(/proxy/i);
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
 		expect(result?.notes?.[0]).toContain("site=serverfault");
 	});
 
@@ -96,14 +90,8 @@ describe.skipIf(SKIP)("handleStackOverflow", () => {
 	it("returns complete response structure", async () => {
 		const result = await handleStackOverflow("https://stackoverflow.com/questions/218384", 20);
 		expect(result).not.toBeNull();
-		expect(result).toHaveProperty("url");
-		expect(result).toHaveProperty("finalUrl");
 		expect(result).toHaveProperty("contentType", "text/markdown");
 		expect(result).toHaveProperty("method", "stackexchange");
-		expect(result).toHaveProperty("content");
-		expect(result).toHaveProperty("fetchedAt");
-		expect(result).toHaveProperty("truncated");
-		expect(result).toHaveProperty("notes");
 		// Content should have question structure
 		expect(result?.content).toContain("# ");
 		expect(result?.content).toContain("Score:");
diff --git a/packages/coding-agent/test/tools/web-scrapers/standards.test.ts b/packages/coding-agent/test/tools/web-scrapers/standards.test.ts
index e716e83e7..9ad8069e1 100644
--- a/packages/coding-agent/test/tools/web-scrapers/standards.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/standards.test.ts
@@ -23,8 +23,6 @@ describe.skipIf(SKIP)("handleRfc", () => {
 		expect(result?.content).toContain("HTTP/1.1");
 		expect(result?.content).toContain("Hypertext Transfer Protocol");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches RFC 2616 via datatracker URL", async () => {
@@ -66,8 +64,6 @@ describe.skipIf(SKIP)("handleCheatSh", () => {
 		expect(result?.method).toBe("cheat.sh");
 		expect(result?.content).toContain("curl");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches tar cheatsheet", async () => {
@@ -102,8 +98,6 @@ describe.skipIf(SKIP)("handleTldr", () => {
 		expect(result?.method).toBe("tldr");
 		expect(result?.content).toContain("git");
 		expect(result?.contentType).toBe("text/markdown");
-		expect(result?.fetchedAt).toBeTruthy();
-		expect(result?.truncated).toBeDefined();
 	});
 
 	it("fetches curl tldr page", async () => {
diff --git a/packages/coding-agent/test/tools/web-scrapers/wikipedia.test.ts b/packages/coding-agent/test/tools/web-scrapers/wikipedia.test.ts
index cdbaa3fb4..f01ebec53 100644
--- a/packages/coding-agent/test/tools/web-scrapers/wikipedia.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/wikipedia.test.ts
@@ -25,10 +25,6 @@ describe.skipIf(SKIP)("handleWikipedia", () => {
 		expect(result?.finalUrl).toBe("https://en.wikipedia.org/wiki/Computer");
 		expect(result?.truncated).toBe(false);
 		expect(result?.notes).toContain("Fetched via Wikipedia API");
-		expect(result?.fetchedAt).toBeDefined();
-		// Should be a valid ISO timestamp
-		expect(() => new Date(result?.fetchedAt ?? "")).not.toThrow();
-		// The handler should filter out References and External links sections
 		const content = result?.content ?? "";
 		const hasReferencesHeading = /^## References$/m.test(content);
 		const hasExternalLinksHeading = /^## External links$/m.test(content);
diff --git a/packages/coding-agent/test/tools/web-scrapers/youtube-parallel.test.ts b/packages/coding-agent/test/tools/web-scrapers/youtube-parallel.test.ts
index 42de8a4c8..60e041930 100644
--- a/packages/coding-agent/test/tools/web-scrapers/youtube-parallel.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/youtube-parallel.test.ts
@@ -8,7 +8,7 @@ describe("handleYouTube with Parallel extract", () => {
 	beforeEach(async () => {
 		resetSettingsForTest();
 		process.env.PARALLEL_API_KEY = "test-parallel-key";
-		await Settings.init({ inMemory: true, overrides: { "providers.parallelFetch": true } });
+		await Settings.init({ inMemory: true, overrides: { "providers.fetch": "auto" } });
 	});
 
 	afterEach(() => {
diff --git a/packages/coding-agent/test/tools/web-scrapers/youtube.test.ts b/packages/coding-agent/test/tools/web-scrapers/youtube.test.ts
index 88a6effd2..493280f41 100644
--- a/packages/coding-agent/test/tools/web-scrapers/youtube.test.ts
+++ b/packages/coding-agent/test/tools/web-scrapers/youtube.test.ts
@@ -112,18 +112,6 @@ describe.skipIf(SKIP)("handleYouTube", () => {
 		}
 	}, 30000);
 
-	it("handles videos without transcripts gracefully", async () => {
-		// Many music videos lack captions, but this is not guaranteed
-		// Just verify the handler doesn't crash and provides some info
-		const result = await handleYouTube("https://www.youtube.com/watch?v=kJQP7kiw5Fk", 30);
-		expect(result).not.toBeNull();
-
-		if (result?.method === "youtube") {
-			// Should still have basic metadata
-			expect(result.content).toContain("Video ID");
-		}
-	}, 30000);
-
 	it("returns appropriate response when yt-dlp is not available", async () => {
 		// We can't force yt-dlp to be unavailable in tests, but we can verify
 		// the return structure matches expectations for both cases
@@ -132,13 +120,7 @@ describe.skipIf(SKIP)("handleYouTube", () => {
 
 		// Should have one of these methods
 		expect(["parallel", "youtube", "youtube-no-ytdlp"]).toContain(result.method);
-
-		// Both should have required fields
 		expect(result.url).toBe("https://www.youtube.com/watch?v=dQw4w9WgXcQ");
-		expect(result.finalUrl).toContain("youtube.com");
-		expect(result.fetchedAt).toBeTruthy();
-		expect(typeof result.truncated).toBe("boolean");
-		expect(Array.isArray(result.notes)).toBe(true);
 	}, 30000);
 
 	it("normalizes video URLs to canonical format", async () => {
diff --git a/packages/coding-agent/test/tools/web-search-parallel.test.ts b/packages/coding-agent/test/tools/web-search-parallel.test.ts
index 2fa5f6d55..1886417cb 100644
--- a/packages/coding-agent/test/tools/web-search-parallel.test.ts
+++ b/packages/coding-agent/test/tools/web-search-parallel.test.ts
@@ -30,6 +30,12 @@ describe("Parallel web search", () => {
 		hasAuth() {
 			return Boolean(process.env.PARALLEL_API_KEY);
 		},
+		resolver(_provider: string) {
+			return async () => process.env.PARALLEL_API_KEY ?? undefined;
+		},
+		async rotateSessionCredential() {
+			return false;
+		},
 	} as unknown as AuthStorage;
 
 	let capturedRequestBody: unknown;
diff --git a/packages/coding-agent/test/tools/web-search-tavily.test.ts b/packages/coding-agent/test/tools/web-search-tavily.test.ts
index a260179c7..bea6429f1 100644
--- a/packages/coding-agent/test/tools/web-search-tavily.test.ts
+++ b/packages/coding-agent/test/tools/web-search-tavily.test.ts
@@ -21,6 +21,12 @@ describe("Tavily web search provider", () => {
 		hasAuth() {
 			return Boolean(process.env.TAVILY_API_KEY);
 		},
+		resolver(_provider: string) {
+			return async () => process.env.TAVILY_API_KEY ?? undefined;
+		},
+		async rotateSessionCredential() {
+			return false;
+		},
 	} as unknown as AuthStorage;
 
 	function makeParams(query: string) {
diff --git a/packages/coding-agent/test/ttsr.test.ts b/packages/coding-agent/test/ttsr.test.ts
index 2053cdc9c..4f04ad3e0 100644
--- a/packages/coding-agent/test/ttsr.test.ts
+++ b/packages/coding-agent/test/ttsr.test.ts
@@ -1,9 +1,21 @@
 import { describe, expect, it } from "bun:test";
 import * as path from "node:path";
 import { parseRuleConditionAndScope, type Rule } from "@oh-my-pi/pi-coding-agent/capability/rule";
+import type { TtsrSettings } from "@oh-my-pi/pi-coding-agent/config/settings";
 import { EDIT_MODE_STRATEGIES } from "@oh-my-pi/pi-coding-agent/edit";
 import { TtsrManager } from "@oh-my-pi/pi-coding-agent/export/ttsr";
 
+function ttsrManager(overrides: Partial<TtsrSettings> = {}): TtsrManager {
+	return new TtsrManager({
+		enabled: true,
+		contextMode: "discard",
+		interruptMode: "always",
+		repeatMode: "once",
+		repeatGap: 10,
+		...overrides,
+	});
+}
+
 function makeRule(partial: Partial<Rule>): Rule {
 	return {
 		name: partial.name ?? "rule",
@@ -274,6 +286,52 @@ describe("TtsrManager scope matching", () => {
 	});
 });
 
+describe("TtsrManager enabled gate", () => {
+	it("rejects registration when ttsr is disabled", () => {
+		const manager = ttsrManager({ enabled: false });
+		const rule = makeRule({
+			name: "no-foo",
+			condition: ["FORBIDDEN"],
+			scope: ["text"],
+		});
+
+		expect(manager.addRule(rule)).toBe(false);
+	});
+
+	it("reports no rules when ttsr is disabled, even after a registration attempt", () => {
+		const manager = ttsrManager({ enabled: false });
+		manager.addRule(
+			makeRule({
+				name: "no-foo",
+				condition: ["FORBIDDEN"],
+				scope: ["text"],
+			}),
+		);
+
+		expect(manager.hasRules()).toBe(false);
+	});
+
+	it("returns no matches from stream deltas when ttsr is disabled", () => {
+		const manager = ttsrManager({ enabled: false });
+
+		expect(manager.checkDelta("contains FORBIDDEN token", { source: "text" })).toEqual([]);
+		expect(manager.checkDelta("FORBIDDEN", { source: "tool", toolName: "edit" })).toEqual([]);
+	});
+
+	it("preserves the default (enabled) registration and matching contract", () => {
+		const manager = ttsrManager();
+		const rule = makeRule({
+			name: "no-foo",
+			condition: ["FORBIDDEN"],
+			scope: ["text"],
+		});
+
+		expect(manager.addRule(rule)).toBe(true);
+		expect(manager.hasRules()).toBe(true);
+		expect(manager.checkDelta("FORBIDDEN", { source: "text" })).toEqual([rule]);
+	});
+});
+
 describe("TtsrManager snapshot matching", () => {
 	it("matches source-level conditions against a tool digest where the raw patch grammar fails", () => {
 		const manager = new TtsrManager();
@@ -291,7 +349,7 @@ describe("TtsrManager snapshot matching", () => {
 			streamKey: "toolcall:tc-1",
 		};
 		const patch = [
-			"¶src/repo.ts#AB12",
+			"[src/repo.ts#AB12]",
 			"replace block 1:",
 			"+export async function isRepository(cwd: string): Promise<boolean> {",
 			"+\treturn repo.isRepository(cwd);",
diff --git a/packages/coding-agent/test/utils/enhanced-paste.test.ts b/packages/coding-agent/test/utils/enhanced-paste.test.ts
new file mode 100644
index 000000000..d319c1362
--- /dev/null
+++ b/packages/coding-agent/test/utils/enhanced-paste.test.ts
@@ -0,0 +1,109 @@
+import { describe, expect, it } from "bun:test";
+import { EnhancedPasteController } from "../../src/utils/enhanced-paste";
+
+const ST = "\x1b\\";
+const OSC = "\x1b]5522;";
+
+function packet(metadata: string, payload?: string): string {
+	return `${OSC}${metadata}${payload === undefined ? "" : `;${payload}`}${ST}`;
+}
+
+describe("EnhancedPasteController", () => {
+	it("requests image data from an OSC 5522 paste event and preserves chunk boundaries", () => {
+		const writes: string[] = [];
+		const pastedImages: Array<{ data: string; mimeType: string }> = [];
+		const statuses: string[] = [];
+		const controller = new EnhancedPasteController({
+			write: data => writes.push(data),
+			pasteText: () => statuses.push("unexpected text paste"),
+			pasteImage: image => {
+				pastedImages.push({ data: image.data, mimeType: image.mimeType });
+			},
+			showStatus: message => statuses.push(message),
+		});
+
+		controller.enable();
+		expect(writes).toEqual(["\x1b[?5522h"]);
+
+		const imageMime = Buffer.from("image/png", "utf8").toString("base64");
+		const textMime = Buffer.from("text/plain", "utf8").toString("base64");
+		const password = Buffer.from("secret123", "utf8").toString("base64");
+		controller.handleInput(packet(`type=read:status=OK:pw=${password}`));
+		controller.handleInput(packet(`type=read:status=DATA:mime=${textMime}`));
+		controller.handleInput(packet(`type=read:status=DATA:mime=${imageMime}`));
+		controller.handleInput(packet("type=read:status=DONE"));
+
+		const pasteEventName = Buffer.from("Paste event", "utf8").toString("base64");
+		expect(writes.at(-1)).toBe(`${OSC}type=read:mime=${imageMime}:pw=${password}:name=${pasteEventName}${ST}`);
+
+		controller.handleInput(packet("type=read:status=OK"));
+		controller.handleInput(
+			packet(`type=read:status=DATA:mime=${imageMime}`, Buffer.from("image-", "utf8").toString("base64")),
+		);
+		controller.handleInput(
+			packet(`type=read:status=DATA:mime=${imageMime}`, Buffer.from("bytes", "utf8").toString("base64")),
+		);
+		controller.handleInput(packet("type=read:status=DONE"));
+
+		expect(pastedImages).toEqual([
+			{
+				data: Buffer.from("image-bytes", "utf8").toString("base64"),
+				mimeType: "image/png",
+			},
+		]);
+		expect(statuses).toEqual([]);
+	});
+
+	it("falls back to text/plain and carries primary-selection location into the read request", () => {
+		const writes: string[] = [];
+		const pastedText: string[] = [];
+		const controller = new EnhancedPasteController({
+			write: data => writes.push(data),
+			pasteText: text => pastedText.push(text),
+			pasteImage: () => {
+				throw new Error("unexpected image paste");
+			},
+			showStatus: message => pastedText.push(`status:${message}`),
+		});
+
+		const textMime = Buffer.from("text/plain", "utf8").toString("base64");
+		const password = Buffer.from("secret456", "utf8").toString("base64");
+		expect(controller.handleInput("plain text")).toBe(false);
+		controller.handleInput(packet(`type=read:status=OK:loc=primary:pw=${password}`));
+		controller.handleInput(packet(`type=read:status=DATA:mime=${textMime}`));
+		controller.handleInput(packet("type=read:status=DONE"));
+
+		expect(writes).toHaveLength(1);
+		expect(writes[0]).toContain(`mime=${textMime}`);
+		expect(writes[0]).toContain("loc=primary");
+		expect(writes[0]).toContain(`pw=${password}`);
+
+		controller.handleInput(packet("type=read:status=OK"));
+		controller.handleInput(
+			packet(`type=read:status=DATA:mime=${textMime}`, Buffer.from("hello ", "utf8").toString("base64")),
+		);
+		controller.handleInput(
+			packet(`type=read:status=DATA:mime=${textMime}`, Buffer.from("world", "utf8").toString("base64")),
+		);
+		controller.handleInput(packet("type=read:status=DONE"));
+
+		expect(pastedText).toEqual(["hello world"]);
+	});
+
+	it("reports unsupported paste events instead of leaking OSC packets to the editor", () => {
+		const statuses: string[] = [];
+		const controller = new EnhancedPasteController({
+			write: () => {},
+			pasteText: () => {},
+			pasteImage: () => {},
+			showStatus: message => statuses.push(message),
+		});
+
+		const htmlMime = Buffer.from("text/html", "utf8").toString("base64");
+		expect(controller.handleInput(packet("type=read:status=OK"))).toBe(true);
+		expect(controller.handleInput(packet(`type=read:status=DATA:mime=${htmlMime}`))).toBe(true);
+		expect(controller.handleInput(packet("type=read:status=DONE"))).toBe(true);
+
+		expect(statuses).toEqual(["Clipboard paste has no supported text or image data"]);
+	});
+});
diff --git a/packages/coding-agent/test/web/search/abort-and-timeout.test.ts b/packages/coding-agent/test/web/search/abort-and-timeout.test.ts
index 9a347d196..2e82e4d27 100644
--- a/packages/coding-agent/test/web/search/abort-and-timeout.test.ts
+++ b/packages/coding-agent/test/web/search/abort-and-timeout.test.ts
@@ -123,6 +123,7 @@ describe("Anthropic provider hard-timeout wiring", () => {
 			systemPrompt: "",
 			authStorage: {
 				getApiKey: async () => "sk-fallback",
+				resolver: vi.fn(() => async () => "sk-fallback"),
 			} as unknown as AuthStorage,
 		});
 
diff --git a/packages/coding-agent/test/web/search/perplexity.test.ts b/packages/coding-agent/test/web/search/perplexity.test.ts
index 1d2ef6781..a582141ab 100644
--- a/packages/coding-agent/test/web/search/perplexity.test.ts
+++ b/packages/coding-agent/test/web/search/perplexity.test.ts
@@ -1,6 +1,6 @@
 import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
 import type { AuthStorage } from "@oh-my-pi/pi-ai";
-import { searchPerplexity } from "@oh-my-pi/pi-coding-agent/web/search/providers/perplexity";
+import { PerplexityProvider, searchPerplexity } from "@oh-my-pi/pi-coding-agent/web/search/providers/perplexity";
 import { hookFetch } from "@oh-my-pi/pi-utils";
 
 const API_URL = "https://api.perplexity.ai/chat/completions";
@@ -97,3 +97,173 @@ describe("Perplexity API-key request shape", () => {
 		expect(response.relatedQuestions).toBeUndefined();
 	});
 });
+
+const OAUTH_ASK_URL = "https://www.perplexity.ai/rest/sse/perplexity_ask";
+
+// OAuth path: getOAuthAccess returns a bearer (no `.`-delimited exp claim, so it
+// is treated as non-expiring), making findPerplexityAuth pick the oauth branch.
+const oauthAuthStorage = {
+	async getOAuthAccess() {
+		return { accessToken: "test-oauth-token" };
+	},
+	hasAuth() {
+		return true;
+	},
+} as unknown as AuthStorage;
+
+const anonymousAuthStorage = {
+	async getOAuthAccess() {
+		return undefined;
+	},
+	hasAuth() {
+		return false;
+	},
+} as unknown as AuthStorage;
+
+function mockOAuth(capture: (body: Record<string, unknown>, headers: Headers) => void) {
+	const event = {
+		final: true,
+		display_model: "turbo",
+		uuid: "req-oauth",
+		blocks: [
+			{ intended_usage: "ask_text", markdown_block: { answer: "OAuth answer" } },
+			{
+				intended_usage: "web_results",
+				web_result_block: { web_results: [{ name: "T", url: "https://example.com", snippet: "s" }] },
+			},
+		],
+	};
+	const sseBody = `data: ${JSON.stringify(event)}\n\n`;
+	return hookFetch(async (input, init) => {
+		const url = typeof input === "string" ? input : input instanceof URL ? input.toString() : input.url;
+		if (url === OAUTH_ASK_URL) {
+			capture(JSON.parse(init?.body as string), new Headers(init?.headers));
+			return new Response(sseBody, { status: 200, headers: { "Content-Type": "text/event-stream" } });
+		}
+		return new Response("not mocked", { status: 500 });
+	});
+}
+
+function mockAnonymous(capture: (body: Record<string, unknown>, headers: Headers) => void) {
+	const answerPayload = {
+		answer: "Anonymous answer",
+		web_results: [{ name: "Example", url: "https://example.com", snippet: "s" }],
+		chunks: ["Anonymous ", "answer"],
+		structured_answer: [{ type: "markdown", text: "Anonymous answer", chunks: ["Anonymous ", "answer"] }],
+	};
+	const event = {
+		final: true,
+		display_model: "turbo",
+		uuid: "req-anon",
+		text: JSON.stringify([{ step_type: "FINAL", content: { answer: JSON.stringify(answerPayload) }, uuid: "" }]),
+	};
+	const sseBody = `data: ${JSON.stringify(event)}\n\n`;
+	return hookFetch(async (input, init) => {
+		const url = typeof input === "string" ? input : input instanceof URL ? input.toString() : input.url;
+		if (url === OAUTH_ASK_URL) {
+			capture(JSON.parse(init?.body as string), new Headers(init?.headers));
+			return new Response(sseBody, { status: 200, headers: { "Content-Type": "text/event-stream" } });
+		}
+		return new Response("not mocked", { status: 500 });
+	});
+}
+
+describe("Perplexity OAuth request shape", () => {
+	const savedCookies = process.env.PERPLEXITY_COOKIES;
+
+	beforeEach(() => {
+		delete process.env.PERPLEXITY_COOKIES; // cookies take precedence over oauth; keep them out
+	});
+
+	afterEach(() => {
+		vi.restoreAllMocks();
+		if (savedCookies === undefined) delete process.env.PERPLEXITY_COOKIES;
+		else process.env.PERPLEXITY_COOKIES = savedCookies;
+	});
+
+	it("sends the bare query, never the API-style system prompt, to the ask endpoint", async () => {
+		let body: Record<string, unknown> | undefined;
+		let headers: Headers | undefined;
+		using _hook = mockOAuth((b, h) => {
+			body = b;
+			headers = h;
+		});
+
+		const response = await searchPerplexity({
+			query: "quic vs tcp",
+			system_prompt: "Research assistant with web search. Synthesize comprehensive answers.",
+			authStorage: oauthAuthStorage,
+		});
+
+		// The consumer ask endpoint has no system slot; prepending the prompt makes
+		// the model refuse ("I don't have web-search tools in this turn").
+		expect(body?.query_str).toBe("quic vs tcp");
+		expect((body?.params as Record<string, unknown>).query_str).toBe("quic vs tcp");
+		expect((body?.params as Record<string, unknown>).model_preference).toBe("experimental");
+		// The ask endpoint authenticates via the next-auth session cookie; a bearer
+		// header is ignored and silently downgrades to the anonymous `turbo` model.
+		expect(headers?.get("cookie")).toBe("__Secure-next-auth.session-token=test-oauth-token");
+		expect(headers?.has("authorization")).toBe(false);
+		expect(response.authMode).toBe("oauth");
+		expect(response.answer).toBe("OAuth answer");
+	});
+});
+
+describe("Perplexity anonymous fallback", () => {
+	const savedKey = process.env.PERPLEXITY_API_KEY;
+	const savedPplxKey = process.env.PPLX_API_KEY;
+	const savedCookies = process.env.PERPLEXITY_COOKIES;
+
+	beforeEach(() => {
+		delete process.env.PERPLEXITY_API_KEY;
+		delete process.env.PPLX_API_KEY;
+		delete process.env.PERPLEXITY_COOKIES;
+	});
+
+	afterEach(() => {
+		vi.restoreAllMocks();
+		if (savedKey === undefined) delete process.env.PERPLEXITY_API_KEY;
+		else process.env.PERPLEXITY_API_KEY = savedKey;
+		if (savedPplxKey === undefined) delete process.env.PPLX_API_KEY;
+		else process.env.PPLX_API_KEY = savedPplxKey;
+		if (savedCookies === undefined) delete process.env.PERPLEXITY_COOKIES;
+		else process.env.PERPLEXITY_COOKIES = savedCookies;
+	});
+
+	it("uses the browser ask endpoint without credential headers when no key is configured", async () => {
+		let body: Record<string, unknown> | undefined;
+		let headers: Headers | undefined;
+		using _hook = mockAnonymous((b, h) => {
+			body = b;
+			headers = h;
+		});
+
+		const response = await searchPerplexity({ query: "anonymous search", authStorage: anonymousAuthStorage });
+		const requestParams = body?.params as Record<string, unknown>;
+
+		expect(headers?.has("authorization")).toBe(false);
+		expect(headers?.has("cookie")).toBe(false);
+		expect(headers?.get("user-agent")).toContain("Mozilla/5.0");
+		expect(requestParams.model_preference).toBe("experimental");
+		expect(requestParams.send_back_text_in_streaming_api).toBe(true);
+		expect(requestParams.source).toBe("default");
+		expect(response.authMode).toBe("anonymous");
+		expect(response.answer).toBe("Anonymous answer");
+		expect(response.sources).toEqual([
+			{
+				title: "Example",
+				url: "https://example.com",
+				snippet: "s",
+				publishedDate: undefined,
+				ageSeconds: undefined,
+			},
+		]);
+	});
+
+	it("keeps anonymous Perplexity out of auto provider selection but allows explicit selection", () => {
+		const provider = new PerplexityProvider();
+
+		expect(provider.isAvailable(anonymousAuthStorage)).toBe(false);
+		expect(provider.isExplicitlyAvailable(anonymousAuthStorage)).toBe(true);
+	});
+});
diff --git a/packages/coding-agent/test/web/search/render.test.ts b/packages/coding-agent/test/web/search/render.test.ts
index d9a26ccc5..e2b689dc7 100644
--- a/packages/coding-agent/test/web/search/render.test.ts
+++ b/packages/coding-agent/test/web/search/render.test.ts
@@ -75,13 +75,25 @@ describe("renderSearchResult", () => {
 		expect(answer).not.toMatch(/more line/);
 	});
 
-	it("truncates the answer with a summary when collapsed", async () => {
+	it("shows the full answer when collapsed by default", async () => {
 		const uiTheme = (await getThemeByName("dark"))!;
 		const component = renderSearchResult(buildResult(ANSWER), { expanded: false, isPartial: false }, uiTheme, {
 			query: "test query",
 		});
 		const answer = answerSection(component.render(120).map(l => sanitizeText(l)));
-		// Collapsed view caps the answer and signals how many lines were hidden.
+		// TUI collapsed view keeps the answer intact; only explicit compact mode caps it.
+		expect(answer).toContain("FINAL_UNIQUE_MARKER");
+		expect(answer).not.toMatch(/more line/);
+	});
+
+	it("truncates the answer only when compact mode provides maxAnswerLines", async () => {
+		const uiTheme = (await getThemeByName("dark"))!;
+		const component = renderSearchResult(buildResult(ANSWER), { expanded: false, isPartial: false }, uiTheme, {
+			query: "test query",
+			maxAnswerLines: 3,
+		});
+		const answer = answerSection(component.render(120).map(l => sanitizeText(l)));
+
 		expect(answer).toMatch(/more line/);
 		expect(answer).not.toContain("FINAL_UNIQUE_MARKER");
 	});
diff --git a/packages/coding-agent/test/web/search/tavily.test.ts b/packages/coding-agent/test/web/search/tavily.test.ts
index 689a0a495..8eb7d2e7a 100644
--- a/packages/coding-agent/test/web/search/tavily.test.ts
+++ b/packages/coding-agent/test/web/search/tavily.test.ts
@@ -57,6 +57,7 @@ describe("Tavily searchTavily request shape (integration)", () => {
 		async getApiKey() {
 			return process.env.TAVILY_API_KEY ?? undefined;
 		},
+		resolver: vi.fn(() => async () => process.env.TAVILY_API_KEY ?? undefined),
 		hasAuth() {
 			return Boolean(process.env.TAVILY_API_KEY);
 		},
diff --git a/packages/coding-agent/test/write-hashline-header.test.ts b/packages/coding-agent/test/write-hashline-header.test.ts
index 2e1e965d2..61b2e0764 100644
--- a/packages/coding-agent/test/write-hashline-header.test.ts
+++ b/packages/coding-agent/test/write-hashline-header.test.ts
@@ -30,7 +30,7 @@ function resultText(result: { content: { type: string; text?: string }[] }): str
 		.join("\n");
 }
 
-const HASHLINE_HEADER_LINE = /^¶(\S+)#([0-9A-F]{4})$/;
+const HASHLINE_HEADER_LINE = /^\[([^#\r\n]+)#([0-9A-F]{4})\]$/;
 
 describe("write tool hashline header", () => {
 	let tmpDir: string;
@@ -47,7 +47,7 @@ describe("write tool hashline header", () => {
 		await fs.rm(tmpDir, { recursive: true, force: true });
 	});
 
-	it("insert heads a fresh ¶path#TAG header that maps to the written content", async () => {
+	it("inserts a fresh [path#TAG] header that maps to the written content", async () => {
 		const filePath = path.join(tmpDir, "module.ts");
 		const session = createSession(tmpDir);
 		const tool = new WriteTool(session);
@@ -111,7 +111,7 @@ describe("write tool hashline header", () => {
 
 		const result = await tool.execute("call-1", { path: filePath, content });
 		const text = resultText(result);
-		expect(text.startsWith("¶")).toBe(false);
+		expect(text.startsWith("[")).toBe(false);
 		expect(text).toBe(`Successfully wrote ${content.length} bytes to ${path.relative(tmpDir, filePath)}`);
 	});
 });
diff --git a/packages/coding-agent/test/write-streaming-preview-expand.test.ts b/packages/coding-agent/test/write-streaming-preview-expand.test.ts
new file mode 100644
index 000000000..cbbd0db8e
--- /dev/null
+++ b/packages/coding-agent/test/write-streaming-preview-expand.test.ts
@@ -0,0 +1,59 @@
+import { afterEach, describe, expect, it, vi } from "bun:test";
+import { initTheme } from "@oh-my-pi/pi-coding-agent/modes/theme/theme";
+import type { TUI } from "@oh-my-pi/pi-tui";
+import { ToolExecutionComponent } from "../src/modes/components/tool-execution";
+
+const stripAnsi = (s: string): string => s.replace(/\u001b\[[0-9;]*m/g, "");
+const hasLine = (lines: string[], n: number): boolean =>
+	new RegExp(`\\bline ${n}\\b`).test(stripAnsi(lines.join("\n")));
+
+describe("write streaming preview honors Ctrl+O expansion", () => {
+	let initialized = false;
+
+	afterEach(() => {
+		vi.restoreAllMocks();
+	});
+
+	async function makePendingWrite(lineCount: number) {
+		if (!initialized) {
+			await initTheme();
+			initialized = true;
+		}
+		const uiStub = { requestRender() {} } as unknown as TUI;
+		const content = Array.from({ length: lineCount }, (_, i) => `line ${i + 1}`).join("\n");
+		// No updateResult() -> the call stays pending, exercising the streaming
+		// `renderCall` path (formatStreamingContent), not the merged result render.
+		return new ToolExecutionComponent("write", { file_path: "/tmp/foo.ts", content }, {}, undefined, uiStub);
+	}
+
+	it("collapses a streaming write to a bounded tail and lifts the cap on expand", async () => {
+		// 40 lines > WRITE_STREAMING_PREVIEW_LINES (12): the head must be hidden
+		// while collapsed and the streaming edge (tail) kept visible.
+		const comp = await makePendingWrite(40);
+
+		const collapsed = comp.render(80);
+		// Tail-anchored: the streaming edge (last lines) is visible...
+		expect(hasLine(collapsed, 40)).toBe(true);
+		// ...but the head is capped away with an "earlier lines" marker.
+		expect(hasLine(collapsed, 1)).toBe(false);
+		expect(stripAnsi(collapsed.join("\n"))).toContain("earlier line");
+
+		comp.setExpanded(true);
+		const expanded = comp.render(80);
+		// Ctrl+O lifts the cap: the full file (head through tail) is shown,
+		// and the "earlier lines" marker is gone.
+		expect(hasLine(expanded, 1)).toBe(true);
+		expect(hasLine(expanded, 40)).toBe(true);
+		expect(stripAnsi(expanded.join("\n"))).not.toContain("earlier line");
+		// Expanding must strictly grow the preview, not just reformat it.
+		expect(expanded.length).toBeGreaterThan(collapsed.length);
+	});
+
+	it("does not cap a short streaming write that already fits the window", async () => {
+		const comp = await makePendingWrite(4);
+		const collapsed = comp.render(80);
+		expect(hasLine(collapsed, 1)).toBe(true);
+		expect(hasLine(collapsed, 4)).toBe(true);
+		expect(stripAnsi(collapsed.join("\n"))).not.toContain("earlier line");
+	});
+});
diff --git a/packages/hashline/CHANGELOG.md b/packages/hashline/CHANGELOG.md
index f4f5fbf63..42e6dc391 100644
--- a/packages/hashline/CHANGELOG.md
+++ b/packages/hashline/CHANGELOG.md
@@ -2,6 +2,16 @@
 
 ## [Unreleased]
 
+## [15.9.67] - 2026-06-06
+
+### Breaking Changes
+
+- Changed hashline file section headers from `¶PATH#TAG` to `[PATH#TAG]` so model-authored edits use ASCII delimiters instead of a pilcrow sigil.
+
+### Fixed
+
+- Fixed missing-header diagnostics and copied-content prefix stripping to consistently teach and recognize 4-hex snapshot tags.
+
 ## [15.8.2] - 2026-06-03
 
 ### Fixed
diff --git a/packages/hashline/README.md b/packages/hashline/README.md
index b98cc7e3e..545f98826 100644
--- a/packages/hashline/README.md
+++ b/packages/hashline/README.md
@@ -23,10 +23,10 @@ const snapshots = new InMemorySnapshotStore();
 const before = `const greeting = "hi";\nexport { greeting };\n`;
 await fs.writeText("hello.ts", before);
 
-const tag = snapshots.recordContiguous("hello.ts", 1, before.split("\n"), { fullText: before });
+const tag = snapshots.record("hello.ts", before);
 const patcher = new Patcher({ fs, snapshots });
-const patch = Patch.parse(String.raw`¶hello.ts#${tag}
-@@ 1..1 @@
+const patch = Patch.parse(String.raw`[hello.ts#${tag}]
+replace 1..1:
 +const greeting = "hello";`);
 const result = await patcher.apply(patch);
 
@@ -39,19 +39,19 @@ console.log(await fs.readText("hello.ts"));
 See [`src/prompt.md`](./src/prompt.md) for the user-facing description and
 [`src/grammar.lark`](./src/grammar.lark) for the formal grammar.
 
-Each file section starts with `¶PATH#TAG`. The tag is a 3-hex opaque
-pointer into the `SnapshotStore` that minted it; it is not content-derived
-and is not meaningful outside that store. The patcher protects against
-stale anchors by resolving the tag, verifying the recorded snapshot lines
-against live file content, and refusing or attempting session-aware
-recovery on mismatch.
+Each file section starts with `[PATH#TAG]`. The tag is a 4-hex
+content hash of the full normalized file text recorded by the
+`SnapshotStore`, and it is not meaningful outside that store. The patcher
+protects against stale anchors by resolving the tag, verifying the live file
+still matches the recorded content hash, and refusing or attempting
+session-aware recovery on mismatch.
 
 Inside a section:
-- `@@ A..B @@` — open a hunk on lines A..B (use `@@ A,A @@` for a single line; bare `@@ A @@` is also accepted).
-- `@@ BOF @@` / `@@ EOF @@` — virtual hunks at the beginning/end of file.
+- `replace A..B:` — replace lines A..B with following `+TEXT` body rows.
+- `replace block A:` — replace the syntactic block beginning on line A.
+- `delete A..B` / `delete block A` — delete concrete lines or a resolved block.
+- `insert before A:` / `insert after A:` / `insert head:` / `insert tail:` — insert following body rows.
 - `+TEXT` — literal body row (use `+` alone for a blank line).
-- `&A..B` — repeat original file lines A..B inline (`&A` for one line).
-- Empty body — delete the selected range.
 
 ## Abstractions
 
@@ -67,9 +67,10 @@ text-document protocol, a Git tree, anything.
 
 ### `SnapshotStore`
 
-Required. Hashline tags are opaque store pointers, so `Patcher` must receive
-the store that minted them. Recovery replays edits against the cached pre-edit
-snapshot and 3-way-merges onto current content when the live file diverged.
+Required. Hashline tags are full-file content hashes recorded per path, so
+`Patcher` must receive the store that observed them. Recovery replays edits
+against the cached pre-edit snapshot and 3-way-merges onto current content
+when the live file diverged.
 
 ### `Patcher`
 
diff --git a/packages/hashline/package.json b/packages/hashline/package.json
index b6237aba2..cf4a8dd30 100644
--- a/packages/hashline/package.json
+++ b/packages/hashline/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/hashline",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "Hashline: a compact, line-anchored patch language and applier. Pluggable FS/IO so it works over disk, in-memory, or any custom backend.",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
diff --git a/packages/hashline/src/format.ts b/packages/hashline/src/format.ts
index 2f4df998e..ef752aa97 100644
--- a/packages/hashline/src/format.ts
+++ b/packages/hashline/src/format.ts
@@ -6,8 +6,9 @@
 
 import type { Cursor } from "./types";
 
-/** File-section header prefix: `¶path#hash`. */
-export const HL_FILE_PREFIX = "¶";
+/** File-section header delimiters: `[path#hash]`. */
+export const HL_FILE_PREFIX = "[";
+export const HL_FILE_SUFFIX = "]";
 
 /** Payload sigil for literal body rows. */
 export const HL_PAYLOAD_REPLACE = "+";
@@ -118,7 +119,7 @@ export function describeAnchorExamples(linePrefix = ""): string {
 
 /** Format a hashline section header for a file path and snapshot tag. */
 export function formatHashlineHeader(filePath: string, fileHash: string): string {
-	return `${HL_FILE_PREFIX}${filePath}${HL_FILE_HASH_SEP}${fileHash}`;
+	return `${HL_FILE_PREFIX}${filePath}${HL_FILE_HASH_SEP}${fileHash}${HL_FILE_SUFFIX}`;
 }
 
 /** Formats a single numbered line as `LINE:TEXT`. */
diff --git a/packages/hashline/src/grammar.lark b/packages/hashline/src/grammar.lark
index 8a056d4e1..ae11cb32b 100644
--- a/packages/hashline/src/grammar.lark
+++ b/packages/hashline/src/grammar.lark
@@ -3,9 +3,9 @@ begin_patch: "*** Begin Patch" LF
 end_patch:   "*** End Patch" LF?
 
 file_patch: file_header hunk+
-file_header: "¶" filename "#" file_hash LF
+file_header: "[" filename "#" file_hash "]" LF
 file_hash: /[0-9A-F]{4}/
-filename: /[^\s#]+/
+filename: /[^#\r\n]+/
 
 hunk: replace_hunk | replace_block_hunk | insert_hunk | delete_hunk | delete_block_hunk
 replace_hunk: replace_anchor LF emit_op*
diff --git a/packages/hashline/src/input.ts b/packages/hashline/src/input.ts
index 9864d95b6..3513d1aea 100644
--- a/packages/hashline/src/input.ts
+++ b/packages/hashline/src/input.ts
@@ -1,6 +1,6 @@
 /**
  * Top-level patch parser. Splits an authored hashline input into a list of
- * {@link PatchSection}s, each rooted at a `¶PATH#HASH` header, then exposes
+ * {@link PatchSection}s, each rooted at a `[PATH#HASH]` header, then exposes
  * a {@link Patch} class that gives lazy access to the parsed edits per
  * section.
  *
@@ -10,7 +10,7 @@
 import * as path from "node:path";
 import { applyEdits } from "./apply";
 import { resolveBlockEdits } from "./block";
-import { HL_FILE_HASH_LENGTH, HL_FILE_HASH_SEP, HL_FILE_PREFIX } from "./format";
+import { HL_FILE_HASH_EXAMPLES, HL_FILE_HASH_LENGTH, HL_FILE_HASH_SEP, HL_FILE_PREFIX, HL_FILE_SUFFIX } from "./format";
 import { parsePatch, parsePatchStreaming } from "./parser";
 import { Tokenizer } from "./tokenizer";
 import type { ApplyResult, BlockResolver, Edit, SplitOptions } from "./types";
@@ -47,21 +47,41 @@ function stripApplyPatchPathNoise(pathText: string): string {
 }
 
 /**
- * Best-effort recovery for `¶`-prefixed lines the strict tokenizer
+ * Best-effort recovery for bracketed header lines the strict tokenizer
  * rejects. Strips apply_patch keyword noise (`Update File:`, `Update:`,
- * etc.) and an extra leading `***` (some models emit a hybrid `¶***foo.ts`
- * shape), then expects `PATH(#HASH)?` with no embedded whitespace.
+ * etc.) and an extra leading `***` (some models emit a hybrid
+ * `[***foo.ts#HASH]` shape), then expects `PATH(#HASH)?`.
  * Returns `null` when no clean path can be salvaged.
  */
 function tryParseRecoveryHeader(line: string, cwd?: string): RawSection | null {
-	if (!line.startsWith(HL_FILE_PREFIX)) return null;
-	const body = stripApplyPatchPathNoise(line.slice(HL_FILE_PREFIX.length).trim());
+	if (!line.startsWith(HL_FILE_PREFIX) || !line.endsWith(HL_FILE_SUFFIX)) return null;
+	const body = stripApplyPatchPathNoise(line.slice(HL_FILE_PREFIX.length, line.length - HL_FILE_SUFFIX.length).trim());
 	if (body.length === 0) return null;
-	const match = new RegExp(`^(\\S+?)(?:#([0-9A-Fa-f]{${HL_FILE_HASH_LENGTH}}))?\\s*$`).exec(body);
-	if (match === null) return null;
-	const path = normalizeHashlinePath(match[1], cwd);
+
+	// Trailing `#XXXX` is the tag; everything before it is the path. The
+	// path may contain whitespace (Windows OneDrive folders, Program Files,
+	// etc.), so we anchor the tag at end-of-body rather than scanning
+	// forward and stopping at the first space.
+	const trailing = new RegExp(`#([0-9A-Fa-f]{${HL_FILE_HASH_LENGTH}})\\s*$`).exec(body);
+	let pathText: string;
+	let fileHash: string | undefined;
+	if (trailing !== null) {
+		pathText = body.slice(0, trailing.index);
+		fileHash = trailing[1].toUpperCase();
+	} else {
+		pathText = body.replace(/\s+$/, "");
+	}
+
+	// Same rule as the strict tokenizer: the hashline header grammar uses
+	// `#` as the path/tag separator and does not allow `#` inside
+	// filenames. Anything `#` left in the path body — short tags, non-hex
+	// tags, over-long tags, stale-tag copy-paste, line-suffixed tags —
+	// means the header is malformed, not a path with an embedded hash.
+	if (pathText.includes("#")) return null;
+
+	const path = normalizeHashlinePath(pathText, cwd);
 	if (path.length === 0) return null;
-	return match[2] !== undefined ? { path, fileHash: match[2].toUpperCase(), diff: "" } : { path, diff: "" };
+	return fileHash !== undefined ? { path, fileHash, diff: "" } : { path, diff: "" };
 }
 
 function normalizeHashlinePath(rawPath: string, cwd?: string): string {
@@ -79,9 +99,9 @@ interface RawSection {
 }
 
 /**
- * Parse a `¶PATH[#hash]` header line. Returns `null` for lines that do
- * not start with `¶`. Throws the strict "Input header must be …" error
- * when a `¶`-prefixed line fails the strict shape (so malformed paths
+ * Parse a `[PATH]` or `[PATH#hash]` header line. Returns `null` for lines that do
+ * not start with `[`. Throws the strict "Input header must be …" error
+ * when a bracketed line fails the strict shape (so malformed paths
  * surface immediately instead of being silently re-classified as payload).
  */
 function parseHashlineHeaderLine(line: string, cwd?: string): RawSection | null {
@@ -91,18 +111,18 @@ function parseHashlineHeaderLine(line: string, cwd?: string): RawSection | null
 	const token = TOKENIZER.tokenize(trimmed);
 	if (token.kind !== "header") {
 		// Recovery: try to extract a path from the raw line after stripping
-		// apply_patch noise. This handles `*** Update File:foo.ts#CB5` and
+		// apply_patch noise. This handles `[*** Update File:foo.ts#CB5A]` and
 		// the half-dozen variants models actually emit.
 		const recovered = tryParseRecoveryHeader(trimmed, cwd);
 		if (recovered !== null) return recovered;
 		throw new Error(
-			`Input header must be ${HL_FILE_PREFIX}PATH or ${HL_FILE_PREFIX}PATH${HL_FILE_HASH_SEP}TAG with a ${HL_FILE_HASH_LENGTH}-hex content-hash tag; got ${JSON.stringify(trimmed)}.`,
+			`Input header must be ${HL_FILE_PREFIX}PATH${HL_FILE_SUFFIX} or ${HL_FILE_PREFIX}PATH${HL_FILE_HASH_SEP}TAG${HL_FILE_SUFFIX} with a ${HL_FILE_HASH_LENGTH}-hex content-hash tag; got ${JSON.stringify(trimmed)}.`,
 		);
 	}
 
 	const parsedPath = normalizeHashlinePath(token.path, cwd);
 	if (parsedPath.length === 0) {
-		throw new Error(`Input header "${HL_FILE_PREFIX}" is empty; provide a file path.`);
+		throw new Error(`Input header "${HL_FILE_PREFIX}${HL_FILE_SUFFIX}" is empty; provide a file path.`);
 	}
 	return token.fileHash !== undefined
 		? { path: parsedPath, fileHash: token.fileHash, diff: "" }
@@ -145,7 +165,7 @@ function normalizeFallbackInput(input: string, options: SplitOptions): string {
 	if (!options.path || !containsRecognizableHashlineOperations(input)) return input;
 	const fallbackPath = normalizeHashlinePath(options.path, options.cwd);
 	if (fallbackPath.length === 0) return input;
-	return `${HL_FILE_PREFIX}${fallbackPath}\n${input}`;
+	return `${HL_FILE_PREFIX}${fallbackPath}${HL_FILE_SUFFIX}\n${input}`;
 }
 
 function splitRawSections(input: string, options: SplitOptions = {}): RawSection[] {
@@ -160,13 +180,13 @@ function splitRawSections(input: string, options: SplitOptions = {}): RawSection
 		if (/^@@\s+[-+]?\d+,\d+\s+[-+]?\d+,\d+\s+@@/.test(firstTrimmed)) {
 			throw new Error(
 				"unified-diff hunk header (`@@ -N,M +N,M @@`) is not valid in hashline. " +
-					"File sections start with `¶path#HASH`; use `replace`, `delete`, or `insert` ops.",
+					`File sections start with \`${HL_FILE_PREFIX}path${HL_FILE_HASH_SEP}HASH${HL_FILE_SUFFIX}\`; use \`replace\`, \`delete\`, or \`insert\` ops.`,
 			);
 		}
 		const preview = JSON.stringify(firstLine.slice(0, 120));
 		throw new Error(
-			`input must begin with "${HL_FILE_PREFIX}PATH${HL_FILE_HASH_SEP}HASH" on the first non-blank line for anchored edits; got: ${preview}. ` +
-				`Example: "${HL_FILE_PREFIX}src/foo.ts${HL_FILE_HASH_SEP}0A3" then edit ops.`,
+			`input must begin with "${HL_FILE_PREFIX}PATH${HL_FILE_HASH_SEP}HASH${HL_FILE_SUFFIX}" on the first non-blank line for anchored edits; got: ${preview}. ` +
+				`Example: "${HL_FILE_PREFIX}src/foo.ts${HL_FILE_HASH_SEP}${HL_FILE_HASH_EXAMPLES[0]}${HL_FILE_SUFFIX}" then edit ops.`,
 		);
 	}
 
@@ -187,7 +207,7 @@ function splitRawSections(input: string, options: SplitOptions = {}): RawSection
 		if (token.kind === "envelope-end" || token.kind === "abort") break;
 		if (token.kind === "envelope-begin") continue;
 
-		// Route every `¶`-prefixed line through parseHashlineHeaderLine so
+		// Route every bracket-prefixed line through parseHashlineHeaderLine so
 		// malformed headers still raise the strict "Input header must be …"
 		// diagnostic (the tokenizer alone would silently classify them as
 		// payload).
@@ -323,7 +343,7 @@ export class PatchSection {
 
 /**
  * A parsed hashline patch — zero or more {@link PatchSection}s, each rooted
- * at a `¶PATH#HASH` header. Construct via {@link Patch.parse}.
+ * at a `[PATH#HASH]` header. Construct via {@link Patch.parse}.
  *
  * `Patch` is pure data: parsing is line-anchored and does not look at the
  * filesystem. To apply a patch, hand it to {@link Patcher.apply}.
diff --git a/packages/hashline/src/messages.ts b/packages/hashline/src/messages.ts
index 5eff387b3..e5e33640d 100644
--- a/packages/hashline/src/messages.ts
+++ b/packages/hashline/src/messages.ts
@@ -5,7 +5,7 @@
  * them.
  */
 
-import { HL_FILE_HASH_SEP, HL_FILE_PREFIX } from "./format";
+import { HL_FILE_HASH_SEP, HL_FILE_PREFIX, HL_FILE_SUFFIX } from "./format";
 
 /** Lines of context shown either side of a hash mismatch. */
 export const MISMATCH_CONTEXT = 2;
@@ -124,5 +124,5 @@ export const HEADTAIL_DRIFT_WARNING =
  * this single builder to stay in lockstep.
  */
 export function missingSnapshotTagMessage(sectionPath: string): string {
-	return `Missing hashline snapshot tag for edit to ${sectionPath}; use \`${HL_FILE_PREFIX}${sectionPath}${HL_FILE_HASH_SEP}tag\` from your latest read/search output. To create a new file, use the write tool.`;
+	return `Missing hashline snapshot tag for edit to ${sectionPath}; use \`${HL_FILE_PREFIX}${sectionPath}${HL_FILE_HASH_SEP}tag${HL_FILE_SUFFIX}\` from your latest read/search output. To create a new file, use the write tool.`;
 }
diff --git a/packages/hashline/src/mismatch.ts b/packages/hashline/src/mismatch.ts
index b77d02454..5e9bed476 100644
--- a/packages/hashline/src/mismatch.ts
+++ b/packages/hashline/src/mismatch.ts
@@ -6,7 +6,7 @@
  * plus a couple of lines of surrounding context. The {@link MismatchError}
  * formats this into a message at construction time.
  */
-import { formatNumberedLine, HL_FILE_HASH_EXAMPLES, HL_FILE_HASH_SEP, HL_FILE_PREFIX } from "./format";
+import { formatNumberedLine, HL_FILE_HASH_EXAMPLES, HL_FILE_HASH_SEP, HL_FILE_PREFIX, HL_FILE_SUFFIX } from "./format";
 import { MISMATCH_CONTEXT } from "./messages";
 
 const LINE_REF_RE = /^\s*[>+\-*]*\s*(\d+)(?::.*)?\s*$/;
@@ -15,7 +15,7 @@ export function formatFullAnchorRequirement(raw?: string): string {
 	const received = raw === undefined ? "" : ` Received ${JSON.stringify(raw)}.`;
 	return (
 		`a bare line number from read/search output plus the section header content-hash tag ` +
-		`(for example ${HL_FILE_PREFIX}src/foo.ts${HL_FILE_HASH_SEP}${HL_FILE_HASH_EXAMPLES[0]} and line "160")${received}`
+		`(for example ${HL_FILE_PREFIX}src/foo.ts${HL_FILE_HASH_SEP}${HL_FILE_HASH_EXAMPLES[0]}${HL_FILE_SUFFIX} and line "160")${received}`
 	);
 }
 
@@ -99,12 +99,12 @@ export class MismatchError extends Error {
 		if (!hashRecognized) {
 			return [
 				`Edit rejected${pathText}: hash ${HL_FILE_HASH_SEP}${details.expectedFileHash} is not from this session.`,
-				`The current file hashes to ${HL_FILE_HASH_SEP}${details.actualFileHash}. Re-read the file with \`read\` to copy a current ${HL_FILE_PREFIX}path${HL_FILE_HASH_SEP}tag header — never invent the tag and never reuse one from a prior session.`,
+				`The current file hashes to ${HL_FILE_HASH_SEP}${details.actualFileHash}. Re-read the file with \`read\` to copy a current ${HL_FILE_PREFIX}path${HL_FILE_HASH_SEP}tag${HL_FILE_SUFFIX} header — never invent the tag and never reuse one from a prior session.`,
 			];
 		}
 		return [
 			`Edit rejected${pathText}: file changed between read and edit.`,
-			`Section is bound to ${HL_FILE_HASH_SEP}${details.expectedFileHash}, but the current file hashes to ${HL_FILE_HASH_SEP}${details.actualFileHash}. If a prior edit in this session modified this file, copy the ${HL_FILE_PREFIX}path${HL_FILE_HASH_SEP}newhash header from that edit's response; otherwise re-read the file with \`read\` to refresh the tag before retrying.`,
+			`Section is bound to ${HL_FILE_HASH_SEP}${details.expectedFileHash}, but the current file hashes to ${HL_FILE_HASH_SEP}${details.actualFileHash}. If a prior edit in this session modified this file, copy the ${HL_FILE_PREFIX}path${HL_FILE_HASH_SEP}newhash${HL_FILE_SUFFIX} header from that edit's response; otherwise re-read the file with \`read\` to refresh the tag before retrying.`,
 		];
 	}
 
diff --git a/packages/hashline/src/parser.ts b/packages/hashline/src/parser.ts
index 918aa6d4d..47e56b699 100644
--- a/packages/hashline/src/parser.ts
+++ b/packages/hashline/src/parser.ts
@@ -43,7 +43,7 @@ function detectApplyPatchContamination(text: string, _hasPending: boolean): stri
 		const preview = trimmed.length > 48 ? `${trimmed.slice(0, 48)}…` : trimmed;
 		return (
 			`apply_patch sentinel ${JSON.stringify(preview)} is not valid in hashline. ` +
-			"File sections start with `¶path#HASH` (no `Update File:` / `Add File:` keyword). " +
+			"File sections start with `[path#HASH]` (no `Update File:` / `Add File:` keyword). " +
 			"Use `replace N..M:`, `delete N..M`, or `insert before|after|head|tail:` ops."
 		);
 	}
diff --git a/packages/hashline/src/patcher.ts b/packages/hashline/src/patcher.ts
index 7204867a6..de257fa40 100644
--- a/packages/hashline/src/patcher.ts
+++ b/packages/hashline/src/patcher.ts
@@ -64,9 +64,9 @@ export interface PatchSectionResult {
 	persisted: string;
 	/** Final text that the {@link Filesystem} actually wrote (may differ if the FS transformed it). */
 	written: string;
-	/** 3-hex opaque snapshot tag for `after`. Use to anchor follow-up edits. */
+	/** 4-hex content-hash tag for `after`. Use to anchor follow-up edits. */
 	fileHash: string;
-	/** Hashline section header (`¶path#tag`) of the post-edit content. */
+	/** Hashline section header (`[path#tag]`) of the post-edit content. */
 	header: string;
 	/** 1-indexed first changed line in `after`, or `undefined` for noops. */
 	firstChangedLine?: number;
diff --git a/packages/hashline/src/prefixes.ts b/packages/hashline/src/prefixes.ts
index 56515d55a..6e18950da 100644
--- a/packages/hashline/src/prefixes.ts
+++ b/packages/hashline/src/prefixes.ts
@@ -14,9 +14,11 @@
  * otherwise turn every content line into a (malformed) op.
  */
 
+import { HL_FILE_HASH_LENGTH } from "./format";
+
 const HL_PREFIX_RE = /^\s*(?:>>>|>>)?\s*(?:[+*-]\s*)?\d+:/;
 const HL_PREFIX_PLUS_RE = /^\s*(?:>>>|>>)?\s*\+\s*\d+:/;
-const HL_HEADER_RE = /^\s*¶\S+#[0-9a-fA-F]{3}\s*$/;
+const HL_HEADER_RE = new RegExp(`^\\s*\\[[^#\\r\\n]+#[0-9a-fA-F]{${HL_FILE_HASH_LENGTH}}\\]\\s*$`);
 const DIFF_PLUS_RE = /^[+](?![+])/;
 const READ_TRUNCATION_NOTICE_RE = /^\[(?:Showing lines \d+-\d+ of \d+|\d+ more lines? in (?:file|\S+))\b.*\bUse :L?\d+/;
 
diff --git a/packages/hashline/src/prompt.md b/packages/hashline/src/prompt.md
index 623d13d89..6547ff2e6 100644
--- a/packages/hashline/src/prompt.md
+++ b/packages/hashline/src/prompt.md
@@ -1,11 +1,11 @@
 Your patch language names lines to replace, delete, or insert at, then lists the new content. Rule of thumb: a header ending in `:` is followed by `+` body rows; `delete` has no body.
 
 <headers>
-Every file section starts with `¶PATH#TAG`. `TAG` is the 4-hex snapshot tag from your latest `read`/`search`, and is REQUIRED on every section — there is no hashless form. To create a new file, use the `write` tool; hashline only edits files that already exist.
+Every file section starts with `[PATH#TAG]`. `TAG` is the 4-hex snapshot tag from your latest `read`/`search`, and is REQUIRED on every section — there is no hashless form. To create a new file, use the `write` tool; hashline only edits files that already exist.
 </headers>
 
 <ops>
-replace N..M:      replace original lines N..M with the body rows below.
+replace N..M:      replace original lines N..M with the body rows below. CAUTION, IT IS INCLUSIVE! MAKE SURE YOU INTEND TO DELETE BOTH ENDS!
 replace block N:   replace the whole syntactic block that BEGINS on line N — its header line through its closing line — resolved with tree-sitter. Body rows below. Point N at the line that OPENS the construct (the `if`/`function`/`def`/`{`-bearing line), not a closing `}` or a blank line.
 delete N..M        delete original lines N..M. No body.
 delete block N     delete the whole syntactic block that BEGINS on line N.
@@ -23,21 +23,24 @@ There is NO other body row kind. NEVER write `-old` or a bare/context line. To k
 </body-rows>
 
 <rules>
-- Line numbers come from `read`/`search` (`LINE:TEXT`). Copy the `¶PATH#TAG` header; use the bare LINE numbers.
+- Line numbers come from `read`/`search` (`LINE:TEXT`). Copy the `[PATH#TAG]` header; use the bare LINE numbers.
 - Numbers refer to the ORIGINAL file and stay valid for the whole patch — they do not shift as hunks apply.
-- Across calls they do NOT survive: each applied edit mints a fresh `#TAG` and renumbers the file, so the tag and line numbers you just used are dead. Anchor the next edit on the `¶PATH#TAG` and lines from the edit response (or re-`read`), never on pre-edit numbers.
+- Across calls they do NOT survive: each applied edit mints a fresh `#TAG` and renumbers the file, so the tag and line numbers you just used are dead. Anchor the next edit on the `[PATH#TAG]` and lines from the edit response (or re-`read`), never on pre-edit numbers.
 - A line number is an offset, not a structural boundary: never `insert after N` into a construct you have not read, and never start or end a `replace`/`delete` range mid-expression or mid-block. If unsure what is on those lines, `read` them first.
+- A valid `#TAG` is NOT permission to patch the whole file — it certifies the snapshot, not your knowledge of it. Authority to touch a line comes from having literally seen that line as a `LINE:TEXT` row in a `read`/`search`, not from holding the tag. Every line in a hunk's range, and the lines bounding it, must be lines you actually saw.
+- An elided or partial read is NOT a read of the gap. A `…` (or any collapsed/truncated region) between two excerpts means those lines are UNSEEN — treat them exactly like lines you never opened. Never place a hunk on, or span a range across, an elided region; `read` that range explicitly first. Reconstructing it from memory of "what the code probably looks like" is how ranges drift off-by-N and shred neighboring blocks.
 - On a stale-tag rejection — or any result you cannot fully account for — STOP and re-`read`. Never stack more line-numbered edits onto output you have not re-grounded; that compounds corruption.
 - One hunk per range; the body is the final content, never an old/new pair.
 - Keep every range as tight as the change: a range must cover ONLY lines whose content actually changes. Never widen it to swallow an unchanged signature, brace, or neighboring statement just to rewrite a few lines inside — change one line with `replace N..N`, not the whole block around it. (A range where every line genuinely changes is correctly long; tightness is about excluding unchanged lines, not about being short.) This bounds the blast radius if a number is off: a stale single-line replace corrupts one line, while a stale block replace shreds the whole block and its structure.
 - To change lines 2 and 5 while keeping 3–4, issue two hunks (`replace 2..2:` and `replace 5..5:`). Untouched lines are simply absent from every range.
+- Pure additions use `insert`, never a widened `replace`. If the change only adds lines, `insert before/after` the spot and keep every existing line out of all ranges. Do NOT `replace` a span of keepers and retype them around the new line "to preserve" them — those retyped keepers are exactly what gets silently dropped when one is forgotten. A keeper that never enters your body cannot be lost. `replace` is only for lines whose own text changes.
 - NEVER use this tool to format code — reordering imports, re-indenting, aligning columns, or any mechanical restyling. That is the project formatter's job; run it instead of hand-editing layout here.
 </rules>
 
 <example>
 Original (the exact shape `read` returns):
 ```
-¶greet.py#A1B2
+[greet.py#A1B2]
 1:def greet(name):
 2:    msg = "Hello, " + name
 3:    print(msg)
@@ -46,14 +49,14 @@ Original (the exact shape `read` returns):
 
 Insert a guard after line 1:
 ```
-¶greet.py#A1B2
+[greet.py#A1B2]
 insert after 1:
 +    if not name: name = "stranger"
 ```
 
 Replace line 2 with two lines:
 ```
-¶greet.py#A1B2
+[greet.py#A1B2]
 replace 2..2:
 +    greeting = "Hi"
 +    msg = f"{greeting}, {name}"
@@ -61,13 +64,13 @@ replace 2..2:
 
 Delete line 3:
 ```
-¶greet.py#A1B2
+[greet.py#A1B2]
 delete 3
 ```
 
 Add a header and trailer:
 ```
-¶greet.py#A1B2
+[greet.py#A1B2]
 insert head:
 +# generated header
 insert tail:
@@ -76,7 +79,7 @@ insert tail:
 
 Replace the whole `greet` function block — `replace block 1:` resolves lines 1–3 (the `def` header through `print(msg)`); line 4 is a separate statement and stays:
 ```
-¶greet.py#A1B2
+[greet.py#A1B2]
 replace block 1:
 +def greet(name):
 +    print(f"Hello, {name}")
@@ -99,6 +102,16 @@ replace 3..3:
 # RIGHT
 replace 3..3:
 +   return msg
+
+# WRONG — a pure insertion done as a widened `replace`: you only want to add one line after 2,
+# but you replace 2..4, retype the keepers in the body, and drop one (here line 4, `greet("world")`).
+replace 2..4:
++    msg = "Hello, " + name
++    extra = compute(name)
++    print(msg)
+# RIGHT — touch nothing you keep; the new line is the whole body.
+insert after 2:
++    extra = compute(name)
 </anti-patterns>
 
 <critical>
diff --git a/packages/hashline/src/tokenizer.ts b/packages/hashline/src/tokenizer.ts
index 19ffe7164..491fd7dc3 100644
--- a/packages/hashline/src/tokenizer.ts
+++ b/packages/hashline/src/tokenizer.ts
@@ -3,7 +3,7 @@
  *
  * Format shape:
  * ```
- * ¶path/to/file.ts#0A3
+ * [path/to/file.ts#1A2B]
  * replace 5..7:
  * +literal new line
  * ```
@@ -15,6 +15,7 @@ import {
 	HL_FILE_HASH_LENGTH,
 	HL_FILE_HASH_SEP,
 	HL_FILE_PREFIX,
+	HL_FILE_SUFFIX,
 	HL_HEADER_COLON,
 	HL_INSERT_AFTER,
 	HL_INSERT_BEFORE,
@@ -45,6 +46,7 @@ const CHAR_LOWER_F = 102;
 const CHAR_PAYLOAD_REPLACE = HL_PAYLOAD_REPLACE.charCodeAt(0);
 const CHAR_COLON = HL_HEADER_COLON.charCodeAt(0);
 const FILE_PREFIX_LENGTH = HL_FILE_PREFIX.length;
+const FILE_SUFFIX_LENGTH = HL_FILE_SUFFIX.length;
 
 function isDigitCode(code: number): boolean {
 	return code >= CHAR_ZERO && code <= CHAR_NINE;
@@ -137,7 +139,7 @@ export function parseLid(raw: string, lineNum: number): Anchor {
 	if (number === null || skipWhitespace(raw, number.nextIndex, end) !== end) {
 		throw new Error(
 			`line ${lineNum}: expected a line number such as ${describeAnchorExamples("119")}; ` +
-				`got ${JSON.stringify(raw)}. Use ${HL_FILE_PREFIX}PATH${HL_FILE_HASH_SEP}hash from your latest read for file-version binding.`,
+				`got ${JSON.stringify(raw)}. Use ${HL_FILE_PREFIX}PATH${HL_FILE_HASH_SEP}hash${HL_FILE_SUFFIX} from your latest read for file-version binding.`,
 		);
 	}
 	return { line: number.line };
@@ -312,28 +314,44 @@ function tryParseHunkHeader(line: string): ParsedHunkHeader | null {
 function tryParseHeader(line: string): { path: string; fileHash?: string } | null {
 	if (!line.startsWith(HL_FILE_PREFIX)) return null;
 	const end = trimEndIndex(line);
-	let index = FILE_PREFIX_LENGTH;
-	if (index >= end) return null;
-	const pathStart = index;
-	while (index < end) {
-		const code = line.charCodeAt(index);
-		if (code === CHAR_HASH || code === CHAR_SPACE || code === CHAR_TAB) break;
-		index++;
-	}
-	if (index === pathStart) return null;
-	const path = line.slice(pathStart, index);
+	if (FILE_PREFIX_LENGTH + FILE_SUFFIX_LENGTH >= end) return null;
+	if (!line.endsWith(HL_FILE_SUFFIX, end)) return null;
+	const bodyEnd = end - FILE_SUFFIX_LENGTH;
+	if (FILE_PREFIX_LENGTH >= bodyEnd) return null;
+
+	// The snapshot tag, when present, is the trailing `#XXXX` block inside the
+	// bracketed header. We detect it from the suffix so the path may
+	// legitimately contain whitespace (e.g. `OneDrive - Company/file.ts`).
+	let pathEnd = bodyEnd;
 	let fileHash: string | undefined;
-	if (index < end && line.charCodeAt(index) === CHAR_HASH) {
-		const hashStart = index + 1;
-		const hashEnd = hashStart + HL_FILE_HASH_LENGTH;
-		if (hashEnd > end) return null;
-		for (let probe = hashStart; probe < hashEnd; probe++) {
-			if (!isHexDigitCode(line.charCodeAt(probe))) return null;
+	const trailingHashStart = bodyEnd - HL_FILE_HASH_LENGTH - 1;
+	if (trailingHashStart >= FILE_PREFIX_LENGTH && line.charCodeAt(trailingHashStart) === CHAR_HASH) {
+		let allHex = true;
+		for (let probe = trailingHashStart + 1; probe < bodyEnd; probe++) {
+			if (!isHexDigitCode(line.charCodeAt(probe))) {
+				allHex = false;
+				break;
+			}
+		}
+		if (allHex) {
+			pathEnd = trailingHashStart;
+			fileHash = line.slice(trailingHashStart + 1, bodyEnd).toUpperCase();
 		}
-		fileHash = line.slice(hashStart, hashEnd).toUpperCase();
-		index = hashEnd;
 	}
-	if (skipWhitespace(line, index, end) !== end) return null;
+
+	// The hashline header grammar uses `#` as the path/tag separator and
+	// does not allow `#` inside filenames. Anything `#` left in the path
+	// body — short tags (`#1A2`), non-hex tags (`#1A2G`), over-long tags
+	// (`#1A2B5`), stale-tag copy-paste (`#1A2B copied from read`), or
+	// line-suffixed tags (`#1A2B:42`) — means the header is malformed.
+	// Surface the focused diagnostic instead of silently mis-routing the
+	// edit or reporting a missing tag downstream.
+	for (let i = FILE_PREFIX_LENGTH; i < pathEnd; i++) {
+		if (line.charCodeAt(i) === CHAR_HASH) return null;
+	}
+
+	if (pathEnd === FILE_PREFIX_LENGTH) return null;
+	const path = line.slice(FILE_PREFIX_LENGTH, pathEnd);
 	return fileHash !== undefined ? { path, fileHash } : { path };
 }
 
diff --git a/packages/hashline/src/types.ts b/packages/hashline/src/types.ts
index 55a9ca010..82326c628 100644
--- a/packages/hashline/src/types.ts
+++ b/packages/hashline/src/types.ts
@@ -72,7 +72,7 @@ export interface SplitOptions {
 	/** Resolves absolute paths inside hashline headers to cwd-relative form. */
 	cwd?: string;
 	/**
-	 * Fallback path used when the input lacks a `¶PATH` header but contains
+	 * Fallback path used when the input lacks a `[PATH]` header but contains
 	 * recognizable hashline operations. Lets streaming previews work before
 	 * the model has written the header.
 	 */
diff --git a/packages/hashline/test/block.test.ts b/packages/hashline/test/block.test.ts
index 9ded37f47..507b7f1b0 100644
--- a/packages/hashline/test/block.test.ts
+++ b/packages/hashline/test/block.test.ts
@@ -91,8 +91,8 @@ describe("PatchSection.applyTo / applyPartialTo with block edits", () => {
 	const text = "function x() {\n  if (y) {\n  }\n}\n";
 
 	it("applyTo resolves a block edit and matches the equivalent `replace`", () => {
-		const blockSection = Patch.parseSingle(`¶${PATH}#1A2B\nreplace block 2:\n+  if (y || z) {\n+  }`);
-		const replaceSection = Patch.parseSingle(`¶${PATH}#1A2B\nreplace 2..3:\n+  if (y || z) {\n+  }`);
+		const blockSection = Patch.parseSingle(`[${PATH}#1A2B]\nreplace block 2:\n+  if (y || z) {\n+  }`);
+		const replaceSection = Patch.parseSingle(`[${PATH}#1A2B]\nreplace 2..3:\n+  if (y || z) {\n+  }`);
 
 		const blockResult = blockSection.applyTo(text, stubResolver);
 		const replaceResult = replaceSection.applyTo(text);
@@ -102,12 +102,12 @@ describe("PatchSection.applyTo / applyPartialTo with block edits", () => {
 	});
 
 	it("applyTo throws when a block edit has no resolver", () => {
-		const section = Patch.parseSingle(`¶${PATH}#1A2B\nreplace block 2:\n+X`);
+		const section = Patch.parseSingle(`[${PATH}#1A2B]\nreplace block 2:\n+X`);
 		expect(() => section.applyTo(text)).toThrow("replace block");
 	});
 
 	it("applyPartialTo drops an unresolvable block edit instead of throwing", () => {
-		const section = Patch.parseSingle(`¶${PATH}#1A2B\nreplace block 2:\n+X`);
+		const section = Patch.parseSingle(`[${PATH}#1A2B]\nreplace block 2:\n+X`);
 		// No resolver → drop. The lone block edit vanishes, so the text is unchanged.
 		const result = section.applyPartialTo(text);
 		expect(result.text).toBe(text);
@@ -123,7 +123,7 @@ describe("Patcher with a block resolver", () => {
 		const tag = snapshots.record(PATH, text);
 		const patcher = new Patcher({ fs, snapshots, blockResolver: stubResolver });
 
-		const result = await patcher.apply(Patch.parse(`¶${PATH}#${tag}\nreplace block 2:\n+  if (y || z) {\n+  }`));
+		const result = await patcher.apply(Patch.parse(`[${PATH}#${tag}]\nreplace block 2:\n+  if (y || z) {\n+  }`));
 
 		expect(result.sections[0]?.op).toBe("update");
 		expect(fs.get(PATH)).toBe("function x() {\n  if (y || z) {\n  }\n}\n");
@@ -140,7 +140,7 @@ describe("Patcher with a block resolver", () => {
 
 		// `block 2` resolves against the SNAPSHOT → span [2,3] → replace
 		// "line1","line2"; recovery 3-way-merges the change onto the live file.
-		const result = await patcher.apply(Patch.parse(`¶${PATH}#${tag}\nreplace block 2:\n+NEW`));
+		const result = await patcher.apply(Patch.parse(`[${PATH}#${tag}]\nreplace block 2:\n+NEW`));
 
 		expect(result.sections[0]?.op).toBe("update");
 		expect(fs.get(PATH)).toBe("line0\nNEW\nline3\nline4\nline5\n");
@@ -155,7 +155,7 @@ describe("Patcher with a block resolver", () => {
 		const bogus = live === "FFFF" ? "0000" : "FFFF";
 		const patcher = new Patcher({ fs, snapshots, blockResolver: stubResolver });
 
-		await expect(patcher.apply(Patch.parse(`¶${PATH}#${bogus}\nreplace block 2:\n+NEW`))).rejects.toBeInstanceOf(
+		await expect(patcher.apply(Patch.parse(`[${PATH}#${bogus}]\nreplace block 2:\n+NEW`))).rejects.toBeInstanceOf(
 			MismatchError,
 		);
 		expect(fs.get(PATH)).toBe(liveText);
@@ -167,7 +167,7 @@ describe("Patcher with a block resolver", () => {
 		const tag = snapshots.record(PATH, text);
 		const patcher = new Patcher({ fs, snapshots, blockResolver: () => null });
 
-		await expect(patcher.apply(Patch.parse(`¶${PATH}#${tag}\nreplace block 2:\n+X`))).rejects.toThrow(
+		await expect(patcher.apply(Patch.parse(`[${PATH}#${tag}]\nreplace block 2:\n+X`))).rejects.toThrow(
 			"could not resolve a syntactic block",
 		);
 		expect(fs.get(PATH)).toBe(text);
@@ -201,13 +201,13 @@ describe("delete block", () => {
 	});
 
 	it("applyTo deletes the resolved block span", () => {
-		const section = Patch.parseSingle(`¶${PATH}#1A2B\ndelete block 2`);
+		const section = Patch.parseSingle(`[${PATH}#1A2B]\ndelete block 2`);
 		// stub span [2,3] → drop "  if (y) {" and "  }".
 		expect(section.applyTo(text, stubResolver).text).toBe("function x() {\n}\n");
 	});
 
 	it("applyPartialTo drops an unresolvable delete-block edit instead of throwing", () => {
-		const section = Patch.parseSingle(`¶${PATH}#1A2B\ndelete block 2`);
+		const section = Patch.parseSingle(`[${PATH}#1A2B]\ndelete block 2`);
 		expect(section.applyPartialTo(text).text).toBe(text);
 	});
 
@@ -217,7 +217,7 @@ describe("delete block", () => {
 		const tag = snapshots.record(PATH, text);
 		const patcher = new Patcher({ fs, snapshots, blockResolver: stubResolver });
 
-		const result = await patcher.apply(Patch.parse(`¶${PATH}#${tag}\ndelete block 2`));
+		const result = await patcher.apply(Patch.parse(`[${PATH}#${tag}]\ndelete block 2`));
 
 		expect(result.sections[0]?.op).toBe("update");
 		expect(fs.get(PATH)).toBe("function x() {\n}\n");
diff --git a/packages/hashline/test/leniency.test.ts b/packages/hashline/test/leniency.test.ts
index f2f2bb2aa..b28f57c27 100644
--- a/packages/hashline/test/leniency.test.ts
+++ b/packages/hashline/test/leniency.test.ts
@@ -1,5 +1,5 @@
 import { describe, expect, it } from "bun:test";
-import { applyEdits, parsePatch } from "@oh-my-pi/hashline";
+import { applyEdits, Patch, parsePatch } from "@oh-my-pi/hashline";
 
 function applyPatch(text: string, diff: string): string {
 	return applyEdits(text, parsePatch(diff).edits).text;
@@ -7,6 +7,62 @@ function applyPatch(text: string, diff: string): string {
 
 const FILE = "a\nb\nc\nd\ne";
 
+describe("hashline section headers", () => {
+	it("accepts paths with spaces in anchored section headers", () => {
+		const section = Patch.parseSingle("[dir with spaces/file.ts#1a2b]\nreplace 1..1:\n+after");
+
+		expect(section.path).toBe("dir with spaces/file.ts");
+		expect(section.fileHash).toBe("1A2B");
+		expect(section.applyTo("before").text).toBe("after");
+	});
+
+	it("recovers apply_patch-contaminated headers whose paths contain spaces", () => {
+		const section = Patch.parseSingle("[*** Update File: dir with spaces/file.ts#1A2B]\nreplace 1..1:\n+after");
+
+		expect(section.path).toBe("dir with spaces/file.ts");
+		expect(section.fileHash).toBe("1A2B");
+		expect(section.applyTo("before").text).toBe("after");
+	});
+
+	it("rejects trailing junk after a snapshot tag", () => {
+		expect(() => Patch.parse("[src/a.ts#1A2B copied from read]\nreplace 1..1:\n+after")).toThrow(
+			/Input header must be/,
+		);
+		expect(() => Patch.parse("[src/a.ts#1A2B:812]\nreplace 1..1:\n+after")).toThrow(/Input header must be/);
+	});
+
+	it("rejects trailing junk after a snapshot tag even with apply_patch noise", () => {
+		expect(() => Patch.parse("[Update File: src/a.ts#1A2B copied from read]\nreplace 1..1:\n+after")).toThrow(
+			/Input header must be/,
+		);
+		expect(() => Patch.parse("[Update File: src/a.ts#1A2B:812]\nreplace 1..1:\n+after")).toThrow(
+			/Input header must be/,
+		);
+	});
+
+	it("rejects malformed snapshot tags", () => {
+		expect(() => Patch.parse("[src/a.ts#1A2]\nreplace 1..1:\n+after")).toThrow(/Input header must be/);
+		expect(() => Patch.parse("[src/a.ts#1A2G]\nreplace 1..1:\n+after")).toThrow(/Input header must be/);
+		expect(() => Patch.parse("[src/a.ts#1A2B5]\nreplace 1..1:\n+after")).toThrow(/Input header must be/);
+	});
+
+	it("rejects malformed snapshot tags even with apply_patch noise", () => {
+		expect(() => Patch.parse("[Update File: src/a.ts#1A2G]\nreplace 1..1:\n+after")).toThrow(/Input header must be/);
+	});
+
+	it("reports bracket syntax with a 4-hex example when the header is missing", () => {
+		try {
+			Patch.parse("delete 38..40");
+			throw new Error("expected missing-header error");
+		} catch (error) {
+			const message = error instanceof Error ? error.message : String(error);
+			expect(message).toContain('input must begin with "[PATH#HASH]"');
+			expect(message).toContain('Example: "[src/foo.ts#1A2B]"');
+			expect(message).not.toContain("#0A3");
+		}
+	});
+});
+
 describe("hashline core — verb header forms", () => {
 	it("rejects a bare single-number hunk header with verb guidance", () => {
 		expect(() => parsePatch("2\n+B")).toThrow(/hunk headers need a verb/);
diff --git a/packages/hashline/test/patcher.test.ts b/packages/hashline/test/patcher.test.ts
index 1196683cc..82e891969 100644
--- a/packages/hashline/test/patcher.test.ts
+++ b/packages/hashline/test/patcher.test.ts
@@ -25,7 +25,7 @@ describe("Patcher snapshot tag integrity", () => {
 		const tag = snapshots.record(PATH, "before\n");
 		const patcher = new Patcher({ fs, snapshots });
 
-		const result = await patcher.apply(Patch.parse(`¶${PATH}#${tag}\nreplace 1..1:\n+after`));
+		const result = await patcher.apply(Patch.parse(`[${PATH}#${tag}]\nreplace 1..1:\n+after`));
 
 		expect(result.sections[0]?.op).toBe("update");
 		expect(result.sections[0]?.fileHash).toMatch(/^[0-9A-F]{4}$/);
@@ -45,14 +45,14 @@ describe("Patcher snapshot tag integrity", () => {
 		expect(snapshots.byHash(PATH, tag)).toBeNull();
 		const patcher = new Patcher({ fs, snapshots });
 
-		const result = await patcher.apply(Patch.parse(`¶${PATH}#${tag}\nreplace 3..3:\n+L3`));
+		const result = await patcher.apply(Patch.parse(`[${PATH}#${tag}]\nreplace 3..3:\n+L3`));
 
 		expect(result.sections[0]?.op).toBe("update");
 		expect(fs.get(PATH)).toBe("l1\nl2\nL3\nl4\nl5\n");
 	});
 
 	it("normalizes lowercase section tags while parsing", () => {
-		const section = Patch.parseSingle(`¶${PATH}#1a2b\nreplace 1..1:\n+after`);
+		const section = Patch.parseSingle(`[${PATH}#1a2b]\nreplace 1..1:\n+after`);
 
 		expect(section.fileHash).toBe("1A2B");
 	});
@@ -65,7 +65,7 @@ describe("Patcher snapshot tag integrity", () => {
 		const patcher = new Patcher({ fs, snapshots });
 
 		try {
-			await patcher.apply(Patch.parse(`¶${PATH}#${tag}\nreplace 1..1:\n+after`));
+			await patcher.apply(Patch.parse(`[${PATH}#${tag}]\nreplace 1..1:\n+after`));
 			throw new Error("expected MismatchError");
 		} catch (error) {
 			expect(error).toBeInstanceOf(MismatchError);
@@ -89,7 +89,7 @@ describe("Patcher snapshot tag integrity", () => {
 		const bogus = live === "FFFF" ? "0000" : "FFFF";
 
 		try {
-			await patcher.apply(Patch.parse(`¶${PATH}#${bogus}\nreplace 1..1:\n+after`));
+			await patcher.apply(Patch.parse(`[${PATH}#${bogus}]\nreplace 1..1:\n+after`));
 			throw new Error("expected MismatchError");
 		} catch (error) {
 			expect(error).toBeInstanceOf(MismatchError);
@@ -109,7 +109,7 @@ describe("Patcher mandatory snapshot tag policy", () => {
 		const snapshots = new InMemorySnapshotStore();
 		const patcher = new Patcher({ fs, snapshots });
 
-		await expect(patcher.apply(Patch.parse(`¶${PATH}\ninsert tail:\n+c`))).rejects.toThrow(
+		await expect(patcher.apply(Patch.parse(`[${PATH}]\ninsert tail:\n+c`))).rejects.toThrow(
 			/Missing hashline snapshot tag.*use the write tool/s,
 		);
 		expect(fs.get(PATH)).toBe("a\nb\n");
@@ -120,7 +120,7 @@ describe("Patcher mandatory snapshot tag policy", () => {
 		const snapshots = new InMemorySnapshotStore();
 		const patcher = new Patcher({ fs, snapshots });
 
-		await expect(patcher.apply(Patch.parse(`¶${PATH}\nreplace 1..1:\n+X`))).rejects.toThrow(
+		await expect(patcher.apply(Patch.parse(`[${PATH}]\nreplace 1..1:\n+X`))).rejects.toThrow(
 			/Missing hashline snapshot tag/,
 		);
 	});
@@ -130,7 +130,7 @@ describe("Patcher mandatory snapshot tag policy", () => {
 		const snapshots = new InMemorySnapshotStore();
 		const patcher = new Patcher({ fs, snapshots });
 
-		await expect(patcher.apply(Patch.parse(`¶ghost.ts#1A2B\ninsert tail:\n+c`))).rejects.toThrow(
+		await expect(patcher.apply(Patch.parse(`[ghost.ts#1A2B]\ninsert tail:\n+c`))).rejects.toThrow(
 			/File not found.*use the write tool/is,
 		);
 	});
@@ -143,7 +143,7 @@ describe("Patcher mandatory snapshot tag policy", () => {
 		const stale = live === "0000" ? "FFFF" : "0000";
 		const patcher = new Patcher({ fs, snapshots });
 
-		const result = await patcher.apply(Patch.parse(`¶${PATH}#${stale}\ninsert tail:\n+c`));
+		const result = await patcher.apply(Patch.parse(`[${PATH}#${stale}]\ninsert tail:\n+c`));
 
 		const section = result.sections[0];
 		expect(section?.op).toBe("update");
@@ -158,7 +158,7 @@ describe("Patcher mandatory snapshot tag policy", () => {
 		const tag = snapshots.record(PATH, content);
 		const patcher = new Patcher({ fs, snapshots });
 
-		const result = await patcher.apply(Patch.parse(`¶${PATH}#${tag}\ninsert tail:\n+c`));
+		const result = await patcher.apply(Patch.parse(`[${PATH}#${tag}]\ninsert tail:\n+c`));
 
 		const section = result.sections[0];
 		expect(section?.op).toBe("update");
diff --git a/packages/mnemopi/package.json b/packages/mnemopi/package.json
index b443f2bd2..594c2c9ca 100644
--- a/packages/mnemopi/package.json
+++ b/packages/mnemopi/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-mnemopi",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "Local SQLite memory engine for Oh My Pi agents",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
diff --git a/packages/natives/CHANGELOG.md b/packages/natives/CHANGELOG.md
index d41dacea2..c89da28eb 100644
--- a/packages/natives/CHANGELOG.md
+++ b/packages/natives/CHANGELOG.md
@@ -2,6 +2,12 @@
 
 ## [Unreleased]
 
+## [15.10.1] - 2026-06-07
+
+### Fixed
+
+- Fixed `applyBashFixups` corrupting commands that contain multi-byte UTF-8 before a trailing `| head`/`| tail` (or `2>&1`). `brush-parser` reports source positions as Unicode-scalar (char) offsets, but `pi_shell::fixup` sliced the command `&str` by those numbers as if they were byte offsets, so each multi-byte char (e.g. `✓`/`×` in a `grep -E` pattern) shifted the cut earlier and left a mangled command — e.g. `… |✓|×|XCTAssert" | tail -80` became `… |✓|×-80`, orphaning the closing quote and making the shell reject the whole pipeline with `unterminated double quote`. Positions are now translated to byte offsets before slicing.
+
 ## [15.9.0] - 2026-06-04
 
 ### Fixed
diff --git a/packages/natives/native/index.d.ts b/packages/natives/native/index.d.ts
index 0d86bd52c..8451d0d8a 100644
--- a/packages/natives/native/index.d.ts
+++ b/packages/natives/native/index.d.ts
@@ -136,7 +136,7 @@ export declare class Shell {
  * `packages/natives/native/index.js` (which derives the name from
  * `package.json#version`).
  */
-export declare function __piNativesV15_9_4(): void
+export declare function __piNativesV15_10_1(): void
 
 /**
  * Apply conservative pre-execution rewrites to a bash command.
diff --git a/packages/natives/native/index.js b/packages/natives/native/index.js
index b453c1853..e2a302227 100644
--- a/packages/natives/native/index.js
+++ b/packages/natives/native/index.js
@@ -23,7 +23,7 @@ export const PtySession = nativeBindings.PtySession;
 export const Shell = nativeBindings.Shell;
 
 // functions
-export const __piNativesV15_9_4 = nativeBindings.__piNativesV15_9_4;
+export const __piNativesV15_10_1 = nativeBindings.__piNativesV15_10_1;
 export const applyBashFixups = nativeBindings.applyBashFixups;
 export const astEdit = nativeBindings.astEdit;
 export const astGrep = nativeBindings.astGrep;
diff --git a/packages/natives/package.json b/packages/natives/package.json
index 56157bf5a..cc76acb89 100644
--- a/packages/natives/package.json
+++ b/packages/natives/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@oh-my-pi/pi-natives",
-  "version": "15.9.4",
+  "version": "15.10.1",
   "description": "Native Rust bindings for grep, clipboard, image processing, syntax highlighting, PTY, and shell operations via N-API",
   "type": "module",
   "homepage": "https://omp.sh",
diff --git a/packages/stats/package.json b/packages/stats/package.json
index 5301b8720..d44baa743 100644
--- a/packages/stats/package.json
+++ b/packages/stats/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/omp-stats",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "Local observability dashboard for pi AI usage statistics",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
diff --git a/packages/swarm-extension/package.json b/packages/swarm-extension/package.json
index 08a658ec7..ade71b483 100644
--- a/packages/swarm-extension/package.json
+++ b/packages/swarm-extension/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/swarm-extension",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "Swarm orchestration extension for omp",
 	"homepage": "https://omp.sh",
 	"author": "Derek Rynd",
diff --git a/packages/tui/CHANGELOG.md b/packages/tui/CHANGELOG.md
index e3feca332..56903eddc 100644
--- a/packages/tui/CHANGELOG.md
+++ b/packages/tui/CHANGELOG.md
@@ -2,6 +2,92 @@
 
 ## [Unreleased]
 
+## [15.10.1] - 2026-06-07
+### Breaking Changes
+
+- Removed Kitty temp-file image transmission, its startup support probe, the `PI_KITTY_IMAGE_TRANSMISSION` override, and the temp-file helper exports. Kitty/Ghostty image payloads now stay on in-band base64 before placeholder/direct placement, avoiding blank first renders from temp-file load races.
+- Renamed `RenderRequestOptions.allowUnknownViewportMutation` → `allowUnknownViewportTransientRepaint`. The option only permits a transient live-viewport repaint (autocomplete/IME/focused-editor chrome) on hosts that cannot report viewport position; it never authorizes a settled transcript commit. The old name implied any offscreen mutation was safe to push into native scrollback, which led callers to emit duplicate transcript copies.
+
+### Added
+
+- Added `TUI.addStartListener()` so feature hooks can re-enable terminal modes after temporary stop/start cycles such as external-editor handoffs.
+- Added `Editor.pasteText()` to apply terminal-style paste handling for text inserted from non-bracketed paste transports
+- Added an optional `dispose()` lifecycle method to `Component` so components can release timers and subscriptions during permanent teardown
+- Added `Container.dispose()` to propagate teardown to child components when a component tree is permanently discarded
+- Added `Loader.dispose()` to stop the loader animation timer when the component is disposed
+- Added a `ScrollView` `ellipsis` option (defaults to `Ellipsis.Unicode`) so callers that pre-wrap content to width can pass `Ellipsis.Omit` and suppress the stray per-line `…` that lands on trailing padding.
+- Added `ScrollView.handleScrollKey()` plus a `fastScrollLines` option so every scroll view gets shared navigation keys, including Shift+Arrow to scroll faster.
+- Added `OverlayOptions.fullscreen`: while the topmost visible overlay sets it, the engine borrows the terminal's alternate screen buffer for the overlay's lifetime and paints only the modal there — no ED3, no transcript re-commit — so the transcript stays untouched on the normal screen and is not scrollable behind the modal. Mouse tracking (`?1000h`/`?1006h`) is enabled for the modal's lifetime and disabled on exit, so the rest of the app keeps the terminal's native text selection.
+- Added the `submitPinsViewportToTail` terminal capability and `detectSubmitPinsViewportToTail()`: genuine local terminals where a submit keystroke scrolls the host to its tail reconcile deferred native scrollback at the prompt-submit checkpoint even when the viewport position is unprobeable (Ghostty/kitty/iTerm/WezTerm/Alacritty). Restores the pre-regression submit reconciliation without re-enabling it for Windows Terminal/ConPTY, SSH, or multiplexers, where a submit is not proof the host is at the tail.
+
+### Changed
+
+- Changed static `Loader` messages to repaint only at the spinner's 80 ms cadence; time-dependent message colorizers can opt into 16 ms redraws with `animated: true`.
+- Changed keybinding matching to precompute canonical key sets so each input sequence is parsed once per binding check instead of once per candidate key.
+- Made `Component.invalidate()` optional so leaf components without render caches no longer need no-op invalidation hooks.
+- `TERMINAL` is now a `RuntimeTerminal` whose post-construction capabilities (image protocol and the probe-driven flags) are writable, replacing the `as unknown as MutableTerminalInfo` cast pattern and the positional `withTerminalOverrides` rebuild with a prototype-preserving `clone()`.
+
+### Fixed
+
+- Fixed `Loader` text updates to skip identical messages and preserve the rendered `Text` cache instead of invalidating it every timer tick.
+
+- Fixed fullscreen overlay alt-frame rendering to reuse the current line-preparation path instead of calling removed fitting helpers.
+- Reduced TUI render-path line fitting by deferring overlay base-frame fitting until an overlay rebuild and by reusing already-fitted lines in emitters.
+- Reduced live-region pinned repaint output by diffing unchanged viewport rows when no sealed rows are being committed to native scrollback.
+- Fixed no-append live-region pinned repaints to re-anchor the hardware cursor when the logical viewport shifts.
+- Fixed keybinding matching so printable uppercase input preserves `Shift` for bindings such as `shift+a`.
+- Optimized terminal image-line detection and Thai/Lao AM normalization checks to avoid hot-path regex scans and substring allocations.
+- Fixed `Markdown.render()` cache hits returning the cache's mutable backing array, which let callers that append extra rows corrupt cached Markdown and duplicate those rows on every redraw.
+- Fixed first-paint full replays for callers that intentionally replace terminal history by allowing `TUI.start({ clearScrollback: true })`, so they do not briefly append an entire initial frame before the first clean replay.
+- Fixed ED3-risk streaming cap accounting to preserve the native scrollback high-water mark for rows that were already physically committed before transient frames were viewport-capped.
+- Fixed terminal stop and restore cleanup to disable enhanced paste mode so it does not remain enabled after shutdown
+- Removed the per-frame line-fit `Map` cache from the render timer path to avoid forcing JSC rope-string hashing during scheduled viewport repaints.
+- Fixed `visibleWidth()` so terminal column measurements for ANSI and OSC text now match the native truncation/wrapping helpers, including OSC 66 text-sizing spans being counted at their scaled payload width
+- Fixed cursor, padding, and line-fit behavior when strings contain tabs or OSC escapes by aligning `visibleWidth()` with the native text-width model
+- Fixed the transcript — or a re-appearing prior view such as the welcome screen — duplicating itself on terminals without a scroll-position oracle (Ghostty/kitty/iTerm/WezTerm) when a foreground tool completes by rewriting a partly-committed block, or when the transcript is reset. A non-destructive viewport repaint no longer re-paints rows that are byte-identical to what is already committed to native scrollback into the active grid; the repaint anchor is clamped to the committed-and-unchanged prefix (`min(firstChanged, scrollbackHighWater)`).
+
+## [15.10.0] - 2026-06-06
+
+### Changed
+
+- Reworked the DEC 2026 synchronized-output default policy: a positive DECRQM mode-2026 report now **enables** sync (previously a report could only disable it), so conservatively defaulted-off hosts that actually support it — current Zellij, tmux master, foot, contour, mintty — are upgraded at runtime. The static allowlist also covers Alacritty and the VS Code terminal, honors a `TERM_FEATURES` `Sy` advertisement and `WT_SESSION` (Windows Terminal / WSL), and no longer blanket-disables SSH (DEC 2026 passes through to the outer terminal). Risky multiplexers still start off and rely on the probe. Added `synchronizedOutputUserOverride()` as the shared opt-out/force resolver.
+
+### Fixed
+
+- Fixed WSL/Windows Terminal row flicker while typing by repainting changed text rows before clearing only their stale suffix ([#2011](https://github.com/can1357/oh-my-pi/issues/2011)).
+- Fixed terminals that support DEC 2026 still tearing/flickering because the renderer ignored a positive DECRQM capability report and kept synchronized output off — most visibly WSL + Windows Terminal, Alacritty (≥0.13), and the VS Code terminal (≥1.108), which were detected yet refused sync.
+
+## [15.9.69] - 2026-06-06
+
+### Added
+
+- Added `TUI.resetDisplay()` to force an immediate full-frame replay, including native scrollback when the host can safely clear it.
+- Added `setPaddingY` to `Box` so vertical padding can be updated programmatically after creation.
+
+### Fixed
+
+- Fixed DECCARA background-fill optimization running when synchronized output is disabled, which could expose default-background gaps during rapidly updating tool-use panels ([#2000](https://github.com/can1357/oh-my-pi/issues/2000)).
+
+## [15.9.67] - 2026-06-06
+### Added
+
+- Added `setPaddingX` to `Box` so horizontal padding can be updated programmatically after creation
+- Added `ScrollView`, a fixed-height viewport component for pre-rendered lines with optional right-edge scrollbars and imperative scroll/page controls.
+- Added optional `Terminal.hasEagerEraseScrollbackRisk()` so custom/test terminal implementations can override the global ED3-risk profile without mutating the shared `TERMINAL` object.
+
+### Changed
+
+- Changed `SelectList` to render its visible window through `ScrollView`, replacing the `(N/M)` text scroll indicator with a uniform right-edge scrollbar (the type-to-search hint line is preserved).
+
+### Fixed
+
+- Fixed unknown-viewport deferred renders freezing bottom-anchored live chrome; deferred history mutations can now repaint only the active-grid bottom row with relative cursor movement, so spinner/status tails keep advancing without rewriting rows a scrolled reader can still see.
+- Fixed autocomplete popups freezing live repaint on ED3-risk macOS/POSIX terminals with unknown native viewport position; direct autocomplete shrink frames now repaint the live viewport without zero-byte deferral and preserve the old bottom anchor when padding can clear stale popup rows without duplicating committed scrollback.
+- Fixed focused Up/Down navigation on ED3-risk macOS/POSIX terminals replaying the whole transcript after dirty foreground-stream renders; selector/editor frames now repaint non-destructively instead of emitting `CSI 3 J` on every arrow-key move ([#1962](https://github.com/can1357/oh-my-pi/issues/1962)).
+- Fixed tmux (and screen/zellij) pane scrollback losing the head of a long streamed assistant reply once it grew past the visible pane, and stranding the chrome/footer in pane history after a later collapse — producing the "repeating chunks and missing sections" reporters saw when scrolling back through tmux pane history ([#1974](https://github.com/can1357/oh-my-pi/issues/1974)). The renderer's foreground-streaming cap-to-viewport branch (introduced in 15.9.2 for ED3-risk hosts that can checkpoint-rebuild later) also activated inside multiplexers, where checkpoint reconcile is a no-op (`refreshNativeScrollbackIfDirty` short-circuits because `\x1b[3J` cannot erase pane history). Every streaming frame clipped `lines` to the visible tail and reset `#scrollbackHighWater` to 0, so any row that scrolled above the viewport top was committed nowhere — pane history stayed empty until streaming ended. Meanwhile `#planLiveRegionPinnedRender` was explicitly disabled for multiplexers, but its `#emitLiveRegionPinnedRepaint` is built from the exact primitives tmux accepts (relative cursor moves, per-line `\x1b[2K`, `\r\n` to scroll the sealed prefix past the viewport bottom) and never emits `\x1b[2J`/`\x1b[3J`. The pinned planner now runs in multiplexers too, the cap branch skips them, and the diff/append path commits incrementally into pane history; the actively-mutating live tail stays in the visible viewport only.
+
+## [15.9.5] - 2026-06-05
+
 ### Changed
 
 - Changed terminal resize handling so any width or height change always performs a clean reset + redraw: the renderer now unconditionally clears the viewport and native scrollback (`CSI 2 J` / `CSI 3 J`) and replays the full transcript at the new geometry, replacing the previous matrix of conditional viewport-repaint / history-rebuild / deferred-mutation branches. Multiplexer panes still repaint the visible window in place (pane scrollback cannot be erased), but a resize during active ED3-risk foreground streaming now performs the same clean rebuild rather than downgrading to a non-destructive viewport repaint: the terminal already re-wrapped its saved lines at the old width, so the rebuild must erase them (ED 3) instead of leaving the mis-wrapped history on screen. As a deliberate tradeoff this drops the prior no-overflow and confirmed-scrolled guards on resize: a reader scrolled into history snaps back to the bottom and preexisting shell scrollback above the UI is cleared.
@@ -11,6 +97,7 @@
 - Fixed ED3-risk foreground streaming dropping the scrolled-off head of an append-only live block that alone overflows the viewport (a long streamed assistant reply). The live-region pin again committed native scrollback only up to the live-region start, so once the live block grew past the viewport its earlier rows scrolled above the viewport top but were committed nowhere and repainted nowhere — they vanished, leaving the reply looking like a ~viewport-tall circular buffer. The `NativeScrollbackLiveRegion` seam now also reports an optional append-only `getNativeScrollbackCommitSafeEnd`, and the pinned commit boundary is the deeper of the sealed start and that append-only end: rows in `[liveRegionStart, commitSafeEnd)` above the viewport top commit to scrollback, while volatile live blocks (tool previews that collapse) omit the boundary and keep their mutable rows deferred — preserving the pending-box-above-running-box fix.
 
 ## [15.9.4] - 2026-06-05
+
 ### Added
 
 - Added `PI_TUI_SYNC_OUTPUT=0` and `PI_TUI_SYNC_OUTPUT=1` to explicitly disable or force-enable DEC 2026 synchronized-output mode, alongside `PI_FORCE_SYNC_OUTPUT=1` as a force-on alias
diff --git a/packages/tui/bench/width.ts b/packages/tui/bench/width.ts
deleted file mode 100644
index 9f8820b38..000000000
--- a/packages/tui/bench/width.ts
+++ /dev/null
@@ -1,182 +0,0 @@
-/**
- * Benchmark: native visibleWidth vs Bun.stringWidth vs hybrid implementation
- *
- * Run: bun packages/tui/bench/visible-width.ts
- */
-import { visibleWidth as nativeVisibleWidth } from "@oh-my-pi/pi-natives";
-import { getDefaultTabWidth } from "@oh-my-pi/pi-utils";
-import { visibleWidthRaw as hybridVisibleWidth, replaceTabs } from "../src/utils";
-
-const ITERATIONS = 10_000;
-const WARMUP = 500;
-
-// Test cases covering different scenarios
-const samples = {
-	// Pure ASCII - different lengths
-	ascii_short: "hello",
-	ascii_medium: "hello world this is a plain ASCII string with some words",
-	ascii_long: "a".repeat(500),
-
-	// ANSI escape codes
-	ansi_simple: "\x1b[31mred\x1b[0m",
-	ansi_complex: "\x1b[31mred text\x1b[0m and \x1b[4munderlined content\x1b[24m with more \x1b[1;33;44mstyles\x1b[0m",
-	ansi_nested: "\x1b[1m\x1b[31m\x1b[4mbold red underline\x1b[0m normal \x1b[32mgreen\x1b[0m",
-
-	// OSC 8 hyperlinks
-	links: "prefix \x1b]8;;https://example.com\x07link text\x1b]8;;\x07 suffix",
-	links_multiple:
-		"Click \x1b]8;;https://a.com\x07here\x1b]8;;\x07 or \x1b]8;;https://b.com\x07there\x1b]8;;\x07 for info",
-
-	// Wide characters (CJK)
-	cjk_short: "日本語",
-	cjk_medium: "日本語のテキストとemoji",
-	cjk_long: "日本語のテキストと中文字符和한국어문자混合在一起形成很长的字符串",
-
-	// Emoji
-	emoji_simple: "👋🌍",
-	emoji_complex: "Hello 👨‍👩‍👧‍👦 family! 🚀✨🎉 Let's go! 🇺🇸🏳️‍🌈",
-	emoji_zwj: "👨‍💻👩‍🔬👨‍👩‍👧‍👦", // ZWJ sequences
-
-	// Mixed content
-	mixed_short: "Hello 世界 🌍",
-	mixed_medium: "\x1b[32mStatus:\x1b[0m 成功 ✓ (took 42ms)",
-	mixed_long:
-		"\x1b[1;34m[INFO]\x1b[0m Processing 日本語テキスト with emoji 🚀 and \x1b]8;;https://example.com\x07links\x1b]8;;\x07 完了",
-
-	// Edge cases
-	tabs: "col1\tcol2\tcol3\tcol4",
-	empty: "",
-	newlines: "line1\nline2\nline3",
-	control_chars: "text\x00with\x01control\x02chars",
-};
-
-// Bun.stringWidth with ANSI stripping (what hybrid uses for short strings)
-function bunStringWidth(str: string): number {
-	if (!str) return 0;
-	return Bun.stringWidth(replaceTabs(str));
-}
-
-interface BenchResult {
-	name: string;
-	totalMs: number;
-	perOpUs: number;
-}
-
-function bench(name: string, fn: () => void): BenchResult {
-	// Warmup
-	for (let i = 0; i < WARMUP; i++) fn();
-
-	const start = Bun.nanoseconds();
-	for (let i = 0; i < ITERATIONS; i++) {
-		fn();
-	}
-	const totalMs = (Bun.nanoseconds() - start) / 1e6;
-	const perOpUs = (totalMs / ITERATIONS) * 1000;
-
-	return { name, totalMs, perOpUs };
-}
-
-function formatResult(r: BenchResult, baseline?: BenchResult): string {
-	const perOp = r.perOpUs.toFixed(3);
-	if (baseline && baseline !== r) {
-		const ratio = r.perOpUs / baseline.perOpUs;
-		const indicator = ratio < 1 ? "faster" : "slower";
-		return `${r.name.padEnd(20)} ${r.totalMs.toFixed(2).padStart(8)}ms  ${perOp.padStart(8)}µs/op  ${ratio.toFixed(2)}x ${indicator}`;
-	}
-	return `${r.name.padEnd(20)} ${r.totalMs.toFixed(2).padStart(8)}ms  ${perOp.padStart(8)}µs/op  (baseline)`;
-}
-
-console.log(`\n${"=".repeat(80)}`);
-console.log(`visibleWidth benchmark: ${ITERATIONS.toLocaleString()} iterations, ${WARMUP} warmup`);
-console.log(`${"=".repeat(80)}\n`);
-
-for (const [sampleName, sample] of Object.entries(samples)) {
-	console.log(`\n--- ${sampleName} (len=${sample.length}) ---`);
-	if (sample.length > 0 && sample.length < 80) {
-		// Show sample for short strings (escape non-printable)
-		const display = sample.replace(/\x1b/g, "\\e").replace(/\x07/g, "\\a");
-		console.log(`    "${display}"`);
-	}
-
-	const results: BenchResult[] = [];
-
-	const tabW = getDefaultTabWidth();
-	results.push(
-		bench("native", () => {
-			nativeVisibleWidth(sample, tabW);
-		}),
-	);
-
-	results.push(
-		bench("bun+strip", () => {
-			bunStringWidth(sample);
-		}),
-	);
-
-	results.push(
-		bench("hybrid", () => {
-			hybridVisibleWidth(sample);
-		}),
-	);
-
-	// Find fastest as baseline
-	const baseline = results.reduce((a, b) => (a.perOpUs < b.perOpUs ? a : b));
-
-	console.log();
-	for (const r of results) {
-		console.log(`  ${formatResult(r, baseline)}`);
-	}
-
-	// Verify correctness
-	const nativeResult = nativeVisibleWidth(sample, tabW);
-	const bunResult = bunStringWidth(sample);
-	const hybridResult = hybridVisibleWidth(sample);
-
-	if (nativeResult !== hybridResult) {
-		console.log(`  ⚠️  MISMATCH: native=${nativeResult}, hybrid=${hybridResult}`);
-	}
-	if (bunResult !== hybridResult && !sample.includes("\x00")) {
-		// Control chars can differ
-		console.log(`  ⚠️  MISMATCH: bun=${bunResult}, hybrid=${hybridResult}`);
-	}
-}
-
-console.log(`\n${"=".repeat(80)}`);
-console.log("Summary");
-console.log(`${"=".repeat(80)}\n`);
-
-// Aggregate by category
-const categories = {
-	ascii: ["ascii_short", "ascii_medium", "ascii_long"],
-	ansi: ["ansi_simple", "ansi_complex", "ansi_nested"],
-	links: ["links", "links_multiple"],
-	cjk: ["cjk_short", "cjk_medium", "cjk_long"],
-	emoji: ["emoji_simple", "emoji_complex", "emoji_zwj"],
-	mixed: ["mixed_short", "mixed_medium", "mixed_long"],
-};
-
-const benchTabW = getDefaultTabWidth();
-
-for (const [category, sampleNames] of Object.entries(categories)) {
-	const categoryResults = { native: 0, bun: 0, hybrid: 0 };
-
-	for (const name of sampleNames) {
-		const sample = samples[name as keyof typeof samples];
-
-		const nativeTime = bench("", () => nativeVisibleWidth(sample, benchTabW)).perOpUs;
-		const bunTime = bench("", () => bunStringWidth(sample)).perOpUs;
-		const hybridTime = bench("", () => hybridVisibleWidth(sample)).perOpUs;
-
-		categoryResults.native += nativeTime;
-		categoryResults.bun += bunTime;
-		categoryResults.hybrid += hybridTime;
-	}
-
-	const fastest = Math.min(categoryResults.native, categoryResults.bun, categoryResults.hybrid);
-	const winner =
-		fastest === categoryResults.native ? "native" : fastest === categoryResults.bun ? "bun+strip" : "hybrid";
-
-	console.log(
-		`${category.padEnd(10)} native: ${categoryResults.native.toFixed(1)}µs  bun: ${categoryResults.bun.toFixed(1)}µs  hybrid: ${categoryResults.hybrid.toFixed(1)}µs  → ${winner} wins`,
-	);
-}
diff --git a/packages/tui/package.json b/packages/tui/package.json
index 6360e2dbe..76fa4fa1b 100644
--- a/packages/tui/package.json
+++ b/packages/tui/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-tui",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "Terminal User Interface library with differential rendering for efficient text-based applications",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
diff --git a/packages/tui/src/components/box.ts b/packages/tui/src/components/box.ts
index 9562a0d25..ec3ae61e6 100644
--- a/packages/tui/src/components/box.ts
+++ b/packages/tui/src/components/box.ts
@@ -42,6 +42,18 @@ export class Box implements Component {
 		this.#invalidateCache();
 	}
 
+	setPaddingX(paddingX: number): void {
+		if (this.#paddingX === paddingX) return;
+		this.#paddingX = paddingX;
+		this.#invalidateCache();
+	}
+
+	setPaddingY(paddingY: number): void {
+		if (this.#paddingY === paddingY) return;
+		this.#paddingY = paddingY;
+		this.#invalidateCache();
+	}
+
 	setBgFn(bgFn?: (text: string) => string): void {
 		this.#bgFn = bgFn;
 		// Don't invalidate here - we'll detect bgFn changes by sampling output
diff --git a/packages/tui/src/components/editor.ts b/packages/tui/src/components/editor.ts
index 326ade5f7..2ea05e43e 100644
--- a/packages/tui/src/components/editor.ts
+++ b/packages/tui/src/components/editor.ts
@@ -1495,6 +1495,11 @@ export class Editor implements Component, Focusable {
 		this.#insertTextAtCursor(text);
 	}
 
+	/** Apply terminal paste semantics to text from non-bracketed paste transports. */
+	pasteText(text: string): void {
+		this.#handlePaste(text);
+	}
+
 	// All the editor methods from before...
 	#insertCharacter(char: string): void {
 		this.#exitHistoryForEditing();
diff --git a/packages/tui/src/components/loader.ts b/packages/tui/src/components/loader.ts
index b9b2ffca9..1ae5cf4ed 100644
--- a/packages/tui/src/components/loader.ts
+++ b/packages/tui/src/components/loader.ts
@@ -3,21 +3,20 @@ import { sliceByColumn, visibleWidth } from "../utils";
 import { Text } from "./text";
 
 /**
- * Loader component that drives display refresh at ~60fps so callers whose
- * message colorizer is time-dependent (e.g. shimmer/KITT) animate smoothly.
+ * Loader component. Spinner frames advance at `SPINNER_ADVANCE_MS`.
  *
- * Two cadences are interleaved on a single timer:
- *   - **Render tick** (every `RENDER_INTERVAL_MS`) → asks the TUI to redraw.
- *     The TUI already throttles at 16ms (`MIN_RENDER_INTERVAL_MS`), so this
- *     is the natural upper bound; static messageColorFns produce identical
- *     output and the differ drops the no-op redraw at ~zero cost.
- *   - **Spinner advance** (every `SPINNER_ADVANCE_MS`) → bumps the spinner
- *     frame index. Decoupled from the render cadence so the spinner keeps
- *     its classic ~12.5fps step pace regardless of shimmer state.
+ * Message colorizers that are time-dependent can opt into 30fps redraws by
+ * setting `animated` to `true` on the function object.
  */
-const RENDER_INTERVAL_MS = 16;
+const RENDER_INTERVAL_MS = 1000 / 30;
 const SPINNER_ADVANCE_MS = 80;
 
+type ColorFn = (str: string) => string;
+
+export type LoaderMessageColorFn = ColorFn & {
+	readonly animated?: true;
+};
+
 export class Loader extends Text {
 	#frames = ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧", "⠇", "⠏"];
 	#currentFrame = 0;
@@ -27,8 +26,8 @@ export class Loader extends Text {
 
 	constructor(
 		ui: TUI,
-		private spinnerColorFn: (str: string) => string,
-		private messageColorFn: (str: string) => string,
+		private spinnerColorFn: ColorFn,
+		private messageColorFn: LoaderMessageColorFn,
 		private message: string = "Loading...",
 		spinnerFrames?: string[],
 	) {
@@ -54,14 +53,17 @@ export class Loader extends Text {
 	start() {
 		this.#lastSpinnerTick = performance.now();
 		this.#updateDisplay();
+		const intervalMs = this.messageColorFn.animated === true ? RENDER_INTERVAL_MS : SPINNER_ADVANCE_MS;
 		this.#intervalId = setInterval(() => {
 			const now = performance.now();
-			if (now - this.#lastSpinnerTick >= SPINNER_ADVANCE_MS) {
-				this.#currentFrame = (this.#currentFrame + 1) % this.#frames.length;
-				this.#lastSpinnerTick = now;
+			const elapsed = now - this.#lastSpinnerTick;
+			if (elapsed >= SPINNER_ADVANCE_MS) {
+				const steps = Math.floor(elapsed / SPINNER_ADVANCE_MS);
+				this.#currentFrame = (this.#currentFrame + steps) % this.#frames.length;
+				this.#lastSpinnerTick += steps * SPINNER_ADVANCE_MS;
 			}
 			this.#updateDisplay();
-		}, RENDER_INTERVAL_MS);
+		}, intervalMs);
 	}
 
 	stop() {
@@ -71,7 +73,15 @@ export class Loader extends Text {
 		}
 	}
 
+	/** Lifecycle teardown: stop the animation timer. Idempotent. */
+	dispose() {
+		this.stop();
+	}
+
 	setMessage(message: string) {
+		if (message === this.message) {
+			return;
+		}
 		this.message = message;
 		this.#updateDisplay();
 	}
diff --git a/packages/tui/src/components/markdown.ts b/packages/tui/src/components/markdown.ts
index c4fbc8772..db96fd2ed 100644
--- a/packages/tui/src/components/markdown.ts
+++ b/packages/tui/src/components/markdown.ts
@@ -58,7 +58,8 @@ markdownParser.setOptions({
 // (Rust FFI) work for content/layout combinations already seen this session.
 
 const RENDER_CACHE_MAX = 256; // sane cap: ~256 distinct message × width combos
-const renderCache = new LRUCache<string, string[]>({ max: RENDER_CACHE_MAX });
+const EMPTY_RENDER_LINES: readonly string[] = [];
+const renderCache = new LRUCache<string, readonly string[]>({ max: RENDER_CACHE_MAX });
 
 /** Drop all L2 cache entries. Call on theme change to prevent stale styled output. */
 export function clearRenderCache(): void {
@@ -288,10 +289,11 @@ export class Markdown implements Component {
 	/** Number of spaces used to indent code block content. */
 	#codeBlockIndent: number;
 
-	// Cache for rendered output
+	// Cache for rendered output. Cached arrays are internal snapshots; render()
+	// returns caller-owned arrays because several renderers append surrounding rows.
 	#cachedText?: string;
 	#cachedWidth?: number;
-	#cachedLines?: string[];
+	#cachedLines?: readonly string[];
 
 	constructor(
 		text: string,
@@ -324,7 +326,7 @@ export class Markdown implements Component {
 		// L1: per-instance cache — fastest path for repeated renders of the same
 		// instance at the same width (e.g. resize debounce, repeated redraws).
 		if (this.#cachedLines && this.#cachedText === this.#text && this.#cachedWidth === width) {
-			return this.#cachedLines;
+			return this.#cachedLines.slice();
 		}
 
 		// Calculate available width for content (subtract horizontal padding)
@@ -332,12 +334,10 @@ export class Markdown implements Component {
 
 		// Don't render anything if there's no actual text
 		if (!this.#text || this.#text.trim() === "") {
-			const result: string[] = [];
-			// Update per-instance cache
 			this.#cachedText = this.#text;
 			this.#cachedWidth = width;
-			this.#cachedLines = result;
-			return result;
+			this.#cachedLines = EMPTY_RENDER_LINES;
+			return [];
 		}
 
 		// Replace tabs with 3 spaces for consistent rendering
@@ -364,7 +364,7 @@ export class Markdown implements Component {
 			this.#cachedText = this.#text;
 			this.#cachedWidth = width;
 			this.#cachedLines = cached;
-			return cached;
+			return cached.slice();
 		}
 
 		// Parse markdown to HTML-like tokens
@@ -445,14 +445,16 @@ export class Markdown implements Component {
 		const rawResult = [...emptyLines, ...contentLines, ...emptyLines];
 		const result = rawResult.length > 0 ? rawResult : [""];
 
-		// Update L1 per-instance cache
+		// Update caches with a private snapshot. The returned array remains owned by
+		// the caller, so push/splice by tool renderers cannot poison future redraws.
+		const cachedLines = result.slice();
 		this.#cachedText = this.#text;
 		this.#cachedWidth = width;
-		this.#cachedLines = result;
+		this.#cachedLines = cachedLines;
 
 		// Update L2 module-level LRU so future instances with the same key skip
 		// the marked.lexer + highlightCode (Rust FFI) work entirely.
-		renderCache.set(cacheKey, result);
+		renderCache.set(cacheKey, cachedLines);
 
 		return result;
 	}
diff --git a/packages/tui/src/components/scroll-view.ts b/packages/tui/src/components/scroll-view.ts
new file mode 100644
index 000000000..1bb5de9bd
--- /dev/null
+++ b/packages/tui/src/components/scroll-view.ts
@@ -0,0 +1,227 @@
+import { matchesKey } from "../keys";
+import type { Component } from "../tui";
+import { Ellipsis, replaceTabs, truncateToWidth, visibleWidth } from "../utils";
+
+const DEFAULT_TRACK = "│";
+const DEFAULT_THUMB = "█";
+
+type ScrollbarMode = "auto" | "always" | "never";
+
+export interface ScrollViewTheme {
+	track?: (text: string) => string;
+	thumb?: (text: string) => string;
+}
+
+export interface ScrollViewOptions {
+	height: number;
+	/** Defaults to "auto". "auto" reserves a scrollbar column only when content overflows. */
+	scrollbar?: ScrollbarMode | boolean;
+	/** Logical row count for pre-windowed line slices. Defaults to lines.length. */
+	totalRows?: number;
+	theme?: ScrollViewTheme;
+	trackChar?: string;
+	thumbChar?: string;
+	/**
+	 * Indicator appended when a row overflows `contentWidth`. Defaults to
+	 * {@link Ellipsis.Unicode}. Pass {@link Ellipsis.Omit} when callers wrap
+	 * lines to width themselves and only trailing padding can overflow (e.g.
+	 * the plan-review overlay), so no stray `…` lands on every padded row.
+	 */
+	ellipsis?: Ellipsis;
+	/**
+	 * Rows moved per keystroke when {@link ScrollView.handleScrollKey} sees a
+	 * Shift+Arrow (the "scroll faster" affordance). Defaults to 5.
+	 */
+	fastScrollLines?: number;
+}
+
+function normalizeScrollbarMode(scrollbar: ScrollViewOptions["scrollbar"]): ScrollbarMode {
+	if (scrollbar === true) return "auto";
+	if (scrollbar === false) return "never";
+	return scrollbar ?? "auto";
+}
+
+function firstCellGlyph(value: string, fallback: string): string {
+	const glyph = Array.from(value)[0] ?? fallback;
+	return visibleWidth(glyph) === 1 ? glyph : fallback;
+}
+
+/**
+ * Fixed-height viewport over pre-rendered lines, with optional right-edge scrollbar.
+ *
+ * ScrollView owns only the row offset. Callers remain responsible for producing
+ * already-wrapped logical lines appropriate for the current render width.
+ */
+export class ScrollView implements Component {
+	#lines: string[];
+	#height: number;
+	#scrollOffset = 0;
+	#totalRows: number | undefined;
+	#scrollbar: ScrollbarMode;
+	#theme: Required<ScrollViewTheme>;
+	#trackChar: string;
+	#thumbChar: string;
+	#ellipsis: Ellipsis;
+	#fastScrollLines: number;
+
+	constructor(lines: readonly string[], options: ScrollViewOptions) {
+		this.#lines = [...lines];
+		this.#height = Number.isFinite(options.height) ? Math.max(0, Math.trunc(options.height)) : 0;
+		this.#totalRows = options.totalRows === undefined ? undefined : Math.max(0, Math.trunc(options.totalRows));
+		this.#scrollbar = normalizeScrollbarMode(options.scrollbar);
+		this.#theme = {
+			track: options.theme?.track ?? (text => text),
+			thumb: options.theme?.thumb ?? (text => text),
+		};
+		this.#trackChar = firstCellGlyph(options.trackChar ?? DEFAULT_TRACK, DEFAULT_TRACK);
+		this.#thumbChar = firstCellGlyph(options.thumbChar ?? DEFAULT_THUMB, DEFAULT_THUMB);
+		this.#ellipsis = options.ellipsis ?? Ellipsis.Unicode;
+		this.#fastScrollLines = Math.max(1, Math.trunc(options.fastScrollLines ?? 5));
+		this.#clampScrollOffset();
+	}
+
+	setLines(lines: readonly string[]): void {
+		this.#lines = [...lines];
+		this.#clampScrollOffset();
+	}
+
+	setTotalRows(totalRows: number | undefined): void {
+		this.#totalRows = totalRows === undefined ? undefined : Math.max(0, Math.trunc(totalRows));
+		this.#clampScrollOffset();
+	}
+
+	setHeight(height: number): void {
+		this.#height = Number.isFinite(height) ? Math.max(0, Math.trunc(height)) : 0;
+		this.#clampScrollOffset();
+	}
+
+	setScrollbar(scrollbar: ScrollViewOptions["scrollbar"]): void {
+		this.#scrollbar = normalizeScrollbarMode(scrollbar);
+	}
+
+	getScrollOffset(): number {
+		return this.#scrollOffset;
+	}
+
+	getMaxScrollOffset(): number {
+		const rowCount = this.#totalRows ?? this.#lines.length;
+		return Math.max(0, rowCount - this.#height);
+	}
+
+	setScrollOffset(offset: number): void {
+		this.#scrollOffset = Number.isFinite(offset) ? Math.trunc(offset) : 0;
+		this.#clampScrollOffset();
+	}
+
+	scroll(delta: number): void {
+		this.setScrollOffset(this.#scrollOffset + (Number.isFinite(delta) ? Math.trunc(delta) : 0));
+	}
+
+	page(delta: number): void {
+		const step = Math.max(1, this.#height - 1);
+		this.scroll(step * (Number.isFinite(delta) ? Math.trunc(delta) : 0));
+	}
+
+	scrollToTop(): void {
+		this.#scrollOffset = 0;
+	}
+
+	scrollToBottom(): void {
+		this.#scrollOffset = this.getMaxScrollOffset();
+	}
+
+	/**
+	 * Apply a standard navigation key to the viewport. Shift+Arrow scrolls by
+	 * {@link ScrollViewOptions.fastScrollLines} (the "scroll faster" affordance);
+	 * plain Arrow by one line; PageUp/PageDown by a page; Home/End to the ends.
+	 * Returns true when the key was consumed, so callers can fall through to
+	 * their own (e.g. vim-style) bindings. Generic on purpose: every ScrollView
+	 * consumer gets the same scroll keys, including Shift-to-go-faster.
+	 */
+	handleScrollKey(data: string): boolean {
+		if (matchesKey(data, "shift+up")) {
+			this.scroll(-this.#fastScrollLines);
+			return true;
+		}
+		if (matchesKey(data, "shift+down")) {
+			this.scroll(this.#fastScrollLines);
+			return true;
+		}
+		if (matchesKey(data, "up")) {
+			this.scroll(-1);
+			return true;
+		}
+		if (matchesKey(data, "down")) {
+			this.scroll(1);
+			return true;
+		}
+		if (matchesKey(data, "pageUp")) {
+			this.page(-1);
+			return true;
+		}
+		if (matchesKey(data, "pageDown")) {
+			this.page(1);
+			return true;
+		}
+		if (matchesKey(data, "home")) {
+			this.scrollToTop();
+			return true;
+		}
+		if (matchesKey(data, "end")) {
+			this.scrollToBottom();
+			return true;
+		}
+		return false;
+	}
+
+	invalidate(): void {
+		// No cached layout to invalidate.
+	}
+
+	render(width: number): string[] {
+		this.#clampScrollOffset();
+		const safeWidth = Number.isFinite(width) ? Math.max(0, Math.trunc(width)) : 0;
+		if (this.#height === 0) return [];
+		const showScrollbar = safeWidth > 0 && this.#shouldRenderScrollbar();
+		const contentWidth = Math.max(0, safeWidth - (showScrollbar ? 1 : 0));
+		const thumb = showScrollbar ? this.#thumbRange() : undefined;
+		const lines: string[] = [];
+		for (let row = 0; row < this.#height; row++) {
+			const sourceIndex = this.#totalRows === undefined ? this.#scrollOffset + row : row;
+			const source = this.#lines[sourceIndex] ?? "";
+			const truncated = truncateToWidth(replaceTabs(source), contentWidth, this.#ellipsis);
+			if (!showScrollbar) {
+				lines.push(truncated);
+				continue;
+			}
+			const content = `${truncated}${" ".repeat(Math.max(0, contentWidth - visibleWidth(truncated)))}`;
+			const barGlyph = thumb && row >= thumb.start && row < thumb.end ? this.#thumbChar : this.#trackChar;
+			const styledBar =
+				thumb && row >= thumb.start && row < thumb.end ? this.#theme.thumb(barGlyph) : this.#theme.track(barGlyph);
+			lines.push(`${content}${styledBar}`);
+		}
+		return lines;
+	}
+
+	#clampScrollOffset(): void {
+		this.#scrollOffset = Math.max(0, Math.min(this.#scrollOffset, this.getMaxScrollOffset()));
+	}
+
+	#shouldRenderScrollbar(): boolean {
+		if (this.#height <= 0) return false;
+		if (this.#scrollbar === "never") return false;
+		if (this.#scrollbar === "always") return true;
+		return (this.#totalRows ?? this.#lines.length) > this.#height;
+	}
+
+	#thumbRange(): { start: number; end: number } {
+		if (this.#height <= 0) return { start: 0, end: 0 };
+		const rowCount = this.#totalRows ?? this.#lines.length;
+		if (rowCount <= this.#height) return { start: 0, end: this.#height };
+		const thumbSize = Math.max(1, Math.min(Math.floor((this.#height * this.#height) / rowCount), this.#height));
+		const travel = this.#height - thumbSize;
+		const maxOffset = this.getMaxScrollOffset();
+		const start = maxOffset === 0 ? 0 : Math.round((this.#scrollOffset / maxOffset) * travel);
+		return { start, end: start + thumbSize };
+	}
+}
diff --git a/packages/tui/src/components/select-list.ts b/packages/tui/src/components/select-list.ts
index 5387fda44..dfe95a920 100644
--- a/packages/tui/src/components/select-list.ts
+++ b/packages/tui/src/components/select-list.ts
@@ -4,6 +4,7 @@ import { extractPrintableText } from "../keys";
 import type { SymbolTheme } from "../symbols";
 import type { Component } from "../tui";
 import { Ellipsis, padding, replaceTabs, truncateToWidth, visibleWidth } from "../utils";
+import { ScrollView } from "./scroll-view";
 
 const DEFAULT_PRIMARY_COLUMN_WIDTH = 32;
 const PRIMARY_COLUMN_GAP = 2;
@@ -104,17 +105,29 @@ export class SelectList implements Component {
 		const endIndex = Math.min(startIndex + this.maxVisible, this.#filteredItems.length);
 
 		// Render visible items
+		const overflow = this.#filteredItems.length > this.maxVisible;
+		const rowWidth = Math.max(0, width - (overflow ? 1 : 0));
+		const rows: string[] = [];
 		for (let i = startIndex; i < endIndex; i++) {
 			const item = this.#filteredItems[i];
 			if (!item) continue;
 
 			const isSelected = i === this.#selectedIndex;
 			const descriptionText = item.description ? sanitizeSingleLine(item.description) : undefined;
-			lines.push(this.#renderItem(item, isSelected, width, descriptionText, primaryColumnWidth));
+			rows.push(this.#renderItem(item, isSelected, rowWidth, descriptionText, primaryColumnWidth));
 		}
 
-		// Add scroll/search status when needed
-		if (startIndex > 0 || endIndex < this.#filteredItems.length || showSearchStatus) {
+		const sv = new ScrollView(rows, {
+			height: rows.length,
+			scrollbar: "auto",
+			totalRows: this.#filteredItems.length,
+			theme: { track: t => this.theme.scrollInfo(t), thumb: t => this.theme.selectedPrefix(t) },
+		});
+		sv.setScrollOffset(startIndex);
+		lines.push(...sv.render(width));
+
+		// Add search status when relevant (scrollbar now indicates overflow)
+		if (showSearchStatus) {
 			lines.push(this.#renderStatusLine(width));
 		}
 
@@ -247,15 +260,8 @@ export class SelectList implements Component {
 	}
 
 	#renderStatusLine(width: number): string {
-		const selectedCount = this.#filteredItems.length === 0 ? 0 : this.#selectedIndex + 1;
-		const filteredCount = this.#filteredItems.length;
-		const count =
-			this.#filterQuery.trim() && filteredCount !== this.items.length
-				? `${selectedCount}/${filteredCount} of ${this.items.length}`
-				: `${selectedCount}/${filteredCount}`;
 		const query = sanitizeSingleLine(this.#filterQuery);
-		const searchSuffix = this.#shouldRenderSearchStatus() ? (query ? `  Search: ${query}` : "  Type to search") : "";
-		const statusText = `  (${count})${searchSuffix}`;
+		const statusText = query ? `  Search: ${query}` : "  Type to search";
 		return this.theme.scrollInfo(truncateToWidth(statusText, Math.max(1, width - 2), Ellipsis.Omit));
 	}
 
diff --git a/packages/tui/src/components/settings-list.ts b/packages/tui/src/components/settings-list.ts
index 3024ddb88..ef643823b 100644
--- a/packages/tui/src/components/settings-list.ts
+++ b/packages/tui/src/components/settings-list.ts
@@ -1,6 +1,7 @@
 import { getKeybindings } from "../keybindings";
 import type { Component } from "../tui";
 import { Ellipsis, padding, truncateToWidth, visibleWidth, wrapTextWithAnsi } from "../utils";
+import { ScrollView } from "./scroll-view";
 
 export interface SettingItem {
 	/** Unique identifier for this setting */
@@ -90,6 +91,22 @@ export class SettingsList implements Component {
 		return this.#renderMainList(width);
 	}
 
+	#renderItemRow(item: SettingItem, index: number, maxLabelWidth: number, rowWidth: number): string {
+		const isSelected = index === this.#selectedIndex;
+		const prefix = isSelected ? this.#theme.cursor : "  ";
+		const prefixWidth = visibleWidth(prefix);
+		const labelPadded = item.label + padding(Math.max(0, maxLabelWidth - visibleWidth(item.label)));
+		const labelText = this.#theme.label(labelPadded, isSelected, item.changed === true);
+		const separator = "  ";
+		const valueMaxWidth = rowWidth - prefixWidth - maxLabelWidth - visibleWidth(separator) - 2;
+		const valueText = this.#theme.value(
+			truncateToWidth(item.currentValue, valueMaxWidth, Ellipsis.Omit),
+			isSelected,
+			item.changed === true,
+		);
+		return truncateToWidth(prefix + labelText + separator + valueText, Math.max(0, rowWidth));
+	}
+
 	#renderMainList(width: number): string[] {
 		const lines: string[] = [];
 
@@ -98,48 +115,29 @@ export class SettingsList implements Component {
 			return lines;
 		}
 
-		// Calculate visible range with scrolling
+		const viewportHeight = Math.min(this.#maxVisible, this.#items.length);
 		const startIndex = Math.max(
 			0,
-			Math.min(this.#selectedIndex - Math.floor(this.#maxVisible / 2), this.#items.length - this.#maxVisible),
+			Math.min(this.#selectedIndex - Math.floor(viewportHeight / 2), this.#items.length - viewportHeight),
 		);
-		const endIndex = Math.min(startIndex + this.#maxVisible, this.#items.length);
-
-		// Calculate max label width for alignment
 		const maxLabelWidth = Math.min(30, Math.max(...this.#items.map(item => visibleWidth(item.label))));
-
-		// Render visible items
-		for (let i = startIndex; i < endIndex; i++) {
-			const item = this.#items[i];
-			if (!item) continue;
-
-			const isSelected = i === this.#selectedIndex;
-			const prefix = isSelected ? this.#theme.cursor : "  ";
-			const prefixWidth = visibleWidth(prefix);
-
-			// Pad label to align values
-			const labelPadded = item.label + padding(Math.max(0, maxLabelWidth - visibleWidth(item.label)));
-			const labelText = this.#theme.label(labelPadded, isSelected, item.changed === true);
-
-			// Calculate space for value
-			const separator = "  ";
-			const usedWidth = prefixWidth + maxLabelWidth + visibleWidth(separator);
-			const valueMaxWidth = width - usedWidth - 2;
-
-			const valueText = this.#theme.value(
-				truncateToWidth(item.currentValue, valueMaxWidth, Ellipsis.Omit),
-				isSelected,
-				item.changed === true,
-			);
-
-			lines.push(truncateToWidth(prefix + labelText + separator + valueText, width));
-		}
-
-		// Add scroll indicator if needed
-		if (startIndex > 0 || endIndex < this.#items.length) {
-			const scrollText = `  (${this.#selectedIndex + 1}/${this.#items.length})`;
-			lines.push(this.#theme.hint(truncateToWidth(scrollText, width - 2, Ellipsis.Omit)));
-		}
+		const itemRowsOverflow = this.#items.length > viewportHeight;
+		const itemRowWidth = Math.max(0, width - (itemRowsOverflow ? 1 : 0));
+		const visibleItems = this.#items.slice(startIndex, startIndex + viewportHeight);
+		const itemRows = visibleItems.map((item, index) =>
+			this.#renderItemRow(item, startIndex + index, maxLabelWidth, itemRowWidth),
+		);
+		const scrollView = new ScrollView(itemRows, {
+			height: viewportHeight,
+			scrollbar: "auto",
+			totalRows: this.#items.length,
+			theme: {
+				track: text => this.#theme.hint(text),
+				thumb: text => this.#theme.label(text, true, false),
+			},
+		});
+		scrollView.setScrollOffset(startIndex);
+		lines.push(...scrollView.render(width));
 
 		// Add description for selected item
 		const selectedItem = this.#items[this.#selectedIndex];
diff --git a/packages/tui/src/components/text.ts b/packages/tui/src/components/text.ts
index b41145a53..c90c32f7b 100644
--- a/packages/tui/src/components/text.ts
+++ b/packages/tui/src/components/text.ts
@@ -27,6 +27,9 @@ export class Text implements Component {
 	}
 
 	setText(text: string): void {
+		if (text === this.#text) {
+			return;
+		}
 		this.#text = text;
 		this.#cachedText = undefined;
 		this.#cachedWidth = undefined;
diff --git a/packages/tui/src/index.ts b/packages/tui/src/index.ts
index f7e3832f5..33928af46 100644
--- a/packages/tui/src/index.ts
+++ b/packages/tui/src/index.ts
@@ -10,6 +10,7 @@ export * from "./components/image";
 export * from "./components/input";
 export * from "./components/loader";
 export * from "./components/markdown";
+export * from "./components/scroll-view";
 export * from "./components/select-list";
 export * from "./components/settings-list";
 export * from "./components/spacer";
@@ -26,7 +27,7 @@ export * from "./fuzzy";
 export * from "./keybindings";
 // Kitty keyboard protocol helpers
 export * from "./keys";
-// Kitty graphics: Unicode placeholders + temp-file transmission
+// Kitty graphics: Unicode placeholders
 export * from "./kitty-graphics";
 // Mermaid diagram support
 // Input buffering for batch splitting
diff --git a/packages/tui/src/keybindings.ts b/packages/tui/src/keybindings.ts
index 134965c42..e03954f09 100644
--- a/packages/tui/src/keybindings.ts
+++ b/packages/tui/src/keybindings.ts
@@ -1,4 +1,4 @@
-import { type KeyId, matchesKey, parseKey } from "./keys";
+import { type KeyId, parseKey } from "./keys";
 
 /**
  * Global keybinding registry.
@@ -164,6 +164,64 @@ const SHIFTED_SYMBOL_KEYS = new Set<string>([
 	"~",
 ]);
 
+const MODIFIER_ORDER = ["ctrl", "shift", "alt", "super"] as const;
+
+function startsWithModifier(key: string, offset: number, modifier: string): boolean {
+	if (key.length <= offset + modifier.length || key.charCodeAt(offset + modifier.length) !== 43) return false;
+	for (let i = 0; i < modifier.length; i++) {
+		const actual = key.charCodeAt(offset + i);
+		const expected = modifier.charCodeAt(i);
+		if (actual !== expected && actual !== expected - 32) return false;
+	}
+	return true;
+}
+
+function isAsciiUppercaseLetter(key: string): boolean {
+	if (key.length !== 1) return false;
+	const code = key.charCodeAt(0);
+	return code >= 65 && code <= 90;
+}
+
+function canonicalKeyId(key: string): string {
+	let offset = 0;
+	const modifiers: string[] = [];
+	let foundModifier = true;
+
+	while (foundModifier) {
+		foundModifier = false;
+		for (const modifier of MODIFIER_ORDER) {
+			if (startsWithModifier(key, offset, modifier)) {
+				modifiers.push(modifier);
+				offset += modifier.length + 1;
+				foundModifier = true;
+				break;
+			}
+		}
+	}
+	const rawBase = key.slice(offset);
+	const lowerBase = rawBase.toLowerCase();
+	const base = lowerBase === "esc" ? "escape" : lowerBase === "return" ? "enter" : lowerBase;
+	if (isAsciiUppercaseLetter(rawBase) && !modifiers.includes("shift")) {
+		modifiers.push("shift");
+	}
+
+	if (modifiers.length === 0) return base;
+	modifiers.sort(
+		(left, right) =>
+			MODIFIER_ORDER.indexOf(left as (typeof MODIFIER_ORDER)[number]) -
+			MODIFIER_ORDER.indexOf(right as (typeof MODIFIER_ORDER)[number]),
+	);
+	return `${modifiers.join("+")}+${base}`;
+}
+
+function addKeyAliases(keys: Set<string>, key: KeyId): void {
+	const canonical = canonicalKeyId(key);
+	keys.add(canonical);
+	if (SHIFTED_SYMBOL_KEYS.has(canonical)) {
+		keys.add(`shift+${canonical}`);
+	}
+}
+
 const normalizeKeyId = (key: KeyId): KeyId => key.toLowerCase() as KeyId;
 
 function normalizeKeys(keys: KeyId | KeyId[] | undefined): KeyId[] {
@@ -185,6 +243,7 @@ export class KeybindingsManager {
 	#definitions: KeybindingDefinitions;
 	#userBindings: KeybindingsConfig;
 	#keysById = new Map<Keybinding, KeyId[]>();
+	#matchKeysById = new Map<Keybinding, Set<string>>();
 	#conflicts: KeybindingConflict[] = [];
 
 	constructor(definitions: KeybindingDefinitions, userBindings: KeybindingsConfig = {}) {
@@ -195,6 +254,7 @@ export class KeybindingsManager {
 
 	#rebuild(): void {
 		this.#keysById.clear();
+		this.#matchKeysById.clear();
 		this.#conflicts = [];
 
 		const userClaims = new Map<KeyId, Set<Keybinding>>();
@@ -217,21 +277,19 @@ export class KeybindingsManager {
 			const userKeys = this.#userBindings[id];
 			const keys = userKeys === undefined ? normalizeKeys(definition.defaultKeys) : normalizeKeys(userKeys);
 			this.#keysById.set(id as Keybinding, keys);
+			const matchKeys = new Set<string>();
+			for (const key of keys) {
+				addKeyAliases(matchKeys, key);
+			}
+			this.#matchKeysById.set(id as Keybinding, matchKeys);
 		}
 	}
 
 	matches(data: string, keybinding: Keybinding): boolean {
-		const keys = this.#keysById.get(keybinding) ?? [];
-		for (const key of keys) {
-			if (matchesKey(data, key)) return true;
-		}
-
-		// Handle shifted symbol keys (e.g., shift+- produces _ on US layout)
 		const parsed = parseKey(data);
-		if (!parsed?.startsWith("shift+")) return false;
-		const keyName = parsed.slice("shift+".length);
-		if (!SHIFTED_SYMBOL_KEYS.has(keyName)) return false;
-		return keys.includes(keyName as KeyId);
+		if (parsed === undefined) return false;
+		const matchKeys = this.#matchKeysById.get(keybinding);
+		return matchKeys?.has(canonicalKeyId(parsed)) ?? false;
 	}
 
 	getKeys(keybinding: Keybinding): KeyId[] {
diff --git a/packages/tui/src/kitty-graphics.ts b/packages/tui/src/kitty-graphics.ts
index 8b65bf5d8..0b403d217 100644
--- a/packages/tui/src/kitty-graphics.ts
+++ b/packages/tui/src/kitty-graphics.ts
@@ -1,6 +1,6 @@
 /**
- * Kitty graphics: Unicode placeholder placement (`U=1` + U+10EEEE) and temp-file
- * (`t=t`) image transmission, with runtime feature state and env overrides.
+ * Kitty graphics: Unicode placeholder placement (`U=1` + U+10EEEE), with
+ * runtime feature state and env overrides.
  *
  * Unicode placeholders let a transmitted image be displayed by writing ordinary
  * text cells — the placeholder char U+10EEEE plus row/column combining
@@ -14,10 +14,6 @@
  * dependency stays one-way (capabilities → kitty-graphics) and no import cycle
  * forms. Protocol gating (`imageProtocol === Kitty`) lives in the caller.
  */
-import * as fs from "node:fs";
-import * as os from "node:os";
-import * as path from "node:path";
-import { $env, logger } from "@oh-my-pi/pi-utils";
 
 /** Kitty Unicode placeholder base character (U+10EEEE, Plane 16 PUA). */
 export const KITTY_PLACEHOLDER = "\u{10eeee}";
@@ -55,25 +51,9 @@ const ROWCOLUMN_DIACRITICS: readonly number[] = [
 /** Largest row/column index expressible with the diacritic table (one cell each). */
 export const KITTY_PLACEHOLDER_MAX_CELLS = ROWCOLUMN_DIACRITICS.length;
 
-/** A minimal opaque 1x1 PNG (base64) used for the temp-file support probe. */
-const PROBE_PNG_BASE64 =
-	"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8z8BQDwAEhQGAhKmMIQAAAABJRU5ErkJggg==";
-
-export type KittyTransmissionMedium = "direct" | "temp-file";
-
 export interface KittyGraphicsFeatures {
 	/** Display images via Unicode placeholders instead of direct `a=p` placement. */
 	unicodePlaceholders: boolean;
-	/** How image data reaches the terminal: in-band base64 or a temp file. */
-	transmissionMedium: KittyTransmissionMedium;
-}
-
-/** Explicit transmission medium from `PI_KITTY_IMAGE_TRANSMISSION`, else "auto". */
-function transmissionOverride(): KittyTransmissionMedium | "auto" {
-	const raw = $env.PI_KITTY_IMAGE_TRANSMISSION?.trim().toLowerCase();
-	if (raw === "temp-file") return "temp-file";
-	if (raw === "direct") return "direct";
-	return "auto";
 }
 
 /**
@@ -106,8 +86,6 @@ let features: KittyGraphicsFeatures = {
 	// Off until `terminal-capabilities` seeds it from the detected terminal id —
 	// the default-on path corrupts wezterm and tmux-passthrough sessions.
 	unicodePlaceholders: false,
-	// Start direct; a successful probe (or explicit `temp-file` override) promotes.
-	transmissionMedium: transmissionOverride() === "temp-file" ? "temp-file" : "direct",
 };
 
 export function getKittyGraphics(): Readonly<KittyGraphicsFeatures> {
@@ -118,29 +96,11 @@ export function setKittyGraphics(partial: Partial<KittyGraphicsFeatures>): void
 	features = { ...features, ...partial };
 }
 
-/**
- * Whether temp-file transmission may be promoted at runtime: forced via env,
- * disabled via env, otherwise auto (local sessions only — a temp file written
- * locally is not readable by a terminal on the far side of an SSH link).
- */
-export function kittyTempFileAllowed(): boolean {
-	const override = transmissionOverride();
-	if (override === "temp-file") return true;
-	if (override === "direct") return false;
-	// Auto: local sessions only — a temp file written here is not on the SSH peer.
-	return !$env.SSH_CONNECTION && !$env.SSH_CLIENT && !$env.SSH_TTY;
-}
-
 /** Whether a `columns`×`rows` placeholder grid fits within the diacritic table. */
 export function kittyPlaceholdersFit(columns: number, rows: number): boolean {
 	return columns >= 1 && rows >= 1 && columns <= KITTY_PLACEHOLDER_MAX_CELLS && rows <= KITTY_PLACEHOLDER_MAX_CELLS;
 }
 
-/** True when the base64 payload is a PNG (kitty `f=100` / temp-file path only). */
-export function isPngBase64(base64Data: string): boolean {
-	return base64Data.startsWith("iVBORw0KGgo");
-}
-
 function diacritic(index: number): string {
 	const cp = ROWCOLUMN_DIACRITICS[index];
 	return cp === undefined ? "" : String.fromCodePoint(cp);
@@ -209,62 +169,3 @@ export function renderKittyPlaceholderLines(opts: {
 	}
 	return grid;
 }
-
-/**
- * Path for a temp-file transmission. Kitty deletes a `t=t` file after reading it
- * **only** when the path contains the substring `tty-graphics-protocol`, so it is
- * embedded in the filename to keep the temp dir self-cleaning.
- */
-function tempGraphicsPath(tag: string): string {
-	return path.join(os.tmpdir(), `tty-graphics-protocol-${tag}-${process.pid}-${Date.now()}.png`);
-}
-
-/**
- * Transmit a PNG via a temp file (`t=t`): decode the base64 to bytes once, write
- * them to a temp file, and send the base64-encoded file path as payload. Returns
- * the APC string, or `null` on any failure (caller falls back to direct base64).
- *
- * Synchronous filesystem writes are mandated by the synchronous render pipeline
- * (`Image.render` → `renderImage` are sync); there is no async seam here.
- */
-export function encodeKittyTempFileTransmit(base64Png: string, imageId: number): string | null {
-	try {
-		const bytes = Buffer.from(base64Png, "base64");
-		if (bytes.length === 0) return null;
-		const file = tempGraphicsPath(`i${imageId}`);
-		fs.writeFileSync(file, bytes);
-		const encodedPath = Buffer.from(file, "utf8").toString("base64");
-		return `\x1b_Ga=t,f=100,t=t,S=${bytes.length},q=2,i=${imageId};${encodedPath}\x1b\\`;
-	} catch (err) {
-		logger.debug("Kitty temp-file transmit failed; using direct base64", { err: String(err) });
-		return null;
-	}
-}
-
-/**
- * Encode a temp-file support probe: write a tiny PNG to a temp file and ask the
- * terminal to query it (`a=q,t=t`). A conforming terminal replies
- * `ESC _ G i=<probeId>;OK ESC \`. Returns the query APC plus a `cleanup` that
- * removes the probe file (best-effort; kitty self-deletes the magic-named file).
- * Returns `null` if the temp file cannot be written.
- */
-export function encodeKittyTempFileProbe(probeId: number): { sequence: string; cleanup: () => void } | null {
-	try {
-		const bytes = Buffer.from(PROBE_PNG_BASE64, "base64");
-		const file = tempGraphicsPath(`probe${probeId}`);
-		fs.writeFileSync(file, bytes);
-		const encodedPath = Buffer.from(file, "utf8").toString("base64");
-		const sequence = `\x1b_Ga=q,t=t,f=100,S=${bytes.length},q=2,i=${probeId};${encodedPath}\x1b\\`;
-		const cleanup = () => {
-			try {
-				fs.rmSync(file, { force: true });
-			} catch {
-				// Best effort; kitty deletes tty-graphics-protocol files itself.
-			}
-		};
-		return { sequence, cleanup };
-	} catch (err) {
-		logger.debug("Kitty temp-file probe setup failed", { err: String(err) });
-		return null;
-	}
-}
diff --git a/packages/tui/src/terminal-capabilities.ts b/packages/tui/src/terminal-capabilities.ts
index bea4ed1f8..6b429d7cc 100644
--- a/packages/tui/src/terminal-capabilities.ts
+++ b/packages/tui/src/terminal-capabilities.ts
@@ -2,9 +2,7 @@ import { encodeSixel } from "@oh-my-pi/pi-natives";
 import { $env, isBunTestRuntime } from "@oh-my-pi/pi-utils";
 import {
 	detectKittyUnicodePlaceholdersSupport,
-	encodeKittyTempFileTransmit,
 	getKittyGraphics,
-	isPngBase64,
 	KITTY_PLACEHOLDER,
 	kittyPlaceholdersFit,
 	renderKittyPlaceholderLines,
@@ -25,7 +23,31 @@ export enum NotifyProtocol {
 
 export type TerminalId = "kitty" | "ghostty" | "wezterm" | "iterm2" | "vscode" | "alacritty" | "base" | "trueColor";
 
-const SIXEL_DCS_START_REGEX = /\x1bP(?:[0-9;]*)q/u;
+function hasNeedleBefore(line: string, needle: string, limit: number): boolean {
+	const index = line.indexOf(needle);
+	return index !== -1 && index + needle.length <= limit;
+}
+
+function hasSixelDcsStart(line: string): boolean {
+	const limit = Math.min(line.length, 128);
+	let from = 0;
+	for (;;) {
+		const start = line.indexOf("\x1bP", from);
+		if (start === -1 || start + 3 > limit) return false;
+		let i = start + 2;
+		while (i < limit) {
+			const code = line.charCodeAt(i);
+			if ((code >= 0x30 && code <= 0x39) || code === 0x3b) {
+				i++;
+				continue;
+			}
+			break;
+		}
+		if (i < limit && line.charCodeAt(i) === 0x71) return true;
+		from = start + 2;
+	}
+}
+
 /** Terminal capability details used for rendering and protocol selection. */
 export class TerminalInfo {
 	constructor(
@@ -41,13 +63,31 @@ export class TerminalInfo {
 		public readonly textSizing: boolean = false,
 	) {}
 
+	/**
+	 * Whether a prompt-submit keystroke scrolls this host to its tail, so the
+	 * native-scrollback reconciliation checkpoint may ED3-rebuild even when the
+	 * viewport position is unprobeable. Assigned by the TERMINAL builder from
+	 * {@link detectSubmitPinsViewportToTail}; readonly but tests opt in via the
+	 * {@link setTerminalSubmitPinsViewportToTail} mutable-cast setter.
+	 */
+	readonly submitPinsViewportToTail: boolean = false;
+
+	/**
+	 * Mutable clone for the {@link TERMINAL} singleton: copies every field and
+	 * keeps the prototype methods, so the builder and runtime setters flip
+	 * runtime-resolved {@link RuntimeTerminal} capabilities in place instead of
+	 * reconstructing positional constructor args.
+	 */
+	clone(): RuntimeTerminal {
+		return Object.assign(Object.create(TerminalInfo.prototype), this) as RuntimeTerminal;
+	}
+
 	isImageLine(line: string): boolean {
 		if (!this.imageProtocol) return false;
 		if (this.imageProtocol === ImageProtocol.Sixel) {
-			return SIXEL_DCS_START_REGEX.test(line.slice(0, 128));
+			return hasSixelDcsStart(line);
 		}
-		const head = line.slice(0, 64);
-		return head.includes(this.imageProtocol) || head.includes(KITTY_PLACEHOLDER);
+		return hasNeedleBefore(line, this.imageProtocol, 64) || hasNeedleBefore(line, KITTY_PLACEHOLDER, 64);
 	}
 
 	formatNotification(message: string | TerminalNotification): string {
@@ -187,41 +227,128 @@ export function detectTerminalEagerEraseScrollbackRisk(
 	return true;
 }
 
-/** Whether DEC 2026 synchronized-output wrappers should be enabled by default. */
-export function shouldEnableSynchronizedOutputByDefault(
+/**
+ * Whether a prompt-submit keystroke scrolls this terminal to its tail, making the
+ * native-scrollback reconciliation checkpoint (`refreshNativeScrollbackIfDirty`)
+ * safe to ED3-rebuild even when the viewport position cannot be probed.
+ *
+ * True only for recognized genuine *local* terminals where typing into the prompt
+ * brings the host viewport to the bottom. False — the checkpoint keeps deferring
+ * until a positive at-tail probe — for hosts whose scrollback a keystroke does not
+ * move: Windows consoles/ConPTY, Windows Terminal (incl. WSL), SSH, multiplexers,
+ * and unrecognized profiles. This is the per-terminal counterpart to the blanket
+ * block from #1610/#1682/#1746: those hosts genuinely cannot treat a submit as
+ * proof of at-tail, but genuine local terminals can.
+ */
+export function detectSubmitPinsViewportToTail(
 	env: NodeJS.ProcessEnv = Bun.env,
 	platform: NodeJS.Platform = process.platform,
-	terminalId: TerminalId = TERMINAL_ID,
 ): boolean {
-	if (env.PI_NO_SYNC_OUTPUT || env.PI_TUI_SYNC_OUTPUT === "0") return false;
-	if (env.PI_FORCE_SYNC_OUTPUT === "1" || env.PI_TUI_SYNC_OUTPUT === "1") return true;
+	if (env.PI_TUI_ED3_SAFE === "1") return true;
 	if (platform === "win32") return false;
-
+	if (env.WT_SESSION) return false;
+	if (env.SSH_CONNECTION || env.SSH_CLIENT || env.SSH_TTY) return false;
 	const term = env.TERM?.toLowerCase() ?? "";
-	const termProgram = env.TERM_PROGRAM?.toLowerCase() ?? "";
-	if (
-		env.SSH_CONNECTION ||
-		env.SSH_CLIENT ||
-		env.SSH_TTY ||
-		env.TMUX ||
-		env.STY ||
-		env.ZELLIJ ||
-		term.startsWith("tmux") ||
-		term.startsWith("screen")
-	) {
+	if (env.TMUX || env.STY || env.ZELLIJ || term.startsWith("tmux") || term.startsWith("screen")) {
 		return false;
 	}
-	if (env.VTE_VERSION) return false;
-	switch (termProgram) {
-		case "gnome-terminal":
-		case "kgx":
-		case "ptyxis":
-		case "xfce4-terminal":
-			return false;
-		default:
-			break;
+	if (
+		env.WEZTERM_PANE ||
+		env.KITTY_WINDOW_ID ||
+		env.GHOSTTY_RESOURCES_DIR ||
+		env.ALACRITTY_WINDOW_ID ||
+		env.ITERM_SESSION_ID ||
+		env.VTE_VERSION
+	) {
+		return true;
+	}
+	switch (env.TERM_PROGRAM?.toLowerCase() ?? "") {
+		case "alacritty":
+		case "apple_terminal":
+		case "ghostty":
+		case "gnome-terminal":
+		case "iterm.app":
+		case "kgx":
+		case "kitty":
+		case "ptyxis":
+		case "wezterm":
+		case "xfce4-terminal":
+			return true;
+		default:
+			return false;
+	}
+}
+
+/**
+ * Resolve an explicit user override for DEC 2026 synchronized output. Returns
+ * `false` for an opt-out, `true` for a force-on, or `null` when the user has
+ * expressed no preference. Shared by the static default and the runtime DECRQM
+ * probe so both honor the same precedence — an opt-out beats a force-on.
+ */
+export function synchronizedOutputUserOverride(env: NodeJS.ProcessEnv = Bun.env): boolean | null {
+	if (env.PI_NO_SYNC_OUTPUT || env.PI_TUI_SYNC_OUTPUT === "0") return false;
+	if (env.PI_FORCE_SYNC_OUTPUT === "1" || env.PI_TUI_SYNC_OUTPUT === "1") return true;
+	return null;
+}
+
+/**
+ * Whether `TERM_FEATURES` advertises DEC 2026 synchronized output via the `Sy`
+ * capability token. `TERM_FEATURES` is a run of capitalized two-letter codes
+ * (e.g. `…Sy…`), so a case-sensitive substring match is unambiguous: `Sy`
+ * cannot straddle a code boundary because those are always lowercase→uppercase.
+ */
+function advertisesSynchronizedOutput(termFeatures: string | undefined): boolean {
+	return termFeatures?.includes("Sy") ?? false;
+}
+
+/**
+ * Whether DEC 2026 synchronized-output wrappers should be enabled by default.
+ *
+ * Policy (highest precedence first):
+ *   1. Explicit user override (`PI_NO_SYNC_OUTPUT`/`PI_TUI_SYNC_OUTPUT=0` off,
+ *      `PI_FORCE_SYNC_OUTPUT=1`/`PI_TUI_SYNC_OUTPUT=1` on).
+ *   2. Positive `TERM_FEATURES` advertisement (`Sy`) — survives SSH/mux wrapping.
+ *   3. Windows Terminal (1.24+) via `WT_SESSION`, on native win32 and the
+ *      WSL/SSH-fronted host alike.
+ *   4. Known direct terminals with confirmed support. SSH does *not* disable —
+ *      DEC 2026 passes through SSH when the outer terminal honors it.
+ *   5. Everything else starts off, including risky multiplexers; the runtime
+ *      DECRQM probe upgrades any of them when the terminal actually reports
+ *      `?2026` supported (current zellij, tmux master, foot, contour, mintty…).
+ */
+export function shouldEnableSynchronizedOutputByDefault(
+	env: NodeJS.ProcessEnv = Bun.env,
+	terminalId: TerminalId = TERMINAL_ID,
+): boolean {
+	const override = synchronizedOutputUserOverride(env);
+	if (override !== null) return override;
+
+	if (advertisesSynchronizedOutput(env.TERM_FEATURES)) return true;
+	if (env.WT_SESSION) return true;
+
+	// Risky multiplexers start off even when an inner terminal id leaks through:
+	// older tmux/screen synchronized-output handling is flaky and a mux may not
+	// pass DEC 2026 to the outer host. The DECRQM probe re-enables sync when the
+	// mux reports `?2026` supported.
+	const term = env.TERM?.toLowerCase() ?? "";
+	if (env.TMUX || env.STY || env.ZELLIJ || term.startsWith("tmux") || term.startsWith("screen")) {
+		return false;
+	}
+
+	switch (terminalId) {
+		case "kitty":
+		case "ghostty":
+		case "wezterm":
+		case "iterm2":
+		case "alacritty":
+		case "vscode":
+			return true;
+		default:
+			// VTE family, GNU screen, Apple Terminal, legacy native console host
+			// (no WT_SESSION), and bare/unknown xterm profiles stay off until the
+			// DECRQM probe proves support.
+			return false;
 	}
-	return terminalId === "kitty" || terminalId === "ghostty" || terminalId === "wezterm" || terminalId === "iterm2";
 }
 
 /**
@@ -317,54 +444,38 @@ export const TERMINAL_ID: TerminalId = (() => {
 	return "base";
 })();
 
-/** Clone a {@link TerminalInfo} with selected fields overridden, preserving the rest. */
-function withTerminalOverrides(
-	base: TerminalInfo,
-	overrides: {
-		imageProtocol?: ImageProtocol | null;
-		hyperlinks?: boolean;
-		eagerEraseScrollbackRisk?: boolean;
-		deccara?: boolean;
-		supportsScreenToScrollback?: boolean;
-	},
-): TerminalInfo {
-	return new TerminalInfo(
-		base.id,
-		overrides.imageProtocol !== undefined ? overrides.imageProtocol : base.imageProtocol,
-		base.trueColor,
-		overrides.hyperlinks !== undefined ? overrides.hyperlinks : base.hyperlinks,
-		base.notifyProtocol,
-		overrides.eagerEraseScrollbackRisk !== undefined
-			? overrides.eagerEraseScrollbackRisk
-			: base.eagerEraseScrollbackRisk,
-		overrides.deccara !== undefined ? overrides.deccara : base.deccara,
-		overrides.supportsScreenToScrollback !== undefined
-			? overrides.supportsScreenToScrollback
-			: base.supportsScreenToScrollback,
-	);
+/**
+ * The process-wide {@link TERMINAL} singleton: a {@link TerminalInfo} whose
+ * post-construction capabilities — the image protocol and the probe-driven
+ * flags — are writable, so the runtime setters and tests mutate them directly
+ * instead of through an unsound cast. Every other field stays readonly.
+ */
+export interface RuntimeTerminal extends TerminalInfo {
+	imageProtocol: ImageProtocol | null;
+	hyperlinks: boolean;
+	eagerEraseScrollbackRisk: boolean;
+	deccara: boolean;
+	supportsScreenToScrollback: boolean;
+	textSizing: boolean;
+	submitPinsViewportToTail: boolean;
 }
 
-export const TERMINAL = (() => {
-	let resolved = getTerminalInfo(TERMINAL_ID);
-	const eagerEraseScrollbackRisk = detectTerminalEagerEraseScrollbackRisk(Bun.env, process.platform);
-	if (resolved.eagerEraseScrollbackRisk !== eagerEraseScrollbackRisk) {
-		resolved = withTerminalOverrides(resolved, { eagerEraseScrollbackRisk });
-	}
+export const TERMINAL: RuntimeTerminal = (() => {
+	const resolved = getTerminalInfo(TERMINAL_ID).clone();
+	resolved.eagerEraseScrollbackRisk = detectTerminalEagerEraseScrollbackRisk(Bun.env, process.platform);
 
 	const forcedImageProtocol = getForcedImageProtocol();
 	if (forcedImageProtocol !== undefined) {
-		resolved = withTerminalOverrides(resolved, { imageProtocol: forcedImageProtocol });
+		resolved.imageProtocol = forcedImageProtocol;
 	} else if (!resolved.imageProtocol) {
 		const fallbackImageProtocol = getFallbackImageProtocol(resolved.id);
-		if (fallbackImageProtocol) {
-			resolved = withTerminalOverrides(resolved, { imageProtocol: fallbackImageProtocol });
-		}
+		if (fallbackImageProtocol) resolved.imageProtocol = fallbackImageProtocol;
 	}
 	// tmux and screen multiplexers do not reliably forward OSC 8 hyperlinks
 	// to the outer terminal, so force them off regardless of detected terminal.
 	const term = Bun.env.TERM?.toLowerCase() ?? "";
 	if (resolved.hyperlinks && (Bun.env.TMUX || term.startsWith("tmux") || term.startsWith("screen"))) {
-		resolved = withTerminalOverrides(resolved, { hyperlinks: false });
+		resolved.hyperlinks = false;
 	}
 	// DECCARA rectangular-SGR background fills. The static per-terminal capability
 	// lives on KNOWN_TERMINALS; here we fold in runtime context — multiplexer and
@@ -372,10 +483,12 @@ export const TERMINAL = (() => {
 	// off inside the test runtime so the xterm.js-backed virtual terminal (which
 	// ignores DECCARA) exercises the padded-string fallback. Integration tests opt
 	// in explicitly through setTerminalDeccara.
-	const deccara = detectRectangularSgrSupport(resolved.id, Bun.env) && !isBunTestRuntime();
-	if (resolved.deccara !== deccara) {
-		resolved = withTerminalOverrides(resolved, { deccara });
-	}
+	resolved.deccara = detectRectangularSgrSupport(resolved.id, Bun.env) && !isBunTestRuntime();
+	// A genuine local terminal scrolls to its tail on the submit keystroke, so the
+	// reconciliation checkpoint may ED3-rebuild on an unprobeable viewport there.
+	// Forced off under the test runtime (like deccara) so checkpoint tests stay
+	// deterministic and opt in through setTerminalSubmitPinsViewportToTail.
+	resolved.submitPinsViewportToTail = detectSubmitPinsViewportToTail(Bun.env, process.platform) && !isBunTestRuntime();
 	return resolved;
 })();
 
@@ -385,18 +498,11 @@ export const TERMINAL = (() => {
 // glyphs, which is the "ASCII artifact + laggy scrolling" reported in #1877.
 setKittyGraphics({ unicodePlaceholders: detectKittyUnicodePlaceholdersSupport(TERMINAL.id, Bun.env) });
 
-type MutableTerminalInfo = {
-	imageProtocol: ImageProtocol | null;
-	deccara: boolean;
-	supportsScreenToScrollback: boolean;
-	textSizing: boolean;
-};
-
 /**
  * Override terminal image protocol at runtime after capability probes complete.
  */
 export function setTerminalImageProtocol(imageProtocol: ImageProtocol | null): void {
-	(TERMINAL as unknown as MutableTerminalInfo).imageProtocol = imageProtocol;
+	TERMINAL.imageProtocol = imageProtocol;
 }
 
 /**
@@ -405,12 +511,17 @@ export function setTerminalImageProtocol(imageProtocol: ImageProtocol | null): v
  * resolved once at import and force-disabled under the test runtime.
  */
 export function setTerminalDeccara(enabled: boolean): void {
-	(TERMINAL as unknown as MutableTerminalInfo).deccara = enabled;
+	TERMINAL.deccara = enabled;
 }
 
 /** Override screen-to-scrollback clear support for targeted renderer tests. */
 export function setTerminalScreenToScrollback(enabled: boolean): void {
-	(TERMINAL as unknown as MutableTerminalInfo).supportsScreenToScrollback = enabled;
+	TERMINAL.supportsScreenToScrollback = enabled;
+}
+
+/** Override submit-pins-viewport-to-tail for checkpoint reconciliation tests. */
+export function setTerminalSubmitPinsViewportToTail(enabled: boolean): void {
+	TERMINAL.submitPinsViewportToTail = enabled;
 }
 
 /**
@@ -419,7 +530,7 @@ export function setTerminalScreenToScrollback(enabled: boolean): void {
  * capability); tests flip it directly to exercise the scaled-heading path.
  */
 export function setTerminalTextSizing(enabled: boolean): void {
-	(TERMINAL as unknown as MutableTerminalInfo).textSizing = enabled;
+	TERMINAL.textSizing = enabled;
 }
 
 export function getTerminalInfo(terminalId: TerminalId): TerminalInfo {
@@ -772,16 +883,11 @@ export function renderImage(
 		if (options.imageId != null) {
 			const placementId = options.placementId ?? options.imageId;
 			const graphics = getKittyGraphics();
-			// Transmit-once (keyed by id). Prefer a local temp file for PNGs when the
-			// medium has been promoted; otherwise send in-band base64. Repaints reuse
-			// the stored image, so the transmit is only emitted when requested.
+			// Transmit-once (keyed by id). Repaints reuse the stored image, so the
+			// transmit is only emitted when requested.
 			let transmit: string | undefined;
 			if (options.includeTransmit) {
-				const tempFile =
-					graphics.transmissionMedium === "temp-file" && isPngBase64(base64Data)
-						? encodeKittyTempFileTransmit(base64Data, options.imageId)
-						: null;
-				transmit = tempFile ?? encodeKittyTransmit(base64Data, options.imageId);
+				transmit = encodeKittyTransmit(base64Data, options.imageId);
 			}
 			// Unicode placeholders render the image as real text cells (which survive
 			// horizontal slicing, reflow and overlaps) instead of a cursor-positioned
diff --git a/packages/tui/src/terminal.ts b/packages/tui/src/terminal.ts
index c15ac579b..2069f6370 100644
--- a/packages/tui/src/terminal.ts
+++ b/packages/tui/src/terminal.ts
@@ -2,9 +2,8 @@ import { dlopen, FFIType, ptr } from "bun:ffi";
 import * as fs from "node:fs";
 import { $env, isBunTestRuntime, logger } from "@oh-my-pi/pi-utils";
 import { setKittyProtocolActive } from "./keys";
-import { encodeKittyTempFileProbe, getKittyGraphics, kittyTempFileAllowed, setKittyGraphics } from "./kitty-graphics";
 import { StdinBuffer } from "./stdin-buffer";
-import { ImageProtocol, NotifyProtocol, setCellDimensions, setOsc99Supported, TERMINAL } from "./terminal-capabilities";
+import { NotifyProtocol, setCellDimensions, setOsc99Supported, TERMINAL } from "./terminal-capabilities";
 
 const TERMINAL_PROGRESS_KEEPALIVE_MS = 1000;
 const TERMINAL_PROGRESS_ACTIVE_SEQUENCE = "\x1b]9;4;3\x07";
@@ -40,6 +39,7 @@ export function emergencyTerminalRestore(): void {
 					"\x1b[?2004l" + // Disable bracketed paste
 					"\x1b[?2031l" + // Disable Mode 2031 appearance notifications
 					"\x1b[?2048l" + // Disable in-band resize notifications
+					"\x1b[?5522l" + // Disable enhanced paste notifications
 					"\x1b[<u" + // Pop kitty keyboard protocol
 					"\x1b[>4;0m" + // Disable modifyOtherKeys fallback
 					"\x1b[?25h", // Show cursor
@@ -129,6 +129,12 @@ export interface Terminal {
 	 */
 	isNativeViewportAtBottom?(): boolean | undefined;
 
+	/**
+	 * Override the global terminal-profile ED3 risk decision for custom/test
+	 * terminals. `undefined` falls back to the resolved `TERMINAL` profile.
+	 */
+	hasEagerEraseScrollbackRisk?(): boolean | undefined;
+
 	/**
 	 * Register a callback for terminal appearance (dark/light) changes.
 	 * Detection uses OSC 11 background color query with Mode 2031 as a change trigger.
@@ -153,11 +159,9 @@ type Da1SentinelOwner =
 	| { kind: "keyboard" }
 	| { kind: "osc11" }
 	| { kind: "privateMode"; mode: number }
-	| { kind: "kittyGraphicsProbe"; id: number }
 	| { kind: "osc99Probe"; id: string };
 
 let nextOsc99ProbeId = 1;
-let nextKittyGraphicsProbeId = 1;
 
 function parseOsc99KeyValues(section: string): Map<string, string> {
 	const values = new Map<string, string>();
@@ -192,8 +196,6 @@ export class ProcessTerminal implements Terminal {
 	#osc99PendingId: string | undefined;
 	#osc99ResponseBuffer = "";
 	#osc99Capabilities = new Map<string, string>();
-	#kittyGraphicsPendingId: number | undefined;
-	#kittyGraphicsProbeCleanup: (() => void) | undefined;
 	#privateCsiResponseBuffer = "";
 	#da1SentinelOwners: Da1SentinelOwner[] = [];
 	/** Resolved DECRQM support per private mode (mode → supported). */
@@ -279,11 +281,6 @@ export class ProcessTerminal implements Terminal {
 		// without leaking probe bytes to application input.
 		this.#queryOsc99Support();
 
-		// Probe Kitty temp-file (`t=t`) graphics transmission support. Rides the
-		// same DA1 sentinel FIFO; promotes the transmission medium to temp-file
-		// only on an explicit `OK`, so unsupported terminals stay on direct base64.
-		this.#queryKittyGraphicsTempFile();
-
 		// Subscribe to Mode 2031 appearance change notifications.
 		// When the terminal reports a change, we re-query OSC 11 to get the
 		// actual background color (following Neovim convention) with 100ms debounce.
@@ -502,9 +499,6 @@ export class ProcessTerminal implements Terminal {
 						this.#resolveOsc99Support(owner.id, false);
 						break;
 					}
-					case "kittyGraphicsProbe":
-						this.#resolveKittyGraphicsTempFile(owner.id, false);
-						break;
 				}
 				return;
 			}
@@ -570,21 +564,6 @@ export class ProcessTerminal implements Terminal {
 				}
 			}
 
-			// Kitty graphics temp-file probe reply: ESC _ G i=<id>;OK ESC \. The
-			// owner remains in the FIFO and is drained by its DA1 sentinel (no-op
-			// once resolved here).
-			if (this.#kittyGraphicsPendingId !== undefined && sequence.startsWith("\x1b_G")) {
-				const graphicsMatch = sequence.match(/^\x1b_G([^;]*);([\s\S]*?)\x1b\\$/u);
-				if (graphicsMatch) {
-					const idMatch = graphicsMatch[1]!.match(/(?:^|,)i=(\d+)(?:,|$)/);
-					const replyId = idMatch ? parseInt(idMatch[1]!, 10) : undefined;
-					if (replyId === this.#kittyGraphicsPendingId) {
-						this.#resolveKittyGraphicsTempFile(replyId, graphicsMatch[2]!.trim() === "OK");
-						return;
-					}
-				}
-			}
-
 			// Mode 2031 change notification: re-query OSC 11 with 100ms debounce
 			// (Neovim convention — coalesces rapid notifications during transitions)
 			const appearanceMatch = sequence.match(appearanceDsrPattern);
@@ -680,37 +659,6 @@ export class ProcessTerminal implements Terminal {
 		setOsc99Supported(supported);
 	}
 
-	#shouldQueryKittyGraphicsTempFile(): boolean {
-		if (TERMINAL.imageProtocol !== ImageProtocol.Kitty) return false;
-		// Honor the remote/explicit env gate, and skip when temp-file is already on.
-		if (!kittyTempFileAllowed() || getKittyGraphics().transmissionMedium === "temp-file") return false;
-		return !isBunTestRuntime() || $env.PI_TUI_KITTY_GRAPHICS_PROBE === "1";
-	}
-
-	#queryKittyGraphicsTempFile(): void {
-		this.#clearKittyGraphicsProbe();
-		if (this.#dead || !this.#shouldQueryKittyGraphicsTempFile()) return;
-
-		const id = nextKittyGraphicsProbeId++;
-		const probe = encodeKittyTempFileProbe(id);
-		if (!probe) return;
-		this.#kittyGraphicsPendingId = id;
-		this.#kittyGraphicsProbeCleanup = probe.cleanup;
-		this.#da1SentinelOwners.push({ kind: "kittyGraphicsProbe", id });
-		this.#safeWrite(`${probe.sequence}\x1b[c`);
-	}
-
-	#resolveKittyGraphicsTempFile(id: number, supported: boolean): void {
-		if (this.#kittyGraphicsPendingId !== id) return;
-		if (supported) setKittyGraphics({ transmissionMedium: "temp-file" });
-		this.#clearKittyGraphicsProbe();
-	}
-
-	#clearKittyGraphicsProbe(): void {
-		this.#kittyGraphicsPendingId = undefined;
-		this.#kittyGraphicsProbeCleanup?.();
-		this.#kittyGraphicsProbeCleanup = undefined;
-	}
 	/**
 	 * Parse an OSC 11 background color response and compute BT.601 luminance.
 	 * Handles 1-, 2-, 3-, and 4-digit XParseColor hex components.
@@ -937,6 +885,7 @@ export class ProcessTerminal implements Terminal {
 
 		// Disable bracketed paste mode
 		this.#safeWrite("\x1b[?2004l");
+		this.#safeWrite("\x1b[?5522l");
 
 		// Disable Mode 2031 appearance change notifications
 		this.#safeWrite("\x1b[?2031l");
@@ -959,7 +908,6 @@ export class ProcessTerminal implements Terminal {
 		this.#osc99ResponseBuffer = "";
 		this.#osc99Capabilities.clear();
 		setOsc99Supported(false);
-		this.#clearKittyGraphicsProbe();
 		this.#privateCsiResponseBuffer = "";
 		this.#da1SentinelOwners.length = 0;
 		this.#privateModeCallbacks = [];
diff --git a/packages/tui/src/tui.ts b/packages/tui/src/tui.ts
index 0e0a1835d..e71ee6818 100644
--- a/packages/tui/src/tui.ts
+++ b/packages/tui/src/tui.ts
@@ -1,5 +1,14 @@
 /**
- * Minimal TUI implementation with differential rendering
+ * Minimal TUI implementation with differential rendering.
+ *
+ * Before changing the render planner, native-scrollback bookkeeping, capability
+ * detection, or width math, read `docs/tui-core-renderer.md`: it documents the
+ * failure modes (yank / corruption / flash / width crashes) and the invariants
+ * this engine must not violate. The short version: the renderer cannot observe
+ * the terminal's scroll position on most hosts, so ED3 (`CSI 3 J`) is confined
+ * to the destructive `clearScrollback` path, an unobservable viewport probe is
+ * never trusted for passive streaming, and the hot path clamps over-wide lines
+ * instead of throwing.
  */
 import * as fs from "node:fs";
 import * as path from "node:path";
@@ -15,6 +24,7 @@ import {
 	setCellDimensions,
 	setTerminalImageProtocol,
 	shouldEnableSynchronizedOutputByDefault,
+	synchronizedOutputUserOverride,
 	TERMINAL,
 } from "./terminal-capabilities";
 import {
@@ -29,12 +39,14 @@ import {
 
 const SEGMENT_RESET = "\x1b[0m";
 /**
- * Per-line terminator written at the end of every non-image line. Closes both
+ * Per-line terminator written after every non-image content row. It closes both
  * SGR state and any in-flight OSC 8 hyperlink so styles/links cannot bleed
- * across lines in scrollback. Applied by {@link TUI.#applyLineResets} before
- * diffing so `#previousLines` mirrors what was actually written.
+ * across lines in scrollback. Kept out of the diff/width cache because reset
+ * bytes are deterministic write framing, not content.
  */
 const LINE_TERMINATOR = "\x1b[0m\x1b]8;;\x07";
+const ERASE_LINE = "\x1b[2K";
+const ERASE_TO_END_OF_LINE = "\x1b[K";
 // Hide the hardware cursor before each paint/move write. Ghostty-style bar
 // cursors can otherwise leave visual afterimages while the TUI repaints the
 // row under a visible cursor. Paint writes also disable terminal autowrap:
@@ -56,9 +68,15 @@ const CURSOR_BEGIN = `${HIDE_CURSOR}${SYNC_OUTPUT_BEGIN}`;
 const CURSOR_BEGIN_NO_SYNC = HIDE_CURSOR;
 const CURSOR_END = SYNC_OUTPUT_END;
 const CURSOR_END_NO_SYNC = "";
+// Mouse reporting (normal click tracking + SGR extended coordinates), enabled
+// only for the lifetime of a fullscreen overlay so the rest of the app keeps the
+// terminal's native text selection.
+const MOUSE_TRACKING_ON = "\x1b[?1000h\x1b[?1006h";
+const MOUSE_TRACKING_OFF = "\x1b[?1006l\x1b[?1000l";
 
 type InputListenerResult = { consume?: boolean; data?: string } | undefined;
 type InputListener = (data: string) => InputListenerResult;
+type StartListener = () => void;
 
 export interface RenderTimer {
 	cancel(): void;
@@ -74,6 +92,11 @@ export interface TUIOptions {
 	renderScheduler?: RenderScheduler;
 }
 
+export interface TUIStartOptions {
+	/** Clear saved native scrollback before the first paint. */
+	clearScrollback?: boolean;
+}
+
 const DEFAULT_RENDER_SCHEDULER: RenderScheduler = {
 	now: () => performance.now(),
 	scheduleImmediate: callback => {
@@ -112,10 +135,18 @@ export interface Component {
 	wantsKeyRelease?: boolean;
 
 	/**
-	 * Invalidate any cached rendering state.
+	 * Optional hook to invalidate any cached rendering state.
 	 * Called when theme changes or when component needs to re-render from scratch.
 	 */
-	invalidate(): void;
+	invalidate?(): void;
+
+	/**
+	 * Optional teardown. Called when the component is permanently removed from
+	 * the live tree (e.g. a transcript reset). Release timers, intervals, and
+	 * subscriptions here. Must be idempotent. Containers propagate dispose to
+	 * their children; leaf components without resources may omit it.
+	 */
+	dispose?(): void;
 }
 
 /**
@@ -172,15 +203,14 @@ export interface RenderRequestOptions {
 	/** Clear terminal scrollback for intentional transcript replacement. */
 	clearScrollback?: boolean;
 	/**
-	 * Bypass the unknown-Windows-viewport deferral for this render so the
-	 * caller's intentional live UI mutation reaches the terminal even when
-	 * `Terminal#isNativeViewportAtBottom()` cannot answer.
+	 * Allow a transient live-viewport repaint when the terminal cannot report
+	 * whether its native viewport is at the tail.
 	 *
-	 * Use only for renders driven by direct user interaction (autocomplete
-	 * updates, IME, etc.). Any background/offscreen transcript change that
-	 * coalesces into the same frame WILL also bypass the deferral and reach
-	 * native scrollback — that is the trade-off, and the reason ordinary
-	 * `requestRender()` calls must continue to omit this flag.
+	 * This is **not** a settled transcript commit and must not be used for tool
+	 * completion, session replay, or other background/offscreen rewrites. On
+	 * ED3-risk terminals it may deliberately choose a viewport repaint/deferred
+	 * shrink without clearing native scrollback so autocomplete, IME, and focused
+	 * editor chrome stay responsive without yanking a scrolled reader.
 	 */
 	allowUnknownViewportMutation?: boolean;
 }
@@ -286,6 +316,17 @@ export interface OverlayOptions {
 	 * Called each render cycle with current terminal dimensions.
 	 */
 	visible?: (termWidth: number, termHeight: number) => boolean;
+
+	// === Fullscreen ===
+	/**
+	 * Borrow the terminal's alternate screen buffer for this overlay's lifetime
+	 * (vim/less idiom). While the topmost visible overlay sets this, the engine
+	 * paints only the modal on the alt screen and emits no ED3 / scrollback
+	 * bytes, so the transcript on the normal screen stays untouched and is not
+	 * scrollable behind the modal. Defaults off — all other overlays are
+	 * unchanged and still draw over the transcript on the normal screen.
+	 */
+	fullscreen?: boolean;
 }
 
 /**
@@ -327,11 +368,23 @@ export class Container implements Component {
 		}
 	}
 
+	/**
+	 * Propagate teardown to children. Call when the container's children are
+	 * being permanently discarded (not when they are detached for reuse — use
+	 * {@link clear} for that). Idempotent per child via each child's own dispose.
+	 */
+	dispose(): void {
+		for (const child of this.children) {
+			child.dispose?.();
+		}
+	}
+
 	render(width: number): string[] {
 		width = Math.max(1, width);
 		const lines: string[] = [];
 		for (const child of this.children) {
-			lines.push(...child.render(width));
+			const childLines = child.render(width);
+			for (let i = 0; i < childLines.length; i++) lines.push(childLines[i]);
 		}
 		return lines;
 	}
@@ -357,6 +410,10 @@ export class Container implements Component {
  * - `deferredShrink`: pure content shrink would re-expose rows already in
  *   native history. Keep row indices stable with blank tail padding, repaint
  *   only the viewport, and defer the real shorter replay to a checkpoint.
+ * - `deferredTailRepaint`: a deferred history mutation also changed the active
+ *   grid's bottom row; repaint only that row relative to the tracked hardware
+ *   cursor so a bottom-anchored spinner can advance without rewriting rows that
+ *   a slightly-scrolled reader can still see.
  * - `deferredMutation`: a row-inserting edit would reindex native scrollback
  *   while the user is scrolled. Defer all bytes until a safe rebuild checkpoint.
  * - `shrink`: trailing rows were dropped — clear extras inline.
@@ -364,17 +421,24 @@ export class Container implements Component {
  */
 type RenderIntent =
 	| { kind: "noop" }
-	| { kind: "initial" }
+	| { kind: "initial"; clearScrollback: boolean }
 	| { kind: "sessionReplace" }
 	| { kind: "historyRebuild" }
 	| { kind: "overlayRebuild" }
 	| { kind: "liveRegionPinned"; appendFrom: number; appendTo: number; renderViewportTop: number }
 	| { kind: "viewportRepaint"; appendFrom?: number }
 	| { kind: "deferredShrink"; paddedLength: number }
+	| { kind: "deferredTailRepaint"; row: number; line: string }
 	| { kind: "deferredMutation" }
 	| { kind: "shrink" }
 	| { kind: "diff"; firstChanged: number; lastChanged: number; appendedLines: boolean };
 
+interface PreparedLine {
+	raw: string;
+	width: number;
+	line: string;
+}
+
 /**
  * TUI - Main class for managing terminal UI with differential rendering
  */
@@ -385,6 +449,7 @@ export class TUI extends Container {
 	#previousHeight = 0;
 	#focusedComponent: Component | null = null;
 	#inputListeners = new Set<InputListener>();
+	#startListeners = new Set<StartListener>();
 
 	/** Global callback for debug key (Shift+Ctrl+D). Called before input is forwarded to focused component. */
 	onDebug?: () => void;
@@ -392,7 +457,7 @@ export class TUI extends Container {
 	#renderTimer: RenderTimer | undefined;
 	#renderScheduler: RenderScheduler;
 	#lastRenderAt = 0;
-	static readonly #MIN_RENDER_INTERVAL_MS = 16;
+	static readonly #MIN_RENDER_INTERVAL_MS = 1000 / 30;
 	#cursorRow = 0; // Logical cursor row (end of rendered content)
 	#hardwareCursorRow = 0; // Actual terminal cursor row (may differ due to IME positioning)
 	#viewportTopRow = 0; // Content row currently mapped to screen row 0
@@ -421,6 +486,7 @@ export class TUI extends Container {
 	#nativeScrollbackLiveRegionStart: number | undefined;
 	#nativeScrollbackCommitSafeEnd: number | undefined;
 	#nativeScrollbackDirty = false;
+	#deferredTailLine: string | undefined;
 	// Highest `#maxLinesRendered` reached during a foreground tool turn while
 	// intermediate frames were prevented from committing to terminal scrollback.
 	// Used after the tool finishes to push the settled content into scrollback
@@ -455,6 +521,21 @@ export class TUI extends Container {
 	#resizeEventPending = false;
 	#stopped = false;
 
+	// Transient alternate-screen state for a fullscreen overlay. While active, the
+	// engine paints only the modal on the alt buffer and leaves every
+	// normal-screen accounting field (#previousLines, #viewportTopRow, …)
+	// untouched, so exiting reconciles cleanly against the terminal-restored
+	// normal screen. #altPreviousLines is the last alt frame, for repaint-skip.
+	#altActive = false;
+	#altPreviousLines: string[] = [];
+	#altEnterWidth = 0;
+	#altEnterHeight = 0;
+
+	// Last-frame line preparation cache. Entries store normalized, width-fitted
+	// content rows without the per-line terminal terminator; terminators are
+	// appended only at write time so width checks stay on content, not reset bytes.
+	#preparedLineCache: PreparedLine[] = [];
+
 	// Overlay stack for modal components rendered on top of base content
 	overlayStack: {
 		component: Component;
@@ -492,7 +573,7 @@ export class TUI extends Container {
 					this.#nativeScrollbackCommitSafeEnd = offset + boundedEnd;
 				}
 			}
-			lines.push(...childLines);
+			for (let i = 0; i < childLines.length; i++) lines.push(childLines[i]);
 		}
 		return lines;
 	}
@@ -550,12 +631,18 @@ export class TUI extends Container {
 
 	/**
 	 * Whether DEC 2026 synchronized-output wrappers are currently emitted around
-	 * paints. Starts from conservative terminal/env detection and is force-disabled
-	 * at runtime if the terminal reports mode 2026 unsupported via DECRQM.
+	 * paints. Starts from conservative terminal/env detection and is reconciled at
+	 * runtime against the terminal's DECRQM mode-2026 report — enabled on a
+	 * positive report, disabled on a negative one.
 	 */
 	get synchronizedOutput(): boolean {
 		return this.#synchronizedOutputEnabled;
 	}
+	#deccaraFillsEnabled(): boolean {
+		// DECCARA fill rectangles arrive after shortened row text; synchronized
+		// output hides that intermediate default-background state from users.
+		return TERMINAL.deccara && this.#synchronizedOutputEnabled;
+	}
 
 	/**
 	 * When enabled, live render frames rebuild native scrollback on offscreen and
@@ -569,8 +656,8 @@ export class TUI extends Container {
 	 * (the viewport is never observable there and ConPTY hosts erase host
 	 * scrollback on ED3 — #1635/#1746); only the unknown POSIX case is forced to
 	 * rebuild. POSIX hosts known to disturb scrolled readers on xterm ED3
-	 * (`CSI 3 J`, erase saved lines) also defer the eager opt-in; checkpoint and
-	 * direct user-input rebuilds are unaffected.
+	 * (`CSI 3 J`, erase saved lines) also defer the eager opt-in; checkpoint
+	 * rebuilds are unaffected.
 	 *
 	 * Disabling stays active through one already-requested frame: the event batch
 	 * that ends a foreground stream both removes its UI rows (loader/status
@@ -591,7 +678,7 @@ export class TUI extends Container {
 			this.#eagerNativeScrollbackRebuildDisablePending = true;
 			return;
 		}
-		if (process.platform !== "win32" && TERMINAL.eagerEraseScrollbackRisk) {
+		if (this.#hasEagerEraseScrollbackRisk()) {
 			this.#streamingHighWater = 0;
 			this.#markNativeScrollbackDirty();
 		}
@@ -714,28 +801,47 @@ export class TUI extends Container {
 		for (const overlay of this.overlayStack) overlay.component.invalidate?.();
 	}
 
-	start(): void {
+	start(options?: TUIStartOptions): void {
 		this.#stopped = false;
-		// Disable synchronized output if the terminal reports DEC 2026 unsupported
-		// via DECRQM. PI_NO_SYNC_OUTPUT already forces it off at construction, so
-		// only react when the user has not already opted out. Future paints drop
-		// the begin/end markers; the autowrap guards stay (see #1765).
+		// A DECRQM report for mode 2026 is authoritative: enable synchronized
+		// output when the terminal reports support (upgrading conservatively
+		// defaulted-off hosts like zellij/tmux-master/foot) and disable it when
+		// the terminal reports it unsupported. An explicit user opt-out/force
+		// (resolved at construction) still wins, so skip the probe in that case.
 		this.terminal.onPrivateModeReport?.((mode, supported) => {
-			if (mode === 2026 && !supported && !$flag("PI_NO_SYNC_OUTPUT")) {
-				this.#setSynchronizedOutput(false);
-			}
+			if (mode !== 2026) return;
+			if (synchronizedOutputUserOverride() !== null) return;
+			this.#setSynchronizedOutput(supported);
 		});
 		this.terminal.start(
 			data => this.#handleInput(data),
 			() => {
+				// Repaint immediately rather than via the throttled path: a resize must
+				// clear and replay at the fresh geometry before the terminal's reflow
+				// settles into a state a throttled frame would race. Forced render skips
+				// the 30fps coalescing window, matching resetDisplay()'s prompt repaint.
 				this.#resizeEventPending = true;
-				this.requestRender();
+				this.requestRender(true);
 			},
 		);
+		for (const listener of this.#startListeners) {
+			try {
+				listener();
+			} catch {
+				// Startup listeners are feature hooks; one broken hook must not prevent rendering.
+			}
+		}
 		this.terminal.hideCursor();
 		this.#querySixelSupport();
 		this.#queryCellSize();
-		this.requestRender(true);
+		this.requestRender(true, { clearScrollback: options?.clearScrollback === true });
+	}
+
+	addStartListener(listener: StartListener): () => void {
+		this.#startListeners.add(listener);
+		return () => {
+			this.#startListeners.delete(listener);
+		};
 	}
 
 	addInputListener(listener: InputListener): () => void {
@@ -886,8 +992,8 @@ export class TUI extends Container {
 
 	/**
 	 * Toggle synchronized-output (DEC 2026) wrappers on paint/cursor writes and
-	 * recompute the cached begin/end sequences. Honors a DECRQM report that the
-	 * terminal does not support 2026 (#1765 covers the static env opt-out).
+	 * recompute the cached begin/end sequences. Driven by the terminal's DECRQM
+	 * mode-2026 report (#1765 covers the static env opt-out).
 	 */
 	#setSynchronizedOutput(enabled: boolean): void {
 		if (this.#synchronizedOutputEnabled === enabled) return;
@@ -899,6 +1005,13 @@ export class TUI extends Container {
 	}
 
 	stop(): void {
+		// Leave the alt buffer first so the teardown cursor math below runs against
+		// the restored normal screen (which #previousLines still describes).
+		if (this.#altActive) {
+			this.terminal.write(`${MOUSE_TRACKING_OFF}\x1b[?1049l`);
+			this.#altActive = false;
+			this.#altPreviousLines = [];
+		}
 		if (TERMINAL.imageProtocol === ImageProtocol.Kitty) {
 			for (const id of this.#imageBudget.takeAllTransmittedIds()) {
 				this.terminal.write(encodeKittyDeleteImage(id));
@@ -949,21 +1062,52 @@ export class TUI extends Container {
 			return false;
 		}
 		const nativeViewportAtBottom = this.#readNativeViewportAtBottom();
-		if (!this.#canReplayNativeScrollbackAtCheckpoint(nativeViewportAtBottom)) {
-			return false;
-		}
-		this.#prepareForcedRender(true, false);
+		// The checkpoint fires at a prompt submit — a bottom-pinning user action. On a
+		// genuine local terminal the submit keystroke scrolls the host to its tail, so
+		// an unprobeable viewport is safely at-bottom and the ED3 replay will not yank
+		// a scrolled reader (the same explicit-user-action reasoning the resize rebuild
+		// uses). Hosts whose scrollback a keystroke does not move — Windows
+		// console/Terminal, SSH, multiplexers, unknown profiles — stay gated on a
+		// positive at-tail probe (#1610/#1682/#1746); a known-scrolled viewport always
+		// defers regardless of terminal.
+		if (nativeViewportAtBottom === false) return false;
+		if (nativeViewportAtBottom === undefined && !TERMINAL.submitPinsViewportToTail) return false;
+		this.#prepareForcedRender(true);
 		this.#renderRequested = false;
 		this.#lastRenderAt = this.#renderScheduler.now();
 		this.#doRender();
 		return true;
 	}
 
+	/**
+	 * Force an immediate full replay of the current frame, including native
+	 * scrollback. This is the keyboard-accessible equivalent of the resize reset:
+	 * no queued diff frame or terminal scrollback probe can downgrade it to a
+	 * viewport-only repaint.
+	 *
+	 * Invalidates every component first so the replay reflects current state. A
+	 * geometry-driven reset thaws frozen scrollback snapshots implicitly (the new
+	 * width misses every cached snapshot), but a same-width reset would otherwise
+	 * replay stale snapshots — leaving host-frozen blocks (e.g. a transcript whose
+	 * committed rows are immutable on ED3-risk terminals) showing pre-mutation
+	 * content. Invalidation is the generic signal those containers use to retire
+	 * their snapshots, which is exactly what a user-driven display reset wants.
+	 */
+	resetDisplay(): void {
+		if (this.#stopped) return;
+		this.invalidate();
+		this.#prepareForcedRender(!isMultiplexerSession());
+		this.#resizeEventPending = true;
+		this.#renderRequested = false;
+		this.#lastRenderAt = this.#renderScheduler.now();
+		this.#doRender();
+	}
+
 	requestRender(force = false, options?: RenderRequestOptions): void {
 		const allowUnknownViewportMutation = options?.allowUnknownViewportMutation === true;
 		this.#allowUnknownViewportMutationOnNextRender ||= allowUnknownViewportMutation;
 		if (force) {
-			this.#prepareForcedRender(options?.clearScrollback === true, allowUnknownViewportMutation);
+			this.#prepareForcedRender(options?.clearScrollback === true);
 			this.#renderRequested = true;
 			this.#renderScheduler.scheduleImmediate(() => {
 				if (this.#stopped || !this.#renderRequested) {
@@ -980,7 +1124,7 @@ export class TUI extends Container {
 		this.#renderScheduler.scheduleImmediate(() => this.#scheduleRender());
 	}
 
-	#prepareForcedRender(clearScrollback: boolean, _allowUnknownViewportMutation: boolean): void {
+	#prepareForcedRender(clearScrollback: boolean): void {
 		const geometryChanged =
 			(this.#previousWidth > 0 && this.#previousWidth !== this.terminal.columns) ||
 			(this.#previousHeight > 0 && this.#previousHeight !== this.terminal.rows);
@@ -1395,24 +1539,9 @@ export class TUI extends Container {
 		return cursor;
 	}
 
-	/**
-	 * Append the per-line terminator ({@link LINE_TERMINATOR}) to every
-	 * non-image line and normalize for terminal rendering. Mutates the input
-	 * array in place so downstream diffing/storage sees exactly the bytes
-	 * written to the terminal — without this, the diff cache disagrees with
-	 * emitted output and OSC 8 hyperlink state can leak across lines.
-	 */
-	#applyLineResets(lines: string[]): string[] {
-		for (let i = 0; i < lines.length; i++) {
-			const line = lines[i];
-			if (TERMINAL.isImageLine(line)) continue;
-			const normalized = normalizeTerminalOutput(line);
-			// Only close OSC 8 hyperlinks when the line actually opened one;
-			// emitting `\x1b]8;;\x07` on every line just feeds the terminal's OSC
-			// parser for no reason (measurable cost in xterm.js parse loop).
-			lines[i] = normalized + (normalized.includes("\x1b]8;") ? LINE_TERMINATOR : SEGMENT_RESET);
-		}
-		return lines;
+	#terminalLine(line: string): string {
+		if (TERMINAL.isImageLine(line)) return line;
+		return line + (line.includes("\x1b]8;") ? LINE_TERMINATOR : SEGMENT_RESET);
 	}
 
 	/**
@@ -1425,6 +1554,36 @@ export class TUI extends Container {
 		const width = this.terminal.columns;
 		const height = this.terminal.rows;
 
+		// Fullscreen alt-screen short-circuit. While the topmost visible overlay
+		// requests it, borrow the terminal's alternate buffer (saved/restored by
+		// the terminal around 1049h/1049l) and paint only the modal there. This
+		// touches no normal-screen accounting field, so the transcript on the
+		// normal screen stays untouched and unscrollable behind the modal, and
+		// exiting reconciles cleanly against the terminal-restored screen.
+		const wantAlt = this.#wantsAltScreen();
+		if (wantAlt && !this.#altActive) {
+			this.terminal.write(`\x1b[?1049h${MOUSE_TRACKING_ON}`);
+			this.terminal.hideCursor();
+			this.#altActive = true;
+			this.#altPreviousLines = [];
+			this.#altEnterWidth = width;
+			this.#altEnterHeight = height;
+		} else if (!wantAlt && this.#altActive) {
+			this.terminal.write(`${MOUSE_TRACKING_OFF}\x1b[?1049l`);
+			this.#altActive = false;
+			this.#altPreviousLines = [];
+			// A resize while on the alt buffer reflowed the terminal's saved normal
+			// screen; it no longer matches #previousLines, so force the geometry
+			// rebuild path instead of a stale diff.
+			if (width !== this.#altEnterWidth || height !== this.#altEnterHeight) {
+				this.#resizeEventPending = true;
+			}
+		}
+		if (this.#altActive) {
+			this.#renderAltFrame(width, height);
+			return;
+		}
+
 		// 1. Compose the frame. Bracket the transcript render so the image budget
 		// observes every inline image in display order (overlays carry none).
 		this.#imageBudget.beginPass();
@@ -1445,11 +1604,7 @@ export class TUI extends Container {
 		const overlayVisibilityReduced = this.#overlayVisibilityReduced(visibleOverlayComponents);
 		let lines = visibleOverlayComponents.length > 0 ? this.#compositeOverlays(baseLines, width, height) : baseLines;
 		const cursorPos = this.#extractCursorPosition(lines, height);
-		lines = this.#fitLinesToWidth(this.#applyLineResets(lines), width);
-		if (lines !== baseLines) {
-			this.#extractCursorPosition(baseLines, height);
-			baseLines = this.#fitLinesToWidth(this.#applyLineResets(baseLines), width);
-		}
+		lines = this.#prepareLines(lines, width, true);
 
 		// 2. Capture transition + pre-render state before any emitter runs.
 		const prevViewportTop = this.#viewportTopRow;
@@ -1463,7 +1618,7 @@ export class TUI extends Container {
 		const heightChanged =
 			(this.#previousHeight > 0 && this.#previousHeight !== height) ||
 			(resizeEventOccurred && this.#previousHeight > 0);
-		const eagerEraseScrollbackRisk = process.platform !== "win32" && TERMINAL.eagerEraseScrollbackRisk;
+		const eagerEraseScrollbackRisk = this.#hasEagerEraseScrollbackRisk();
 		const eagerRebuildAllowed = this.#eagerNativeScrollbackRebuild && !eagerEraseScrollbackRisk;
 		const explicitViewportMutation = this.#allowUnknownViewportMutationOnNextRender;
 		const allowUnknownViewportMutation = explicitViewportMutation || eagerRebuildAllowed;
@@ -1528,6 +1683,7 @@ export class TUI extends Container {
 			} else if (
 				!explicitReconcile &&
 				nativeViewportAtBottom !== true &&
+				!isMultiplexerSession() &&
 				(intent.kind === "sessionReplace" ||
 					intent.kind === "historyRebuild" ||
 					intent.kind === "overlayRebuild" ||
@@ -1535,9 +1691,20 @@ export class TUI extends Container {
 			) {
 				// Cap the frame to the viewport and keep scrollback dirty: transient
 				// rows never enter history, and the checkpoint reconciles later.
+				// Multiplexers (tmux/screen/zellij) are excluded: their checkpoint
+				// reconcile is a no-op (pane history cannot be erased), so any rows
+				// dropped here are dropped forever. Pane history is append-only
+				// anyway, so a normal diff/append `\r\n` commit is exactly what the
+				// multiplexer needs — and the `liveRegionPinned` planner above
+				// keeps the actively-mutating live tail out of pane history while
+				// committing only the sealed prefix (issue #1974).
+				// Do not lower #scrollbackHighWater here. The viewport repaint below
+				// avoids committing new transient rows, but rows committed by earlier
+				// full/diff paints are still physically present in native scrollback and
+				// must remain in the shrink/de-dup accounting until an ED3 checkpoint
+				// clears them.
 				this.#markNativeScrollbackDirty();
 				this.#streamingHighWater = Math.max(this.#streamingHighWater, lines.length);
-				this.#scrollbackHighWater = 0;
 				lines = lines.slice(-height);
 				intent = { kind: "viewportRepaint" };
 			} else {
@@ -1575,6 +1742,7 @@ export class TUI extends Container {
 				if (
 					this.#eagerNativeScrollbackRebuild &&
 					eagerEraseScrollbackRisk &&
+					!intent.clearScrollback &&
 					!allowUnknownViewportMutation &&
 					liveRegionStart !== undefined &&
 					liveRegionStart < lines.length &&
@@ -1590,8 +1758,12 @@ export class TUI extends Container {
 						this.#nativeScrollbackCommitSafeEnd,
 					);
 				} else {
-					this.#emitFullPaint(lines, width, height, cursorPos, { clearViewport: true, clearScrollback: false });
+					this.#emitFullPaint(lines, width, height, cursorPos, {
+						clearViewport: true,
+						clearScrollback: intent.clearScrollback && !isMultiplexerSession(),
+					});
 				}
+				this.#clearScrollbackOnNextRender = false;
 				this.#hasEverRendered = true;
 				return;
 			}
@@ -1602,6 +1774,7 @@ export class TUI extends Container {
 					clearViewport: true,
 					clearScrollback: !isMultiplexerSession(),
 				});
+				this.#hasEverRendered = true;
 				return;
 			case "historyRebuild":
 				this.#clearNativeScrollbackDirty();
@@ -1612,6 +1785,8 @@ export class TUI extends Container {
 				return;
 			case "overlayRebuild":
 				this.#clearNativeScrollbackDirty();
+				this.#extractCursorPosition(baseLines, height);
+				baseLines = this.#prepareLines(baseLines, width, false);
 				this.#emitFullPaint(baseLines, width, height, null, {
 					clearViewport: true,
 					clearScrollback: !isMultiplexerSession(),
@@ -1633,10 +1808,20 @@ export class TUI extends Container {
 				return;
 			case "viewportRepaint":
 				if (intent.appendFrom !== undefined) {
-					this.#emitAppendTail(lines, intent.appendFrom, height, width, prevViewportTop, prevHardwareCursorRow);
+					this.#emitAppendTail(lines, intent.appendFrom, height, prevViewportTop, prevHardwareCursorRow);
 				}
 				this.#emitViewportRepaint(lines, width, height, cursorPos);
 				return;
+			case "deferredTailRepaint":
+				this.#emitDeferredTailRepaint(
+					intent.line,
+					width,
+					height,
+					intent.row,
+					prevViewportTop,
+					prevHardwareCursorRow,
+				);
+				return;
 			case "deferredMutation":
 				return;
 			case "deferredShrink":
@@ -1686,16 +1871,16 @@ export class TUI extends Container {
 		liveRegionStart: number | undefined,
 		commitSafeEnd: number | undefined,
 	): RenderIntent {
-		// Initial paint after start(): scrollback must keep its prior shell
-		// content, but the viewport must be cleared so stale rows do not bleed
-		// into the new UI.
-		if (!this.#hasEverRendered) return { kind: "initial" };
+		// Initial paint after start(): preserve prior shell scrollback by default,
+		// but honor callers that are replacing terminal history before any frame is
+		// committed. This keeps the first visible commit clean instead of appending
+		// a tall transcript once and wiping on the next render.
+		if (!this.#hasEverRendered) return { kind: "initial", clearScrollback: this.#clearScrollbackOnNextRender };
 
-		// Caller opted into a scrollback wipe via requestRender(true, { clearScrollback: true }).
 		if (this.#clearScrollbackOnNextRender) return { kind: "sessionReplace" };
 
 		const forceViewportRepaint = this.#forceViewportRepaintOnNextRender;
-		const eagerEraseScrollbackRisk = process.platform !== "win32" && TERMINAL.eagerEraseScrollbackRisk;
+		const eagerEraseScrollbackRisk = this.#hasEagerEraseScrollbackRisk();
 		if (overlayVisibilityReduced && !isMultiplexerSession()) {
 			return hasVisibleOverlay ? { kind: "overlayRebuild" } : { kind: "historyRebuild" };
 		}
@@ -1718,6 +1903,11 @@ export class TUI extends Container {
 			return hasVisibleOverlay ? { kind: "overlayRebuild" } : { kind: "historyRebuild" };
 		}
 
+		// Same dirty-scrollback opt-in policy as the non-overlay branch below: an
+		// ED3-risk macOS/POSIX terminal with an unobservable viewport ignores
+		// focused-input unknown opt-ins, so overlay selector Up/Down moves do not
+		// become ED3 clears plus full transcript replays. Non-ED3-risk POSIX still
+		// honors direct-input/IME/autocomplete opt-ins.
 		if (hasVisibleOverlay) {
 			const nativeViewportAtBottom = this.#readNativeViewportAtBottom();
 			// Multiplexer panes never get a destructive scrollback clear
@@ -1725,10 +1915,11 @@ export class TUI extends Container {
 			// "rebuild" would only append a full duplicate copy of the transcript
 			// to pane history on every dirty frame. Keep repainting the viewport
 			// and leave reconciliation to explicit checkpoints.
+			const allowDirtyUnknownViewportMutation = allowUnknownViewportMutation && !eagerEraseScrollbackRisk;
 			if (
 				this.#nativeScrollbackDirty &&
 				!isMultiplexerSession() &&
-				this.#canRebuildNativeScrollbackLive(nativeViewportAtBottom, allowUnknownViewportMutation)
+				this.#canRebuildNativeScrollbackLive(nativeViewportAtBottom, allowDirtyUnknownViewportMutation)
 			) {
 				return { kind: "overlayRebuild" };
 			}
@@ -1759,12 +1950,19 @@ export class TUI extends Container {
 			this.#streamingHighWater = 0;
 		}
 
-		if (
-			this.#nativeScrollbackDirty &&
-			!isMultiplexerSession() &&
-			this.#canRebuildNativeScrollbackLive(this.#readNativeViewportAtBottom(), allowUnknownViewportMutation)
-		) {
-			return { kind: "historyRebuild" };
+		if (this.#nativeScrollbackDirty && !isMultiplexerSession()) {
+			// A dirty flag means older native history is stale; it is not required to
+			// make the current focused-input frame correct. On ED3-risk macOS/POSIX
+			// terminals with an unobservable viewport, ignore focused-input unknown
+			// opt-ins so Up/Down selector moves do not become ED3 clears plus full
+			// transcript replays. Non-ED3-risk POSIX terminals keep their safe
+			// direct-input/IME/autocomplete opt-in.
+			const allowDirtyUnknownViewportMutation = allowUnknownViewportMutation && !eagerEraseScrollbackRisk;
+			if (
+				this.#canRebuildNativeScrollbackLive(this.#readNativeViewportAtBottom(), allowDirtyUnknownViewportMutation)
+			) {
+				return { kind: "historyRebuild" };
+			}
 		}
 
 		const diff = this.#diffLines(newLines);
@@ -1802,18 +2000,28 @@ export class TUI extends Container {
 			//
 			const paddedViewportTop = Math.max(0, this.#previousLines.length - height);
 			// ED3-risk terminals with an unobservable viewport cannot safely clear
-			// saved lines. During an active eager streaming turn the user follows the
-			// live tail, so paint the shrink's bottom-anchored viewport in place
-			// whether it still overflows OR now fits — otherwise the UI freezes on
-			// stale rows until the next input even though the frame has a fresh bottom
-			// viewport to show (issues #1682, foreground-stream fidelity on collapse).
-			// Native history stays dirty and reconciles at the next checkpoint. With no
-			// active eager turn the reader may be scrolled; even a padded shrink repaint
-			// can move ED3-risk unknown host scrollback (WSL/Ghostty-style), so defer
+			// saved lines. Direct user-input frames (autocomplete/IME) may still
+			// repaint the live viewport: the user action pins the host to the tail, and
+			// emitting zero bytes leaves stale autocomplete rows on screen until a later
+			// checkpoint. When the changed rows are at or below the previous viewport
+			// top, keep the old bottom anchor by padding the frame to its previous
+			// length; that clears stale popup rows without re-exposing rows already
+			// committed to native history. If an offscreen edit shifted rows above the
+			// viewport, padding would repaint the wrong seam, so use a viewport repaint
+			// for liveness and keep history dirty. Active eager streaming also uses a
+			// viewport repaint so the live tail keeps moving. With neither direct input
+			// nor active eager streaming, the reader may be scrolled, so defer
 			// completely rather than repainting over their history.
 			if (nativeViewportAtBottom === undefined && eagerEraseScrollbackRisk) {
 				this.#markNativeScrollbackDirty();
-				return this.#eagerNativeScrollbackRebuild ? { kind: "viewportRepaint" } : { kind: "deferredMutation" };
+				if (allowUnknownViewportMutation) {
+					return diff.firstChanged < prevViewportTop
+						? { kind: "viewportRepaint" }
+						: { kind: "deferredShrink", paddedLength: this.#previousLines.length };
+				}
+				return this.#eagerNativeScrollbackRebuild
+					? { kind: "viewportRepaint" }
+					: this.#planDeferredTailRepaint(newLines, prevViewportTop, height);
 			}
 
 			// Non-ED3-risk POSIX with an unobservable viewport. `deferredShrink` is
@@ -1825,7 +2033,7 @@ export class TUI extends Container {
 			}
 			this.#markNativeScrollbackDirty();
 			if (diff.firstChanged < prevViewportTop) {
-				return { kind: "deferredMutation" };
+				return this.#planDeferredTailRepaint(newLines, prevViewportTop, height);
 			}
 			return { kind: "deferredShrink", paddedLength: this.#previousLines.length };
 		}
@@ -1924,21 +2132,30 @@ export class TUI extends Container {
 		const contentGrew = newLines.length > this.#previousLines.length;
 		const pureAppend = diff.appendedLines && diff.firstChanged === this.#previousLines.length;
 		const structuralMutation = newLines.length !== this.#previousLines.length || diff.firstChanged < prevViewportTop;
-		if (pureAppend && contentGrew && this.#previousLines.length > height && !isMultiplexerSession()) {
+		if (pureAppend && contentGrew && this.#previousLines.length >= height && !isMultiplexerSession()) {
 			const nativeViewportAtBottom = this.#readNativeViewportAtBottom();
+			if (this.#nativeViewportIsKnownScrolled(nativeViewportAtBottom)) {
+				this.#markNativeScrollbackDirty();
+				return { kind: "deferredMutation" };
+			}
+			if (nativeViewportAtBottom === undefined && allowUnknownViewportMutation) {
+				// Direct input can grow transient live UI (autocomplete/IME/editor
+				// wraps) while the previous frame already touched the viewport bottom.
+				// A diff append would `\r\n`-scroll those transient rows into native
+				// history, and a later popup shrink would duplicate the stable prefix at
+				// the scrollback seam. Repaint the live viewport in place instead; the
+				// dirty checkpoint owns native-history reconciliation.
+				this.#markNativeScrollbackDirty();
+				return { kind: "viewportRepaint" };
+			}
 			if (this.#nativeViewportIsScrolled(nativeViewportAtBottom, allowUnknownViewportMutation)) {
 				this.#markNativeScrollbackDirty();
-				// Confirmed scrolled (probe returned `false`): the reader is parked in
-				// scrollback and writing the live frame is wasted bytes — defer until
-				// the next checkpoint reconciles. Unknown viewport (e.g. native Windows
-				// Terminal where the probe cannot see WT host scrollback) is a
-				// different case: a no-op there freezes the editor on the keystroke
-				// that grows `lines.length` past the viewport (the wrap keystroke).
-				// Fall through to a non-destructive viewport repaint instead so the
-				// live UI keeps updating without yanking a possibly-scrolled reader.
-				if (this.#nativeViewportIsKnownScrolled(nativeViewportAtBottom)) {
-					return { kind: "deferredMutation" };
-				}
+				// Unknown viewport (e.g. native Windows Terminal where the probe cannot
+				// see WT host scrollback) is a different case: a no-op there freezes the
+				// editor on the keystroke that grows `lines.length` past the viewport
+				// (the wrap keystroke). Fall through to a non-destructive viewport
+				// repaint instead so the live UI keeps updating without yanking a
+				// possibly-scrolled reader.
 				return { kind: "viewportRepaint" };
 			}
 		}
@@ -2022,7 +2239,7 @@ export class TUI extends Container {
 				!cleanTailAppend &&
 				!this.#eagerNativeScrollbackRebuild
 			) {
-				return { kind: "deferredMutation" };
+				return this.#planDeferredTailRepaint(newLines, prevViewportTop, height);
 			}
 			return { kind: "viewportRepaint", appendFrom: cleanTailAppend ? this.#previousLines.length : undefined };
 		}
@@ -2108,6 +2325,11 @@ export class TUI extends Container {
 		this.#nativeScrollbackDirty = false;
 	}
 
+	#hasEagerEraseScrollbackRisk(): boolean {
+		if (process.platform === "win32") return false;
+		return this.terminal.hasEagerEraseScrollbackRisk?.() ?? TERMINAL.eagerEraseScrollbackRisk;
+	}
+
 	#readNativeViewportAtBottom(): boolean | undefined {
 		// A stale positive is destructive: live history rebuilds clear native
 		// scrollback. Require two consecutive at-bottom probes before trusting it.
@@ -2141,15 +2363,14 @@ export class TUI extends Container {
 	 * the native viewport) is safe to emit *during ordinary rendering*. POSIX
 	 * terminals cannot report whether the user has scrolled up
 	 * (`isNativeViewportAtBottom()` is `undefined`), so an unknown position is
-	 * treated as unsafe: defer to a non-destructive viewport repaint and keep
-	 * scrollback dirty until a later render has a positive at-tail proof. A prompt
-	 * submit is no longer treated as proof for unobservable host scrollback.
-	 * this, every offscreen transcript edit while streaming wiped scrollback and
-	 * yanked a scrolled-up reader out of their current context.
-	 * `allowUnknownViewportMutation` (autocomplete/IME) opts directly
-	 * user-driven POSIX frames back into the rebuild. Native Windows and Windows
-	 * Terminal still cannot trust an unknown probe during live rendering — ConPTY
-	 * may be fronting host scrollback we cannot observe — so they keep deferring.
+	 * treated as unsafe by default: defer to a non-destructive viewport repaint and
+	 * keep scrollback dirty until a later checkpoint/positive at-tail proof.
+	 *
+	 * `allowUnknownViewportMutation` is the narrow exception for direct
+	 * input chrome (autocomplete/IME/editor wrapping): those frames may repaint or,
+	 * on non-Windows hosts, rebuild live UI while the user action pins the prompt
+	 * to the tail. Settled transcript commits should not use this flag; they must
+	 * request an explicit clear+replay instead.
 	 */
 	#canRebuildNativeScrollbackLive(
 		nativeViewportAtBottom: boolean | undefined,
@@ -2174,11 +2395,18 @@ export class TUI extends Container {
 			liveRegionStart >= newLines.length ||
 			!this.#eagerNativeScrollbackRebuild ||
 			!eagerEraseScrollbackRisk ||
-			allowUnknownViewportMutation ||
-			isMultiplexerSession()
+			allowUnknownViewportMutation
 		) {
 			return undefined;
 		}
+		// Multiplexers (tmux/screen/zellij) cannot erase pane history with `\x1b[3J`
+		// and cannot answer a viewport-position probe, so the destructive checkpoint
+		// rebuild path is forever unavailable. The pinned emitter is built from the
+		// opposite primitives — relative cursor moves, per-row rewrite/suffix-clear,
+		// and `\r\n` to scroll sealed rows past the viewport bottom — which are exactly
+		// what tmux pane history accepts. Without this commit-as-you-go path, the
+		// streaming cap below clipped every frame to the visible tail and the
+		// scrolled-off head was committed nowhere (issue #1974).
 		if (newLines.length <= height && this.#scrollbackHighWater === 0) return undefined;
 		if (this.#readNativeViewportAtBottom() !== undefined) return undefined;
 
@@ -2209,29 +2437,115 @@ export class TUI extends Container {
 		return { kind: "liveRegionPinned", appendFrom, appendTo, renderViewportTop };
 	}
 
+	#planDeferredTailRepaint(newLines: string[], prevViewportTop: number, height: number): RenderIntent {
+		const row = prevViewportTop + height - 1;
+		if (row < 0 || row >= this.#previousLines.length || newLines.length !== this.#previousLines.length) {
+			return { kind: "deferredMutation" };
+		}
+		const line = newLines[row] ?? "";
+		const previousLine = this.#deferredTailLine ?? this.#previousLines[row] ?? "";
+		if (line === previousLine) {
+			return { kind: "deferredMutation" };
+		}
+		return { kind: "deferredTailRepaint", row, line };
+	}
+
 	#padDeferredShrinkLines(lines: string[], paddedLength: number): string[] {
 		if (lines.length >= paddedLength) return lines;
 		return [...lines, ...new Array<string>(paddedLength - lines.length).fill("")];
 	}
-	/**
-	 * Truncate a line to the visible viewport width. Image lines are left
-	 * alone, narrow lines pass through unchanged. Truncation re-appends the
-	 * per-line terminator so SGR/OSC 8 state does not leak across rows when
-	 * `truncateToWidth` drops the trailing bytes appended by
-	 * {@link #applyLineResets}.
-	 */
-	#fitLinesToWidth(lines: string[], width: number): string[] {
+	#prepareLines(lines: string[], width: number, useCache: boolean): string[] {
+		const prepared: string[] = new Array(lines.length);
+		const previous = useCache ? this.#preparedLineCache : [];
+		const nextCache: PreparedLine[] | undefined = useCache ? new Array(lines.length) : undefined;
 		for (let i = 0; i < lines.length; i++) {
-			lines[i] = this.#fitLineToWidth(lines[i], width);
+			const raw = lines[i]!;
+			const cached = previous[i];
+			if (cached && cached.raw === raw && cached.width === width) {
+				prepared[i] = cached.line;
+				if (nextCache) nextCache[i] = cached;
+				continue;
+			}
+			const entry = this.#prepareLine(raw, width);
+			prepared[i] = entry.line;
+			if (nextCache) nextCache[i] = entry;
 		}
-		return lines;
+		if (nextCache) this.#preparedLineCache = nextCache;
+		return prepared;
 	}
 
-	#fitLineToWidth(line: string, width: number): string {
-		if (TERMINAL.isImageLine(line)) return line;
-		if (visibleWidth(line) <= width) return line;
-		const truncated = truncateToWidth(line, width, Ellipsis.Omit);
-		return truncated + (truncated.includes("\x1b]8;") ? LINE_TERMINATOR : SEGMENT_RESET);
+	#prepareLine(raw: string, width: number): PreparedLine {
+		if (TERMINAL.isImageLine(raw)) {
+			return { raw, width, line: raw };
+		}
+		const normalized = normalizeTerminalOutput(raw);
+		const asciiWidth = this.#ansiAsciiLineWidth(normalized, width);
+		if ((asciiWidth ?? visibleWidth(normalized)) <= width) {
+			return { raw, width, line: normalized };
+		}
+		const line = truncateToWidth(normalized, width, Ellipsis.Omit);
+		return { raw, width, line };
+	}
+
+	#ansiAsciiLineWidth(line: string, maxWidth: number): number | undefined {
+		let col = 0;
+		for (let i = 0; i < line.length; ) {
+			const code = line.charCodeAt(i);
+			if (code === 0x1b) {
+				const next = line.charCodeAt(i + 1);
+				if (next === 0x5b) {
+					let j = i + 2;
+					while (j < line.length) {
+						const final = line.charCodeAt(j);
+						if (final >= 0x40 && final <= 0x7e) break;
+						j++;
+					}
+					if (j >= line.length) return undefined;
+					i = j + 1;
+					continue;
+				}
+				if (next === 0x5d) {
+					// OSC 66 text-sizing spans carry visible payload inside the OSC.
+					// Fall back to visibleWidth() so scaled cells stay exact.
+					if (
+						line.charCodeAt(i + 2) === 0x36 &&
+						line.charCodeAt(i + 3) === 0x36 &&
+						line.charCodeAt(i + 4) === 0x3b
+					) {
+						return undefined;
+					}
+					let j = i + 2;
+					while (j < line.length) {
+						const osc = line.charCodeAt(j);
+						if (osc === 0x07) {
+							i = j + 1;
+							break;
+						}
+						if (osc === 0x1b && line.charCodeAt(j + 1) === 0x5c) {
+							i = j + 2;
+							break;
+						}
+						j++;
+					}
+					if (j >= line.length) return undefined;
+					continue;
+				}
+				return undefined;
+			}
+			if (code < 0x20 || code > 0x7e) return undefined;
+			col++;
+			if (col > maxWidth) return col;
+			i++;
+		}
+		return col;
+	}
+
+	#lineRewriteSequence(line: string, width: number): string {
+		if (TERMINAL.isImageLine(line)) return ERASE_LINE + line;
+		const terminalLine = this.#terminalLine(line);
+		const asciiWidth = this.#ansiAsciiLineWidth(line, width);
+		const lineWidth = asciiWidth ?? visibleWidth(line);
+		return lineWidth >= width ? terminalLine : terminalLine + ERASE_TO_END_OF_LINE;
 	}
 
 	/**
@@ -2240,6 +2554,7 @@ export class TUI extends Container {
 	 */
 
 	#commit(lines: string[], width: number, height: number, viewportTop: number, hardwareCursorRow: number): void {
+		this.#deferredTailLine = undefined;
 		this.#previousLines = lines;
 		this.#previousVisibleOverlayComponents = this.#visibleOverlayComponentsThisRender;
 		this.#forceViewportRepaintOnNextRender = false;
@@ -2293,10 +2608,10 @@ export class TUI extends Container {
 		const visibleStart = Math.max(0, lines.length - height);
 		let fillSequence = "";
 		let visibleTexts: string[] | null = null;
-		if (TERMINAL.deccara && visibleStart < lines.length) {
+		if (this.#deccaraFillsEnabled() && visibleStart < lines.length) {
 			const visible: string[] = new Array(lines.length - visibleStart);
 			for (let k = 0; k < visible.length; k++) {
-				visible[k] = this.#fitLineToWidth(lines[visibleStart + k], width);
+				visible[k] = lines[visibleStart + k] ?? "";
 			}
 			const plan = planDeccaraFills(visible, width);
 			visibleTexts = plan.texts;
@@ -2304,8 +2619,9 @@ export class TUI extends Container {
 		}
 		for (let i = 0; i < lines.length; i++) {
 			if (i > 0) buffer += "\r\n";
-			buffer +=
-				visibleTexts && i >= visibleStart ? visibleTexts[i - visibleStart] : this.#fitLineToWidth(lines[i], width);
+			buffer += this.#terminalLine(
+				visibleTexts && i >= visibleStart ? visibleTexts[i - visibleStart] : (lines[i] ?? ""),
+			);
 		}
 		buffer += fillSequence;
 		const finalRow = Math.max(0, lines.length - 1);
@@ -2354,12 +2670,12 @@ export class TUI extends Container {
 		let wroteLine = false;
 		for (let i = 0; i < appendTo; i++) {
 			if (wroteLine) buffer += "\r\n";
-			buffer += this.#fitLineToWidth(lines[i] ?? "", width);
+			buffer += this.#terminalLine(lines[i] ?? "");
 			wroteLine = true;
 		}
 		for (let screenRow = 0; screenRow < height; screenRow++) {
 			if (wroteLine) buffer += "\r\n";
-			buffer += this.#fitLineToWidth(lines[viewportTop + screenRow] ?? "", width);
+			buffer += this.#terminalLine(lines[viewportTop + screenRow] ?? "");
 			wroteLine = true;
 		}
 
@@ -2387,20 +2703,38 @@ export class TUI extends Container {
 		cursorPos: { row: number; col: number } | null,
 	): void {
 		this.#fullRedrawCount += 1;
+		// A viewport repaint is a strictly in-place rewrite of the live window: it
+		// homes to the screen top and writes exactly `height` rows, so it must stay
+		// bottom-anchored at `lines.length - height`. Anchoring anywhere else pushes
+		// the live tail off the screen bottom (blank rows below the content) AND — far
+		// worse — persists the off-tail anchor into `#viewportTopRow` via `#commit`.
+		// A later frame then reads that inflated `prevViewportTop`, mis-classifies an
+		// ordinary tail change as an offscreen edit (`diff.firstChanged <
+		// prevViewportTop`), and re-routes into an append/scroll path that re-commits
+		// the same frame into native scrollback every tick — the self-driven "options
+		// drawn again and again" spam.
+		//
+		// This repaint cannot un-commit rows already in native scrollback (no safe ED3
+		// on ED3-risk hosts), so a shrink that re-exposes a committed prefix leaves a
+		// stale copy above the viewport. That is the accepted deferred state: the live
+		// window stays correct here, `#nativeScrollbackDirty` stays set, and the next
+		// at-tail checkpoint (`refreshNativeScrollbackIfDirty`) reconciles history with
+		// a clean clear+replay. Hiding the live tail to paper over the stale history —
+		// the previous "anti-duplication clamp" — traded a transient, off-screen
+		// history artifact for a broken live viewport, which is the worse defect.
 		const viewportTop = Math.max(0, lines.length - height);
 		// Each visible screen row, bottom-anchored, blank past content.
 		const visible: string[] = new Array(height);
 		for (let screenRow = 0; screenRow < height; screenRow++) {
-			visible[screenRow] = this.#fitLineToWidth(lines[viewportTop + screenRow] ?? "", width);
+			visible[screenRow] = lines[viewportTop + screenRow] ?? "";
 		}
-		const { texts, sequence } = TERMINAL.deccara
+		const { texts, sequence } = this.#deccaraFillsEnabled()
 			? planDeccaraFills(visible, width)
 			: { texts: visible, sequence: "" };
 		let buffer = `${this.#paintBeginSequence}\x1b[H`;
 		for (let screenRow = 0; screenRow < height; screenRow++) {
 			if (screenRow > 0) buffer += "\r\n";
-			buffer += "\x1b[2K";
-			buffer += texts[screenRow];
+			buffer += this.#lineRewriteSequence(texts[screenRow], width);
 		}
 		// DECCARA rectangles paint the visible fills before cursor positioning;
 		// the cleared cells written above are what the rectangles repaint.
@@ -2431,6 +2765,61 @@ export class TUI extends Container {
 		this.#commit(lines, width, height, viewportTop, toRow);
 	}
 
+	/** Topmost visible overlay requests the alternate-screen buffer. */
+	#wantsAltScreen(): boolean {
+		for (let i = this.overlayStack.length - 1; i >= 0; i--) {
+			const entry = this.overlayStack[i]!;
+			if (!this.#isOverlayVisible(entry)) continue;
+			return entry.options?.fullscreen === true;
+		}
+		return false;
+	}
+
+	/**
+	 * Compose and paint a single fullscreen overlay frame on the alt buffer.
+	 * Cursor markers are stripped (the modal draws its own in-band caret and
+	 * keeps the hardware cursor hidden), and only the modal is composited over a
+	 * blank base — the transcript is never touched while the alt buffer is up.
+	 */
+	#renderAltFrame(width: number, height: number): void {
+		const base: string[] = new Array(Math.max(0, height)).fill("");
+		let lines = this.#compositeOverlays(base, width, height);
+		this.#extractCursorPosition(lines, height);
+		lines = this.#prepareLines(lines, width, false);
+		this.#emitAltFrame(lines, width, height);
+	}
+
+	/**
+	 * Full per-row viewport rewrite on the alt buffer. Emits only sync-output
+	 * brackets, a cursor home, and per-row rewrites — never ED3, append-tail, or
+	 * any native-scrollback byte, so it is fully isolated from the planner and
+	 * #commit. The hardware cursor stays hidden (it is never re-shown here).
+	 */
+	#emitAltFrame(lines: string[], width: number, height: number): void {
+		const fitted: string[] = new Array(height);
+		for (let r = 0; r < height; r++) fitted[r] = lines[r] ?? "";
+		// Skip an identical repaint (the modal is mostly static between keystrokes).
+		if (this.#altPreviousLines.length === height) {
+			let same = true;
+			for (let r = 0; r < height; r++) {
+				if (fitted[r] !== this.#altPreviousLines[r]) {
+					same = false;
+					break;
+				}
+			}
+			if (same) return;
+		}
+		let buffer = `${this.#paintBeginSequence}\x1b[H`;
+		for (let r = 0; r < height; r++) {
+			if (r > 0) buffer += "\r\n";
+			buffer += this.#lineRewriteSequence(fitted[r], width);
+		}
+		buffer += this.#paintEndSequence;
+		this.terminal.write(buffer);
+		this.#altPreviousLines = fitted;
+		this.#fullRedrawCount += 1;
+	}
+
 	/**
 	 * Foreground-stream live-region paint for ED3-risk terminals with an
 	 * unobservable viewport. Commits the newly-sealed chunk to native scrollback
@@ -2438,8 +2827,8 @@ export class TUI extends Container {
 	 * leaving the transient live region out of saved lines.
 	 *
 	 * Uses only the no-scroll-snap vocabulary of {@link #emitDiff}: relative
-	 * cursor moves, per-line `\x1b[2K`, and `\r\n` to push the sealed chunk into
-	 * history. It deliberately avoids a full-screen erase (`\x1b[2J`) and absolute
+	 * cursor moves, per-row rewrite/suffix-clear, and `\r\n` to push the sealed
+	 * chunk into history. It deliberately avoids a full-screen erase (`\x1b[2J`) and absolute
 	 * cursor home (`\x1b[H`): on Ghostty those snap a reader scrolled into history
 	 * back to the bottom on every frame.
 	 */
@@ -2460,6 +2849,42 @@ export class TUI extends Container {
 		const boundedAppendTo = Math.max(0, Math.min(appendTo, naturalViewportTop, lines.length));
 		const boundedAppendFrom = Math.max(0, Math.min(appendFrom, boundedAppendTo));
 
+		if (boundedAppendFrom === boundedAppendTo && viewportTop === prevViewportTop) {
+			let firstChangedScreenRow = -1;
+			let lastChangedScreenRow = -1;
+			for (let screenRow = 0; screenRow < height; screenRow++) {
+				const nextLine = lines[viewportTop + screenRow] ?? "";
+				const previousLine = this.#previousLines[prevViewportTop + screenRow] ?? "";
+				if (nextLine === previousLine) continue;
+				if (firstChangedScreenRow === -1) firstChangedScreenRow = screenRow;
+				lastChangedScreenRow = screenRow;
+			}
+
+			let buffer = this.#paintBeginSequence;
+			let cursorFromRow = prevHardwareCursorRow;
+			if (firstChangedScreenRow !== -1) {
+				const clampedCursor = Math.min(prevHardwareCursorRow, prevViewportTop + height - 1);
+				const currentScreenRow = Math.max(0, Math.min(height - 1, clampedCursor - prevViewportTop));
+				const rowDelta = firstChangedScreenRow - currentScreenRow;
+				if (rowDelta > 0) buffer += `\x1b[${rowDelta}B`;
+				else if (rowDelta < 0) buffer += `\x1b[${-rowDelta}A`;
+				buffer += "\r";
+				for (let screenRow = firstChangedScreenRow; screenRow <= lastChangedScreenRow; screenRow++) {
+					if (screenRow > firstChangedScreenRow) buffer += "\r\n";
+					buffer += this.#lineRewriteSequence(lines[viewportTop + screenRow] ?? "", width);
+				}
+				cursorFromRow = viewportTop + lastChangedScreenRow;
+			}
+			const { seq, toRow } = this.#cursorControlSequence(cursorPos, lines.length, cursorFromRow);
+			buffer += seq;
+			buffer += this.#paintEndSequence;
+			this.terminal.write(buffer);
+
+			this.#maxLinesRendered = Math.max(lines.length, viewportTop + height);
+			this.#commit(lines, width, height, viewportTop, toRow);
+			return;
+		}
+
 		// Position at the top visible row with a relative move. Terminals clamp the
 		// hardware cursor to the viewport on resize, so clamp our tracking to match
 		// before computing the delta (mirrors #emitDiff).
@@ -2471,17 +2896,18 @@ export class TUI extends Container {
 
 		// Write the sealed chunk followed by the full viewport from the top row.
 		// The first (boundedAppendTo - boundedAppendFrom) rows scroll into native
-		// history; the trailing `height` rows fill the viewport. Each row clears
-		// itself with `\x1b[2K` instead of relying on a screen-wide erase.
+		// history; the trailing `height` rows fill the viewport. Text rows overwrite
+		// first and clear only the suffix so non-synchronized hosts do not visibly
+		// blank stable content before repainting it.
 		let wroteLine = false;
 		for (let i = boundedAppendFrom; i < boundedAppendTo; i++) {
 			if (wroteLine) buffer += "\r\n";
-			buffer += `\x1b[2K${this.#fitLineToWidth(lines[i] ?? "", width)}`;
+			buffer += this.#lineRewriteSequence(lines[i] ?? "", width);
 			wroteLine = true;
 		}
 		for (let screenRow = 0; screenRow < height; screenRow++) {
 			if (wroteLine) buffer += "\r\n";
-			buffer += `\x1b[2K${this.#fitLineToWidth(lines[viewportTop + screenRow] ?? "", width)}`;
+			buffer += this.#lineRewriteSequence(lines[viewportTop + screenRow] ?? "", width);
 			wroteLine = true;
 		}
 
@@ -2511,7 +2937,6 @@ export class TUI extends Container {
 		lines: string[],
 		start: number,
 		height: number,
-		width: number,
 		prevViewportTop: number,
 		prevHardwareCursorRow: number,
 	): void {
@@ -2526,7 +2951,7 @@ export class TUI extends Container {
 		if (moveToBottom > 0) buffer += `\x1b[${moveToBottom}B`;
 		for (let i = start; i < lines.length; i++) {
 			buffer += "\r\n";
-			buffer += this.#fitLineToWidth(lines[i], width);
+			buffer += this.#terminalLine(lines[i] ?? "");
 		}
 		buffer += this.#paintEndSequence;
 		this.terminal.write(buffer);
@@ -2536,6 +2961,41 @@ export class TUI extends Container {
 		}
 	}
 
+	/**
+	 * Paint only the active-grid bottom row while a scrollback mutation remains
+	 * deferred. If the native viewport is unknown and the user is scrolled up by a
+	 * single line, every active-grid row except the bottom can still be visible in
+	 * their scrollback window; touching only this row keeps that reader's viewport
+	 * unchanged while allowing bottom-anchored live chrome (spinner/status tail) to
+	 * advance for users at the tail.
+	 */
+	#emitDeferredTailRepaint(
+		line: string,
+		width: number,
+		height: number,
+		row: number,
+		prevViewportTop: number,
+		prevHardwareCursorRow: number,
+	): void {
+		const viewportBottom = prevViewportTop + height - 1;
+		if (row !== viewportBottom) return;
+
+		let buffer = this.#paintBeginSequence;
+		const clampedCursor = Math.min(prevHardwareCursorRow, viewportBottom);
+		const currentScreenRow = Math.max(0, Math.min(height - 1, clampedCursor - prevViewportTop));
+		const moveDown = height - 1 - currentScreenRow;
+		if (moveDown > 0) buffer += `\x1b[${moveDown}B`;
+		buffer += `\r${this.#lineRewriteSequence(line, width)}\x1b[?25l`;
+		buffer += this.#paintEndSequence;
+		this.terminal.write(buffer);
+
+		this.#deferredTailLine = line;
+		this.#previousWidth = width;
+		this.#previousHeight = height;
+		this.#viewportTopRow = prevViewportTop;
+		this.#hardwareCursorRow = row;
+	}
+
 	/**
 	 * Trailing-shrink: prior content shared a prefix with the new content; the
 	 * extra rows below the new tail need to be cleared without scrolling. Falls
@@ -2668,10 +3128,15 @@ export class TUI extends Container {
 		const fillStart = Math.max(firstChanged, fillViewportTop);
 		let fillSequence = "";
 		let fillTexts: string[] | null = null;
-		if (TERMINAL.deccara && !appendStart && moveTargetRow <= prevViewportBottom && renderEnd >= fillStart) {
+		if (
+			this.#deccaraFillsEnabled() &&
+			!appendStart &&
+			moveTargetRow <= prevViewportBottom &&
+			renderEnd >= fillStart
+		) {
 			const slice: string[] = new Array(renderEnd - fillStart + 1);
 			for (let i = fillStart; i <= renderEnd; i++) {
-				slice[i - fillStart] = this.#fitLineToWidth(lines[i], width);
+				slice[i - fillStart] = lines[i] ?? "";
 			}
 			const plan = planDeccaraFills(slice, width, fillStart - fillViewportTop);
 			fillTexts = plan.texts;
@@ -2679,8 +3144,7 @@ export class TUI extends Container {
 		}
 		for (let i = firstChanged; i <= renderEnd; i++) {
 			if (i > firstChanged) buffer += "\r\n";
-			buffer += "\x1b[2K";
-			buffer += fillTexts && i >= fillStart ? fillTexts[i - fillStart] : this.#fitLineToWidth(lines[i], width);
+			buffer += this.#lineRewriteSequence(fillTexts && i >= fillStart ? fillTexts[i - fillStart] : lines[i], width);
 		}
 
 		// If the prior frame was taller, clear the trailing rows.
@@ -2740,8 +3204,14 @@ export class TUI extends Container {
 					? `${intent.kind}(append=${intent.appendFrom}..${intent.appendTo}, viewportTop=${intent.renderViewportTop})`
 					: intent.kind === "viewportRepaint" && intent.appendFrom !== undefined
 						? `${intent.kind}(appendFrom=${intent.appendFrom})`
-						: intent.kind;
-		const msg = `[${new Date().toISOString()}] render: ${detail} (prev=${this.#previousLines.length}, new=${newLength}, height=${height})\n`;
+						: intent.kind === "deferredTailRepaint"
+							? `${intent.kind}(row=${intent.row})`
+							: intent.kind;
+		const state =
+			`shw=${this.#scrollbackHighWater}, max=${this.#maxLinesRendered}, vpTop=${this.#viewportTopRow}, ` +
+			`dirty=${this.#nativeScrollbackDirty}, eager=${this.#eagerNativeScrollbackRebuild}, ` +
+			`lrStart=${this.#nativeScrollbackLiveRegionStart}, commitSafeEnd=${this.#nativeScrollbackCommitSafeEnd}`;
+		const msg = `[${new Date().toISOString()}] render: ${detail} (prev=${this.#previousLines.length}, new=${newLength}, height=${height}, ${state})\n`;
 		fs.appendFileSync(getDebugLogPath(), msg);
 	}
 
diff --git a/packages/tui/src/utils.ts b/packages/tui/src/utils.ts
index 058a0ff62..243741dd3 100644
--- a/packages/tui/src/utils.ts
+++ b/packages/tui/src/utils.ts
@@ -4,7 +4,6 @@ import {
 	extractSegments as nativeExtractSegments,
 	sliceWithWidth as nativeSliceWithWidth,
 	truncateToWidth as nativeTruncateToWidth,
-	visibleWidth as nativeVisibleWidth,
 	wrapTextWithAnsi as nativeWrapTextWithAnsi,
 	type SliceResult,
 } from "@oh-my-pi/pi-natives";
@@ -145,44 +144,6 @@ export function padding(n: number): string {
 // Grapheme segmenter (shared instance)
 const segmenter = new Intl.Segmenter(undefined, { granularity: "grapheme" });
 
-const EXTENDED_PICTOGRAPHIC_REGEX = /\p{Extended_Pictographic}/u;
-
-// Matches CSI (`\x1b[…`) and OSC (`\x1b]…` terminated by BEL/ST) escape
-// sequences. Mirrors the standard ansi-regex coverage so visible-span
-// segmentation lines up with the native ANSI scanner.
-const ANSI_ESCAPE_REGEX =
-	/[\u001b\u009b][[\]()#;?]*(?:(?:(?:(?:;[-a-zA-Z\d/#&.:=?%@~_]+)*|[a-zA-Z\d]+(?:;[-a-zA-Z\d/#&.:=?%@~_]*)*)?\u0007)|(?:(?:\d{1,4}(?:;\d{0,4})*)?[\dA-PR-TZcf-nq-uy=><~]))/g;
-
-function pictographicSpanWidth(span: string): number {
-	let width = 0;
-	for (const { segment } of segmenter.segment(span)) {
-		width += EXTENDED_PICTOGRAPHIC_REGEX.test(segment) ? 2 : nativeVisibleWidth(segment, getDefaultTabWidth());
-	}
-	return width;
-}
-
-// Width fallback for strings that mix ANSI styling with ZWJ pictographic
-// emoji. `Intl.Segmenter` would split an escape sequence into individual
-// graphemes, so the native scanner (which only skips ANSI when handed the
-// complete sequence) double-counts the printable SGR bytes. Excise the ANSI
-// spans first — they contribute zero cells — and apply the pictographic
-// grapheme override only to the visible spans, then sum.
-function visibleWidthByGrapheme(str: string): number {
-	let width = 0;
-	let lastIndex = 0;
-	ANSI_ESCAPE_REGEX.lastIndex = 0;
-	for (let match = ANSI_ESCAPE_REGEX.exec(str); match !== null; match = ANSI_ESCAPE_REGEX.exec(str)) {
-		if (match.index > lastIndex) {
-			width += pictographicSpanWidth(str.slice(lastIndex, match.index));
-		}
-		lastIndex = ANSI_ESCAPE_REGEX.lastIndex;
-	}
-	if (lastIndex < str.length) {
-		width += lastIndex === 0 ? pictographicSpanWidth(str) : pictographicSpanWidth(str.slice(lastIndex));
-	}
-	return width;
-}
-
 /**
  * Get the shared grapheme segmenter instance.
  */
@@ -190,35 +151,106 @@ export function getSegmenter(): Intl.Segmenter {
 	return segmenter;
 }
 
-export function visibleWidthRaw(str: string): number {
-	if (!str) {
-		return 0;
-	}
+// Kitty OSC 66 text-sizing spans: `\x1b]66;<meta>;<payload>` terminated by BEL
+// or ST. `Bun.stringWidth` strips the whole span (payload included) to zero
+// cells, but the payload is visible and scales by the `s=` factor, so each is
+// added back so width matches the native truncate/slice/wrap helpers.
+const OSC66_SPAN_REGEX = /\x1b\]66;([^;]*);([\s\S]*?)(?:\x07|\x1b\\)/g;
+const OSC66_PREFIX = "\x1b]66;";
+const ESC = "\x1b";
+const TAB = "\t";
+const LONG_WIDTH_FAST_PATH_MIN = 128;
 
-	// Fast path: printable ASCII has one cell per code unit. Defer every
-	// control/non-ASCII case (tabs, ANSI/OSC, combining marks, CJK) to the
-	// native text engine so all width/slice/wrap helpers share one Unicode
-	// model instead of mixing Bun.stringWidth quirks with Rust truncation.
-	for (let i = 0; i < str.length; i++) {
-		const code = str.charCodeAt(i);
-		if (code < 0x20 || code > 0x7e) {
-			const tabWidth = getDefaultTabWidth();
-			if (str.includes("\x1b]66;")) return nativeVisibleWidth(str, tabWidth);
-			return str.includes("\u200d") ? visibleWidthByGrapheme(str) : nativeVisibleWidth(str, tabWidth);
-		}
-	}
-	return str.length;
-}
+// Pin Bun.stringWidth semantics to the native width engine and guard against Bun
+// default drift: strip ANSI/OSC (don't count escape bytes) and treat
+// ambiguous-width East Asian chars as narrow (1 cell), matching `unicode-width`'s
+// non-CJK tables that back truncate/slice/wrap. Hoisted so no per-call alloc.
+const STRING_WIDTH_OPTS = { countAnsiEscapeCodes: false, ambiguousIsNarrow: true } as const;
 
 /**
- * Calculate the visible width of a string in terminal columns.
+ * Visible width of a string in terminal columns, excluding ANSI/OSC escapes.
+ *
+ * `Bun.stringWidth` does the heavy lifting (UAX#11 width tables + ANSI/OSC
+ * stripping); this adds the two corrections it omits — tabs (expanded to
+ * `tabWidth` cells) and OSC 66 text-sizing payloads (scaled by `s=`).
  */
 export function visibleWidth(str: string): number {
 	if (!str) return 0;
-	return visibleWidthRaw(str);
+
+	// Long non-escape text is faster through Bun's native scanner than through
+	// a JS printable-ASCII prepass. Escape-bearing strings stay on the scanner
+	// below so CSI/OSC-heavy render output can still bail out at the first ESC.
+	if (str.length >= LONG_WIDTH_FAST_PATH_MIN && !str.includes(ESC)) {
+		let width = Bun.stringWidth(str, STRING_WIDTH_OPTS);
+		let tabCount = 0;
+		for (let tabIndex = str.indexOf(TAB); tabIndex !== -1; tabIndex = str.indexOf(TAB, tabIndex + 1)) {
+			tabCount++;
+		}
+		if (tabCount > 0) width += tabCount * getDefaultTabWidth();
+		return width;
+	}
+
+	let tabCount = 0;
+	let i = 0;
+	for (; i < str.length; i++) {
+		const code = str.charCodeAt(i);
+		if (code < 0x20 || code > 0x7e) {
+			if (code === 0x09) {
+				tabCount++;
+				continue;
+			}
+			break;
+		}
+	}
+	if (i === str.length) {
+		return tabCount === 0 ? str.length : str.length + tabCount * (getDefaultTabWidth() - 1);
+	}
+
+	if (tabCount === 0) {
+		let tabIndex = str.indexOf(TAB, i + 1);
+		if (tabIndex !== -1) {
+			tabCount = 1;
+			for (tabIndex = str.indexOf(TAB, tabIndex + 1); tabIndex !== -1; tabIndex = str.indexOf(TAB, tabIndex + 1)) {
+				tabCount++;
+			}
+		}
+	} else {
+		for (let tabIndex = str.indexOf(TAB, i + 1); tabIndex !== -1; tabIndex = str.indexOf(TAB, tabIndex + 1)) {
+			tabCount++;
+		}
+	}
+
+	// `Bun.stringWidth` is a JSC builtin (no per-call N-API number box, unlike
+	// the native scanner that traps under Bun 1.3.x GC/N-API load). It strips
+	// CSI/OSC to zero cells and shares the native engine's UAX#11 width tables.
+	let width = Bun.stringWidth(str, STRING_WIDTH_OPTS);
+	if (tabCount > 0) width += tabCount * getDefaultTabWidth();
+
+	// OSC 66: add back each stripped span as `scale * (explicit w ?? payload
+	// width)`. Matched rather than replaced to avoid reallocating the string.
+	if (str.includes(OSC66_PREFIX, i)) {
+		OSC66_SPAN_REGEX.lastIndex = 0;
+		for (let m = OSC66_SPAN_REGEX.exec(str); m !== null; m = OSC66_SPAN_REGEX.exec(str)) {
+			let scale = 1;
+			let explicit: number | undefined;
+			for (const part of m[1].split(":")) {
+				// metadata keys are single chars, e.g. `s=2`, `w=5`
+				if (part.indexOf("=") !== 1) continue;
+				const value = Number.parseInt(part.slice(2), 10);
+				if (!Number.isFinite(value)) continue;
+				if (part[0] === "s") {
+					if (value >= 1 && value <= 7) scale = value;
+				} else if (part[0] === "w" && value > 0) {
+					explicit = value;
+				}
+			}
+			width += scale * (explicit ?? Bun.stringWidth(m[2], STRING_WIDTH_OPTS));
+		}
+	}
+
+	return width;
 }
 
-const THAI_LAO_AM_REGEX = /[\u0e33\u0eb3]/;
 const THAI_LAO_AM_GLOBAL_REGEX = /[\u0e33\u0eb3]/g;
 
 /**
@@ -228,7 +260,7 @@ const THAI_LAO_AM_GLOBAL_REGEX = /[\u0e33\u0eb3]/g;
  * width but avoid stale-cell artifacts in terminal renderers.
  */
 export function normalizeTerminalOutput(str: string): string {
-	if (!THAI_LAO_AM_REGEX.test(str)) return str;
+	if (str.indexOf("\u0e33") === -1 && str.indexOf("\u0eb3") === -1) return str;
 	return str.replace(THAI_LAO_AM_GLOBAL_REGEX, char => (char === "\u0e33" ? "\u0e4d\u0e32" : "\u0ecd\u0eb2"));
 }
 
diff --git a/packages/tui/test/container-dispose.test.ts b/packages/tui/test/container-dispose.test.ts
new file mode 100644
index 000000000..7c497a2f6
--- /dev/null
+++ b/packages/tui/test/container-dispose.test.ts
@@ -0,0 +1,30 @@
+import { describe, expect, it } from "bun:test";
+import { type Component, Container } from "@oh-my-pi/pi-tui";
+
+function inert(dispose?: () => void): Component {
+	return { render: () => [], invalidate: () => {}, dispose };
+}
+
+describe("Container.dispose", () => {
+	it("calls dispose on each child and tolerates children without dispose", () => {
+		const order: string[] = [];
+		const container = new Container();
+		container.addChild(inert(() => order.push("a")));
+		container.addChild(inert()); // no dispose — must not throw
+		container.addChild(inert(() => order.push("c")));
+
+		expect(() => container.dispose()).not.toThrow();
+		expect(order).toEqual(["a", "c"]);
+	});
+
+	it("recurses through nested containers", () => {
+		let leafDisposed = 0;
+		const outer = new Container();
+		const inner = new Container();
+		inner.addChild(inert(() => leafDisposed++));
+		outer.addChild(inner);
+
+		outer.dispose();
+		expect(leafDisposed).toBe(1);
+	});
+});
diff --git a/packages/tui/test/deccara-grid-parity.test.ts b/packages/tui/test/deccara-grid-parity.test.ts
deleted file mode 100644
index 9eb7ca28b..000000000
--- a/packages/tui/test/deccara-grid-parity.test.ts
+++ /dev/null
@@ -1,470 +0,0 @@
-import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
-import { performance } from "node:perf_hooks";
-import { applyBackgroundToLine, type Component, setTerminalDeccara, TERMINAL, TUI } from "@oh-my-pi/pi-tui";
-import { VirtualTerminal } from "./virtual-terminal";
-
-const BG_OPEN = "\x1b[48;2;10;20;30m";
-
-class PanelComponent implements Component {
-	#rows: string[];
-	#bg: boolean[];
-	constructor(rows: string[], bg: boolean[]) {
-		this.#rows = [...rows];
-		this.#bg = [...bg];
-	}
-	set(rows: string[], bg: boolean[]): void {
-		this.#rows = [...rows];
-		this.#bg = [...bg];
-	}
-	invalidate(): void {}
-	render(width: number): string[] {
-		return this.#rows.map((row, i) =>
-			this.#bg[i] ? applyBackgroundToLine(row, width, t => `${BG_OPEN}${t}\x1b[49m`) : row,
-		);
-	}
-}
-
-async function settle(term: VirtualTerminal): Promise<void> {
-	const t = Promise.withResolvers<void>();
-	process.nextTick(t.resolve);
-	await t.promise;
-	await Bun.sleep(1);
-	await term.flush();
-}
-
-// ---- Minimal DECCARA-aware grid simulator -------------------------------
-// Tracks per-cell background only (the optimizer's whole job). Background is a
-// normalized SGR param string, or "" for default. Implements the escape subset
-// the renderer emits, including DECSACE rect-mode + DECCARA $r rectangle fills,
-// with background-color-erase (bce) semantics on ED/EL/scroll.
-function normBg(prev: string, params: string): string {
-	if (params.length === 0) return "";
-	const t = params.split(";");
-	let r = prev;
-	for (let i = 0; i < t.length; i++) {
-		const n = t[i].length === 0 ? 0 : Number(t[i]);
-		if (!Number.isInteger(n)) return r; // ignore unknowns
-		if (n === 0 || n === 49) r = "";
-		else if ((n >= 40 && n <= 47) || (n >= 100 && n <= 107)) r = t[i];
-		else if (n === 48) {
-			if (t[i + 1] === "5") {
-				r = `48;5;${t[i + 2]}`;
-				i += 2;
-			} else if (t[i + 1] === "2") {
-				r = `48;2;${t[i + 2]};${t[i + 3]};${t[i + 4]}`;
-				i += 4;
-			}
-		} else if (n === 38) {
-			if (t[i + 1] === "5") i += 2;
-			else if (t[i + 1] === "2") i += 4;
-		}
-	}
-	return r;
-}
-
-class Grid {
-	w: number;
-	h: number;
-	bg: string[][];
-	row = 0;
-	col = 0;
-	curBg = "";
-	wrapPending = false;
-	constructor(w: number, h: number) {
-		this.w = w;
-		this.h = h;
-		this.bg = Array.from({ length: h }, () => new Array(w).fill(""));
-	}
-	#scrollUp(): void {
-		this.bg.shift();
-		this.bg.push(new Array(this.w).fill(this.curBg));
-	}
-	#nl(): void {
-		this.wrapPending = false;
-		if (this.row >= this.h - 1) this.#scrollUp();
-		else this.row++;
-	}
-	#put(): void {
-		if (this.wrapPending) {
-			this.col = 0;
-			this.#nl();
-		}
-		if (this.col < this.w) this.bg[this.row][this.col] = this.curBg;
-		if (this.col >= this.w - 1) this.wrapPending = true;
-		else this.col++;
-	}
-	#erase(r: number, c0: number, c1: number): void {
-		for (let c = c0; c <= c1 && c < this.w; c++) this.bg[r][c] = this.curBg;
-	}
-	write(s: string): void {
-		let i = 0;
-		while (i < s.length) {
-			const code = s.charCodeAt(i);
-			if (code === 0x1b) {
-				const n = s[i + 1];
-				if (n === "[") {
-					let j = i + 2;
-					while (j < s.length && !(s.charCodeAt(j) >= 0x40 && s.charCodeAt(j) <= 0x7e)) j++;
-					const fin = s[j];
-					// Strip trailing intermediates (0x20-0x2f), e.g. '$' in DECCARA ($r)
-					// and '*' in DECSACE (*x), so the param list is clean.
-					let bodyEnd = j;
-					while (bodyEnd > i + 2 && s.charCodeAt(bodyEnd - 1) >= 0x20 && s.charCodeAt(bodyEnd - 1) <= 0x2f)
-						bodyEnd--;
-					const inter = s.slice(bodyEnd, j);
-					const body = s.slice(i + 2, bodyEnd);
-					this.#csi(body, inter + fin);
-					i = j + 1;
-					continue;
-				}
-				if (n === "]") {
-					// OSC ... BEL or ST
-					let j = i + 2;
-					while (j < s.length && s.charCodeAt(j) !== 0x07 && !(s[j] === "\x1b" && s[j + 1] === "\\")) j++;
-					i = s.charCodeAt(j) === 0x07 ? j + 1 : j + 2;
-					continue;
-				}
-				if (n === "P" || n === "X" || n === "^" || n === "_") {
-					let j = i + 2;
-					while (j < s.length && !(s[j] === "\x1b" && s[j + 1] === "\\")) j++;
-					i = j + 2;
-					continue;
-				}
-				i += 2;
-				continue;
-			}
-			if (code === 0x0d) {
-				this.col = 0;
-				this.wrapPending = false;
-				i++;
-				continue;
-			}
-			if (code === 0x0a) {
-				this.#nl();
-				i++;
-				continue;
-			}
-			// printable
-			this.#put();
-			i++;
-		}
-	}
-	#csi(body: string, fin: string): void {
-		const ps = body.startsWith("?") ? body.slice(1) : body;
-		const args = ps.split(";");
-		const a0 = args[0]?.length ? Number(args[0]) : undefined;
-		switch (fin) {
-			case "m":
-				this.curBg = normBg(this.curBg, body);
-				return;
-			case "H":
-			case "f": {
-				const r = (args[0]?.length ? Number(args[0]) : 1) - 1;
-				const c = (args[1]?.length ? Number(args[1]) : 1) - 1;
-				this.row = Math.max(0, Math.min(this.h - 1, r));
-				this.col = Math.max(0, Math.min(this.w - 1, c));
-				this.wrapPending = false;
-				return;
-			}
-			case "A":
-				this.row = Math.max(0, this.row - (a0 ?? 1));
-				this.wrapPending = false;
-				return;
-			case "B":
-				this.row = Math.min(this.h - 1, this.row + (a0 ?? 1));
-				this.wrapPending = false;
-				return;
-			case "C":
-				this.col = Math.min(this.w - 1, this.col + (a0 ?? 1));
-				this.wrapPending = false;
-				return;
-			case "D":
-				this.col = Math.max(0, this.col - (a0 ?? 1));
-				this.wrapPending = false;
-				return;
-			case "G":
-				this.col = Math.max(0, Math.min(this.w - 1, (a0 ?? 1) - 1));
-				this.wrapPending = false;
-				return;
-			case "K": {
-				const mode = a0 ?? 0;
-				if (mode === 0) this.#erase(this.row, this.col, this.w - 1);
-				else if (mode === 1) this.#erase(this.row, 0, this.col);
-				else this.#erase(this.row, 0, this.w - 1);
-				return;
-			}
-			case "J": {
-				const mode = a0 ?? 0;
-				if (mode === 2 || mode === 3 || mode === 22) {
-					for (let r = 0; r < this.h; r++) this.#erase(r, 0, this.w - 1);
-				} else if (mode === 0) {
-					this.#erase(this.row, this.col, this.w - 1);
-					for (let r = this.row + 1; r < this.h; r++) this.#erase(r, 0, this.w - 1);
-				} else if (mode === 1) {
-					for (let r = 0; r < this.row; r++) this.#erase(r, 0, this.w - 1);
-					this.#erase(this.row, 0, this.col);
-				}
-				return;
-			}
-			case "*x":
-				return; // DECSACE extent: renderer always wraps $r in rect mode
-			case "$r": {
-				// DECCARA: Pt;Pl;Pb;Pr; <sgr...>
-				const top = (args[0]?.length ? Number(args[0]) : 1) - 1;
-				const left = (args[1]?.length ? Number(args[1]) : 1) - 1;
-				const bot = (args[2]?.length ? Number(args[2]) : this.h) - 1;
-				const right = (args[3]?.length ? Number(args[3]) : this.w) - 1;
-				const sgr = args.slice(4).join(";");
-				for (let r = Math.max(0, top); r <= Math.min(this.h - 1, bot); r++) {
-					for (let c = Math.max(0, left); c <= Math.min(this.w - 1, right); c++) {
-						this.bg[r][c] = normBg(this.bg[r][c], sgr);
-					}
-				}
-				return;
-			}
-			default:
-				return; // ignore h/l/t/etc.
-		}
-	}
-	dump(): string {
-		return this.bg.map(r => r.map(c => (c === "" ? "." : "#")).join("")).join("\n");
-	}
-}
-
-function captureBytes(term: VirtualTerminal): string[] {
-	const writes: string[] = [];
-	const real = term.write.bind(term);
-	vi.spyOn(term, "write").mockImplementation((d: string) => {
-		writes.push(d);
-		real(d);
-	});
-	return writes;
-}
-
-describe("DECCARA visible-fill parity (simulated grid)", () => {
-	beforeEach(() => {
-		let m = 0;
-		vi.spyOn(performance, "now").mockImplementation(() => {
-			m += 20;
-			return m;
-		});
-	});
-	afterEach(() => {
-		vi.restoreAllMocks();
-	});
-
-	async function runScenario(
-		deccara: boolean,
-		w: number,
-		h: number,
-		frames: { rows: string[]; bg: boolean[] }[],
-	): Promise<Grid> {
-		const saved = TERMINAL.deccara;
-		setTerminalDeccara(deccara);
-		const term = new VirtualTerminal(w, h);
-		const tui = new TUI(term);
-		const panel = new PanelComponent(frames[0].rows, frames[0].bg);
-		tui.addChild(panel);
-		const grid = new Grid(w, h);
-		const bytes = captureBytes(term);
-		try {
-			tui.start();
-			await settle(term);
-			for (let f = 1; f < frames.length; f++) {
-				panel.set(frames[f].rows, frames[f].bg);
-				tui.requestRender();
-				await settle(term);
-			}
-			for (const b of bytes) grid.write(b);
-			return grid;
-		} finally {
-			tui.stop();
-			setTerminalDeccara(saved);
-		}
-	}
-
-	const scenarios: { name: string; w: number; h: number; frames: { rows: string[]; bg: boolean[] }[] }[] = [
-		{
-			// Regression: a diff that grows the buffer past the viewport scrolls the
-			// terminal, so the rewritten rows settle one row higher than where they
-			// were first painted. The absolute DECCARA rectangles must follow the
-			// post-scroll rows (and the row scrolled into history must keep its full
-			// background padding). Pre-fix, the bg fills landed one row too low.
-			name: "diff grows past viewport, scrolling a bg fill row (regression)",
-			w: 20,
-			h: 4,
-			frames: [
-				{ rows: ["x", "y"], bg: [false, false] },
-				{ rows: ["X", "y", "z", "", ""], bg: [false, false, false, true, true] },
-			],
-		},
-		{
-			name: "short content + 2-row input box, then spinner diff",
-			w: 40,
-			h: 12,
-			frames: [
-				{
-					rows: ["Hi! What can I help you with today?", "", "can you ping google", "", "", "", "Working"],
-					bg: [false, false, false, false, true, true, false],
-				},
-				{
-					rows: ["Hi! What can I help you with today?", "", "can you ping google", "", "", "", "Working."],
-					bg: [false, false, false, false, true, true, false],
-				},
-			],
-		},
-		{
-			name: "blank panel filling viewport",
-			w: 40,
-			h: 8,
-			frames: [{ rows: ["", "", "", ""], bg: [true, true, true, true] }],
-		},
-		{
-			name: "content taller than viewport (scroll)",
-			w: 40,
-			h: 8,
-			frames: [
-				{
-					rows: Array.from({ length: 12 }, (_, i) => (i % 2 === 0 ? "" : `line ${i}`)),
-					bg: Array.from({ length: 12 }, () => true),
-				},
-			],
-		},
-		{
-			name: "in-place single-row change among bg rows",
-			w: 40,
-			h: 8,
-			frames: [
-				{ rows: ["AAA", "BBB", "CCC", "DDD"], bg: [true, true, true, true] },
-				{ rows: ["AAA", "BBB", "XXX", "DDD"], bg: [true, true, true, true] },
-			],
-		},
-		{
-			name: "streaming insert above input box (content < height)",
-			w: 40,
-			h: 14,
-			frames: [
-				{
-					rows: ["msg one", "", "can you ping google", "", "", "", "Working"],
-					bg: [false, false, false, false, true, true, false],
-				},
-				{
-					rows: ["msg one", "", "can you ping google", "", "Pinging", "", "", "", "Working"],
-					bg: [false, false, false, false, false, false, true, true, false],
-				},
-				{
-					rows: ["msg one", "", "can you ping google", "", "Pinging google", "now", "", "", "", "Working."],
-					bg: [false, false, false, false, false, false, false, true, true, false],
-				},
-				{
-					rows: ["msg one", "", "can you ping google", "", "Pinging google", "now done", "", "", "", "Working.."],
-					bg: [false, false, false, false, false, false, false, true, true, false],
-				},
-			],
-		},
-		{
-			name: "streaming insert above input box (content > height, scroll)",
-			w: 40,
-			h: 8,
-			frames: [
-				{ rows: ["a", "b", "c", "d", "", "", "stat"], bg: [false, false, false, false, true, true, false] },
-				{
-					rows: ["a", "b", "c", "d", "e", "", "", "stat"],
-					bg: [false, false, false, false, false, true, true, false],
-				},
-				{
-					rows: ["a", "b", "c", "d", "e", "f", "", "", "stat"],
-					bg: [false, false, false, false, false, false, true, true, false],
-				},
-				{
-					rows: ["a", "b", "c", "d", "e", "f", "g", "", "", "stat"],
-					bg: [false, false, false, false, false, false, false, true, true, false],
-				},
-			],
-		},
-		{
-			name: "full-viewport frame, in-place multi-row diff incl bg box",
-			w: 40,
-			h: 6,
-			frames: [
-				{ rows: ["top", "mid", "", "", "stat"], bg: [false, false, true, true, false] },
-				{ rows: ["top", "CHANGED", "", "", "stat2"], bg: [false, false, true, true, false] },
-			],
-		},
-		{
-			name: "shrink frame: assistant block collapses",
-			w: 40,
-			h: 12,
-			frames: [
-				{
-					rows: ["q", "", "aaaa", "bbbb", "cccc", "", "", "stat"],
-					bg: [false, false, false, false, false, true, true, false],
-				},
-				{ rows: ["q", "", "done", "", "", "stat"], bg: [false, false, false, true, true, false] },
-			],
-		},
-	];
-
-	for (const sc of scenarios) {
-		it(`matches padded fallback: ${sc.name}`, async () => {
-			const off = await runScenario(false, sc.w, sc.h, sc.frames);
-			const on = await runScenario(true, sc.w, sc.h, sc.frames);
-			if (off.dump() !== on.dump()) {
-				const offL = off.dump().split("\n");
-				const onL = on.dump().split("\n");
-				const diff = offL
-					.map((l, i) => (l === onL[i] ? null : `row ${i}:\n  off ${l}\n  on  ${onL[i]}`))
-					.filter(Boolean)
-					.join("\n");
-				throw new Error(`bg grid diverged for "${sc.name}":\n${diff}`);
-			}
-			expect(on.dump()).toBe(off.dump());
-		});
-	}
-
-	it("fuzz: random multi-frame sequences match the padded fallback", async () => {
-		const words = ["", "", "ok", "ping", "google", "running cmd", "x", "hello world here", "a longer content line"];
-		const failures: string[] = [];
-		let totalTrials = 0;
-		for (const startSeed of [0x12345678, 0x0badf00d, 0x7fffff00]) {
-			// Deterministic LCG so a failure is reproducible from the seed.
-			let seed = startSeed;
-			const rnd = () => {
-				seed = (seed * 1103515245 + 12345) & 0x7fffffff;
-				return seed / 0x7fffffff;
-			};
-			const pick = <T>(arr: T[]): T => arr[Math.floor(rnd() * arr.length)];
-			const makeFrame = (w: number) => {
-				const n = 1 + Math.floor(rnd() * 16);
-				const rows: string[] = [];
-				const bg: boolean[] = [];
-				for (let i = 0; i < n; i++) {
-					let r = pick(words);
-					if (r.length > w) r = r.slice(0, w);
-					rows.push(r);
-					bg.push(rnd() < 0.5);
-				}
-				return { rows, bg };
-			};
-			for (let trial = 0; trial < 100; trial++) {
-				totalTrials++;
-				const w = 16 + Math.floor(rnd() * 40);
-				const h = 3 + Math.floor(rnd() * 12);
-				const frameCount = 1 + Math.floor(rnd() * 6);
-				const frames = Array.from({ length: frameCount }, () => makeFrame(w));
-				const off = await runScenario(false, w, h, frames);
-				const on = await runScenario(true, w, h, frames);
-				if (off.dump() !== on.dump()) {
-					failures.push(
-						`seed ${startSeed.toString(16)} trial ${trial} (w=${w},h=${h}):\n${JSON.stringify(frames)}\n--- off\n${off.dump()}\n--- on\n${on.dump()}`,
-					);
-					if (failures.length >= 3) break;
-				}
-			}
-			if (failures.length >= 3) break;
-		}
-		if (failures.length > 0) {
-			throw new Error(`DECCARA fuzz divergences (${totalTrials} trials):\n${failures.join("\n\n")}`);
-		}
-	}, 60000);
-});
diff --git a/packages/tui/test/deccara.test.ts b/packages/tui/test/deccara.test.ts
index d5dd82e48..b95bb1b23 100644
--- a/packages/tui/test/deccara.test.ts
+++ b/packages/tui/test/deccara.test.ts
@@ -49,7 +49,7 @@ async function settle(term: VirtualTerminal): Promise<void> {
 	const nextTick = Promise.withResolvers<void>();
 	process.nextTick(nextTick.resolve);
 	await nextTick.promise;
-	await Bun.sleep(1);
+	await Bun.sleep(40);
 	await term.flush();
 }
 
@@ -74,6 +74,41 @@ function countOccurrences(haystack: string, needle: string): number {
 	}
 }
 
+async function withEnvPatch<T>(patch: Record<string, string | undefined>, run: () => T | Promise<T>): Promise<T> {
+	const bunSnapshot: Record<string, string | undefined> = {};
+	const processSnapshot: Record<string, string | undefined> = {};
+	for (const key in patch) {
+		bunSnapshot[key] = Bun.env[key];
+		processSnapshot[key] = process.env[key];
+		const value = patch[key];
+		if (value === undefined) {
+			delete Bun.env[key];
+			delete process.env[key];
+		} else {
+			Bun.env[key] = value;
+			process.env[key] = value;
+		}
+	}
+	try {
+		return await run();
+	} finally {
+		for (const key in patch) {
+			const bunValue = bunSnapshot[key];
+			if (bunValue === undefined) {
+				delete Bun.env[key];
+			} else {
+				Bun.env[key] = bunValue;
+			}
+			const processValue = processSnapshot[key];
+			if (processValue === undefined) {
+				delete process.env[key];
+			} else {
+				process.env[key] = processValue;
+			}
+		}
+	}
+}
+
 describe("detectRectangularSgrSupport", () => {
 	it("enables only kitty, which implements the SGR-background extension", () => {
 		expect(detectRectangularSgrSupport("kitty", {})).toBe(true);
@@ -240,8 +275,17 @@ describe("planDeccaraFills", () => {
 
 describe("TUI DECCARA integration", () => {
 	const savedDeccara = TERMINAL.deccara;
+	const savedForceSyncOutput = Bun.env.PI_FORCE_SYNC_OUTPUT;
+	const savedNoSyncOutput = Bun.env.PI_NO_SYNC_OUTPUT;
+	const savedTuiSyncOutput = Bun.env.PI_TUI_SYNC_OUTPUT;
 
 	beforeEach(() => {
+		Bun.env.PI_FORCE_SYNC_OUTPUT = "1";
+		process.env.PI_FORCE_SYNC_OUTPUT = "1";
+		delete Bun.env.PI_NO_SYNC_OUTPUT;
+		delete process.env.PI_NO_SYNC_OUTPUT;
+		delete Bun.env.PI_TUI_SYNC_OUTPUT;
+		delete process.env.PI_TUI_SYNC_OUTPUT;
 		let monotonic = 0;
 		vi.spyOn(performance, "now").mockImplementation(() => {
 			monotonic += 20;
@@ -250,6 +294,27 @@ describe("TUI DECCARA integration", () => {
 	});
 
 	afterEach(() => {
+		if (savedForceSyncOutput === undefined) {
+			delete Bun.env.PI_FORCE_SYNC_OUTPUT;
+			delete process.env.PI_FORCE_SYNC_OUTPUT;
+		} else {
+			Bun.env.PI_FORCE_SYNC_OUTPUT = savedForceSyncOutput;
+			process.env.PI_FORCE_SYNC_OUTPUT = savedForceSyncOutput;
+		}
+		if (savedNoSyncOutput === undefined) {
+			delete Bun.env.PI_NO_SYNC_OUTPUT;
+			delete process.env.PI_NO_SYNC_OUTPUT;
+		} else {
+			Bun.env.PI_NO_SYNC_OUTPUT = savedNoSyncOutput;
+			process.env.PI_NO_SYNC_OUTPUT = savedNoSyncOutput;
+		}
+		if (savedTuiSyncOutput === undefined) {
+			delete Bun.env.PI_TUI_SYNC_OUTPUT;
+			delete process.env.PI_TUI_SYNC_OUTPUT;
+		} else {
+			Bun.env.PI_TUI_SYNC_OUTPUT = savedTuiSyncOutput;
+			process.env.PI_TUI_SYNC_OUTPUT = savedTuiSyncOutput;
+		}
 		setTerminalDeccara(savedDeccara);
 		vi.restoreAllMocks();
 	});
@@ -299,6 +364,32 @@ describe("TUI DECCARA integration", () => {
 		}
 	});
 
+	it("keeps padded fallback bytes when synchronized output is disabled", async () => {
+		await withEnvPatch(
+			{ PI_NO_SYNC_OUTPUT: "1", PI_FORCE_SYNC_OUTPUT: undefined, PI_TUI_SYNC_OUTPUT: undefined },
+			async () => {
+				setTerminalDeccara(true);
+				const term = new VirtualTerminal(40, 8);
+				const tui = new TUI(term);
+				tui.addChild(new BgPanelComponent(["", "", "", ""]));
+				const writes = captureWrites(term);
+
+				try {
+					tui.start();
+					await settle(term);
+					const out = writes.join("");
+
+					expect(out).not.toContain("$r");
+					expect(out).not.toContain(DECSACE_RECT);
+					expect(out).toContain(`${BG_OPEN}${" ".repeat(40)}`);
+					expect(term.getViewportRowBackgroundColumns(0)).toHaveLength(40);
+				} finally {
+					tui.stop();
+				}
+			},
+		);
+	});
+
 	it("preserves viewport text identically whether DECCARA is on or off", async () => {
 		const rowsContent = ["", "Hello", "", "World", ""];
 		const trimmed = (term: VirtualTerminal) => term.getViewport().map(line => line.trimEnd());
diff --git a/packages/tui/test/image-budget.test.ts b/packages/tui/test/image-budget.test.ts
index 03d9cd9e9..57504afd6 100644
--- a/packages/tui/test/image-budget.test.ts
+++ b/packages/tui/test/image-budget.test.ts
@@ -259,7 +259,7 @@ describe("Image budget + Unicode placeholders", () => {
 		originalCellDims = { ...getCellDimensions() };
 		setCellDimensions({ widthPx: 10, heightPx: 10 });
 		terminal.imageProtocol = ImageProtocol.Kitty;
-		setKittyGraphics({ unicodePlaceholders: true, transmissionMedium: "direct" });
+		setKittyGraphics({ unicodePlaceholders: true });
 	});
 
 	afterEach(() => {
@@ -331,8 +331,10 @@ describe("TUI inline-image budget", () => {
 		setCellDimensions({ widthPx: 10, heightPx: 10 });
 		terminal.imageProtocol = ImageProtocol.Kitty;
 		monotonicNow = 0;
+		// Advance one full 30fps frame (>1000/30ms) per tick so the render
+		// throttle computes a zero delay and every requestRender flushes inline.
 		vi.spyOn(performance, "now").mockImplementation(() => {
-			monotonicNow += 20;
+			monotonicNow += 40;
 			return monotonicNow;
 		});
 	});
@@ -348,7 +350,7 @@ describe("TUI inline-image budget", () => {
 			const tick = Promise.withResolvers<void>();
 			process.nextTick(tick.resolve);
 			await tick.promise;
-			await Bun.sleep(1);
+			await Bun.sleep(40);
 			await term.flush();
 		}
 	}
diff --git a/packages/tui/test/image-render.test.ts b/packages/tui/test/image-render.test.ts
index ce6cd2ad9..4358373a8 100644
--- a/packages/tui/test/image-render.test.ts
+++ b/packages/tui/test/image-render.test.ts
@@ -69,6 +69,18 @@ describe("terminal image rendering", () => {
 		expect(parseKittyParam(result?.sequence ?? "", "r")).toBe(10);
 	});
 
+	it("transmits stable Kitty images in-band before placement", () => {
+		terminal.imageProtocol = ImageProtocol.Kitty;
+		const result = renderImage(BASE64_ONE_PIXEL_PNG, SQUARE_DIMENSIONS, {
+			imageId: 42,
+			includeTransmit: true,
+		});
+
+		expect(result).not.toBeNull();
+		expect(result?.transmit).toBe(`\x1b_Ga=t,f=100,q=2,i=42;${BASE64_ONE_PIXEL_PNG}\x1b\\`);
+		expect(result?.transmit).not.toContain("t=t");
+	});
+
 	it("reduces iTerm2 width when max height is the limiting bound", () => {
 		terminal.imageProtocol = ImageProtocol.Iterm2;
 		const result = renderImage(BASE64_DUMMY, SQUARE_DIMENSIONS, {
diff --git a/packages/tui/test/ime-jamo-cursor.test.ts b/packages/tui/test/ime-jamo-cursor.test.ts
deleted file mode 100644
index 665fa8aae..000000000
--- a/packages/tui/test/ime-jamo-cursor.test.ts
+++ /dev/null
@@ -1,109 +0,0 @@
-/**
- * Regression guard for Korean IME cursor positioning in the Input
- * component.
- *
- * Background: macOS Korean IME (2-bul keyboard) emits Hangul Compatibility
- * Jamo (U+3131..U+318E) during composition. Before commit 79e3170c6 the
- * Input.render() computed `cursorCols = visibleWidth(value.slice(0,
- * cursorIndex))` and `Bun.stringWidth` returned 2 for each jamo (per UAX
- * #11 EAW=W), while every macOS terminal renders them as 1 cell.
- *
- * The user-visible bug was a GROWING horizontal gap between the typed
- * jamo and the IME candidate window — every additional jamo doubled the
- * offset because `cursorCols` was N×2 instead of N×1. With 14 jamo typed,
- * the gap was ~14 cells.
- *
- * After the JS-side width correction, `cursorCols` is N×1. The Rust
- * `pi-natives` `sliceWithWidth` still treats jamo as 2 cells (binary
- * package; follow-up), so the cursor marker placement in `Input.render()`
- * has a residual ≤1-cell-per-jamo offset — but the user-visible "growing
- * gap" is gone because the JS-side `cursorCols` no longer doubles.
- *
- * What this test guards:
- *   - The cursor column for a value containing N jamo is **bounded above
- *     by `PROMPT_WIDTH + N`** (the correct cell count after the fix).
- *     Before the fix it would have been ~`PROMPT_WIDTH + 2N`, which is
- *     what the test catches.
- *   - Pure-ASCII and Hangul-syllable baselines are unchanged.
- *
- * What this test does NOT guard:
- *   - The Rust-side `sliceWithWidth` jamo discrepancy. Tracked as a
- *     follow-up; will tighten this test once the Rust crate is rebuilt.
- */
-import { describe, expect, it } from "bun:test";
-import { CURSOR_MARKER } from "@oh-my-pi/pi-tui";
-import { Input } from "@oh-my-pi/pi-tui/components/input";
-import { visibleWidth } from "@oh-my-pi/pi-tui/utils";
-
-/**
- * Drive `text` through `Input.handleInput()` one Unicode code point at a
- * time (mirrors what the IME does — one code point per emitted sequence),
- * then return the visual column where the hardware cursor marker lands
- * in the rendered output.
- */
-function cursorColAfterTyping(text: string, width = 80): number {
-	const input = new Input();
-	(input as unknown as { focused: boolean }).focused = true;
-	for (const char of text) {
-		input.handleInput(char);
-	}
-	const [line] = input.render(width);
-	const markerIdx = line.indexOf(CURSOR_MARKER);
-	if (markerIdx < 0) {
-		throw new Error(`CURSOR_MARKER not found in rendered line: ${JSON.stringify(line)}`);
-	}
-	return visibleWidth(line.slice(0, markerIdx));
-}
-
-const PROMPT_WIDTH = 2; // "> "
-
-// The jamo-cursor regression (PR #1410 / origin issue) only applies on
-// macOS, where terminals render Hangul Compatibility Jamo as 1 cell while
-// UAX#11 (and `Bun.stringWidth`) report 2. Off-darwin both the terminal and
-// the width helper agree on 2, so the doubling regression cannot occur.
-const IS_DARWIN = process.platform === "darwin";
-
-describe("Input cursor column does not grow at 2× per jamo", () => {
-	it("ASCII baseline: cursor lands exactly after the typed text", () => {
-		expect(cursorColAfterTyping("hello")).toBe(PROMPT_WIDTH + 5);
-	});
-
-	it("Hangul syllables: cursor lands exactly after typed text (2 cells each)", () => {
-		expect(cursorColAfterTyping("안녕")).toBe(PROMPT_WIDTH + 4);
-	});
-
-	it.skipIf(!IS_DARWIN)("single jamo: cursor column is at most `PROMPT_WIDTH + 1`", () => {
-		// Before fix: PROMPT_WIDTH + 2 = 4. After fix: ≤ 3.
-		expect(cursorColAfterTyping("ㅁ")).toBeLessThanOrEqual(PROMPT_WIDTH + 1);
-	});
-
-	it.skipIf(!IS_DARWIN)("8 consecutive jamo: cursor column is at most `PROMPT_WIDTH + 8`", () => {
-		// Before fix: PROMPT_WIDTH + 16 = 18. After fix: ≤ 10.
-		expect(cursorColAfterTyping("ㅁㄴㅁㄴㅇㅂㄴㅂ")).toBeLessThanOrEqual(PROMPT_WIDTH + 8);
-	});
-
-	it.skipIf(!IS_DARWIN)("20 consecutive jamo: cursor column is at most `PROMPT_WIDTH + 20`", () => {
-		// Before fix: PROMPT_WIDTH + 40 = 42 (catastrophic gap). After fix: ≤ 22.
-		const jamo = "ㅁㄴㄷㅂㅈㅎㅋㅌㄱㄹ".repeat(2);
-		expect(cursorColAfterTyping(jamo)).toBeLessThanOrEqual(PROMPT_WIDTH + 20);
-	});
-
-	it.skipIf(!IS_DARWIN)("cursor column grows by ≤1 per typed jamo (not 2)", () => {
-		// The regression: each typed jamo would advance cursor by 2 columns
-		// instead of 1, doubling the offset every keystroke. Assert the
-		// per-step delta never exceeds 1.
-		const input = new Input();
-		(input as unknown as { focused: boolean }).focused = true;
-		const jamo = "ㅁㄴㅇㅂㅈㅎㅋㅌㄷㄹ";
-		let prevCol = PROMPT_WIDTH;
-		for (let i = 0; i < jamo.length; i++) {
-			input.handleInput(jamo[i]);
-			const [line] = input.render(80);
-			const markerIdx = line.indexOf(CURSOR_MARKER);
-			const col = visibleWidth(line.slice(0, markerIdx));
-			const delta = col - prevCol;
-			expect(delta).toBeLessThanOrEqual(1);
-			prevCol = col;
-		}
-	});
-});
diff --git a/packages/tui/test/issue-1610-repro.test.ts b/packages/tui/test/issue-1610-repro.test.ts
index a73b3bf23..5a7b38ad2 100644
--- a/packages/tui/test/issue-1610-repro.test.ts
+++ b/packages/tui/test/issue-1610-repro.test.ts
@@ -36,7 +36,7 @@ async function settle(term: VirtualTerminal): Promise<void> {
 	const nextTick = Promise.withResolvers<void>();
 	process.nextTick(nextTick.resolve);
 	await nextTick.promise;
-	await Bun.sleep(20);
+	await Bun.sleep(40);
 	await term.flush();
 }
 
diff --git a/packages/tui/test/issue-1635-repro.test.ts b/packages/tui/test/issue-1635-repro.test.ts
index d455fc9de..ef196d2d7 100644
--- a/packages/tui/test/issue-1635-repro.test.ts
+++ b/packages/tui/test/issue-1635-repro.test.ts
@@ -49,7 +49,7 @@ async function settle(term: VirtualTerminal): Promise<void> {
 	const nextTick = Promise.withResolvers<void>();
 	process.nextTick(nextTick.resolve);
 	await nextTick.promise;
-	await Bun.sleep(20);
+	await Bun.sleep(40);
 	await term.flush();
 }
 
diff --git a/packages/tui/test/issue-1682-repro.test.ts b/packages/tui/test/issue-1682-repro.test.ts
index dad0124e2..84780790d 100644
--- a/packages/tui/test/issue-1682-repro.test.ts
+++ b/packages/tui/test/issue-1682-repro.test.ts
@@ -55,7 +55,7 @@ async function settle(term: VirtualTerminal): Promise<void> {
 	const nextTick = Promise.withResolvers<void>();
 	process.nextTick(nextTick.resolve);
 	await nextTick.promise;
-	await Bun.sleep(20);
+	await Bun.sleep(40);
 	await term.flush();
 }
 
@@ -229,7 +229,7 @@ describe("issue #1682: TUI eager scrollback rebuild", () => {
 		});
 	});
 
-	it("treats focused keyboard input as a user-input opt-in after an ED3-risk shrink defers", async () => {
+	it("treats focused keyboard input as a non-destructive repaint after an ED3-risk shrink defers", async () => {
 		await withEnvPatch(CLEAR_MULTIPLEXER_ENV, async () => {
 			await withTerminalRisk(true, async () => {
 				const term = new VirtualTerminal(40, 10);
@@ -258,7 +258,7 @@ describe("issue #1682: TUI eager scrollback rebuild", () => {
 					await settle(term);
 
 					expect(term.getViewport().map(line => line.trim())).toContain("prompt> x");
-					expect(eraseScrollbackCount(writes)).toBe(1);
+					expect(eraseScrollbackCount(writes)).toBe(0);
 					expect(tui.refreshNativeScrollbackIfDirty({ allowUnknownViewport: true })).toBe(false);
 				} finally {
 					tui.stop();
@@ -267,6 +267,45 @@ describe("issue #1682: TUI eager scrollback rebuild", () => {
 		});
 	});
 
+	it("preserves focused-input dirty scrollback rebuilds on non-ED3-risk terminals", async () => {
+		await withEnvPatch(CLEAR_MULTIPLEXER_ENV, async () => {
+			await withTerminalRisk(false, async () => {
+				const term = new VirtualTerminal(40, 6);
+				overrideProbe(term, false);
+				const tui = new TUI(term);
+				const transcript = new LineList(Array.from({ length: 12 }, (_value, index) => `init-${index}`));
+				const prompt = new PromptInput();
+				tui.addChild(transcript);
+				tui.addChild(prompt);
+				tui.setFocus(prompt);
+
+				try {
+					tui.start();
+					await settle(term);
+					const writes = capture(term);
+
+					transcript.setLines([
+						"init-0 edited",
+						...Array.from({ length: 11 }, (_value, index) => `init-${index + 1}`),
+					]);
+					tui.requestRender();
+					await settle(term);
+
+					expect(eraseScrollbackCount(writes)).toBe(0);
+					overrideProbe(term, undefined);
+
+					term.sendInput("x");
+					await settle(term);
+
+					expect(term.getViewport().map(line => line.trim())).toContain("prompt> x");
+					expect(eraseScrollbackCount(writes)).toBe(1);
+				} finally {
+					tui.stop();
+				}
+			});
+		});
+	});
+
 	it("keeps eager live rebuilds for other terminal traits", async () => {
 		await withEnvPatch(CLEAR_MULTIPLEXER_ENV, async () => {
 			await withTerminalRisk(false, async () => {
diff --git a/packages/tui/test/issue-1746-repro.test.ts b/packages/tui/test/issue-1746-repro.test.ts
index b6a412e9e..8616f172f 100644
--- a/packages/tui/test/issue-1746-repro.test.ts
+++ b/packages/tui/test/issue-1746-repro.test.ts
@@ -49,7 +49,7 @@ async function settle(term: VirtualTerminal): Promise<void> {
 	const nextTick = Promise.withResolvers<void>();
 	process.nextTick(nextTick.resolve);
 	await nextTick.promise;
-	await Bun.sleep(20);
+	await Bun.sleep(40);
 	await term.flush();
 }
 
diff --git a/packages/tui/test/issue-1765-repro.test.ts b/packages/tui/test/issue-1765-repro.test.ts
index af64666d3..df73fda8c 100644
--- a/packages/tui/test/issue-1765-repro.test.ts
+++ b/packages/tui/test/issue-1765-repro.test.ts
@@ -33,6 +33,21 @@ class FocusedLine implements Component, Focusable {
 	}
 }
 
+// VirtualTerminal does not model DECRQM capability probing, so subclass it to
+// register and replay the renderer's mode-2026 report callback on demand. This
+// exercises the runtime probe path in `TUI.start()` end-to-end.
+class ProbingTerminal extends VirtualTerminal {
+	#privateModeCallbacks: Array<(mode: number, supported: boolean) => void> = [];
+
+	onPrivateModeReport(callback: (mode: number, supported: boolean) => void): void {
+		this.#privateModeCallbacks.push(callback);
+	}
+
+	emitPrivateModeReport(mode: number, supported: boolean): void {
+		for (const callback of this.#privateModeCallbacks) callback(mode, supported);
+	}
+}
+
 const SYNC_BEGIN = "\x1b[?2026h";
 const SYNC_END = "\x1b[?2026l";
 const DISABLE_AUTOWRAP = "\x1b[?7l";
@@ -188,3 +203,148 @@ describe("issue #1765: synchronized-output opt-out", () => {
 		});
 	});
 });
+
+describe("synchronized-output runtime DECRQM probe", () => {
+	it("enables synchronized output after a positive DEC 2026 report on a default-off host", async () => {
+		// TMUX forces the static default off; the positive probe must upgrade it.
+		await withEnvPatch(
+			{
+				TMUX: "1",
+				WT_SESSION: undefined,
+				TERM_FEATURES: undefined,
+				PI_NO_SYNC_OUTPUT: undefined,
+				PI_FORCE_SYNC_OUTPUT: undefined,
+				PI_TUI_SYNC_OUTPUT: undefined,
+			},
+			async () => {
+				const term = new ProbingTerminal(32, 4, 100);
+				const writes = captureWrites(term);
+				const component = new MutableLines(["before probe"]);
+				const tui = new TUI(term);
+				tui.addChild(component);
+
+				try {
+					tui.start();
+					await term.waitForRender();
+					expect(tui.synchronizedOutput).toBe(false);
+					expectNoSyncOutput(writes);
+
+					const mark = writes.length;
+					term.emitPrivateModeReport(2026, true);
+					expect(tui.synchronizedOutput).toBe(true);
+
+					component.lines = ["after probe"];
+					tui.requestRender();
+					await term.waitForRender();
+
+					const after = writes.slice(mark).join("");
+					expect(after).toContain(SYNC_BEGIN);
+					expect(after).toContain(SYNC_END);
+				} finally {
+					tui.stop();
+				}
+			},
+		);
+	});
+
+	it("disables synchronized output after a negative DEC 2026 report on a default-on host", async () => {
+		// WT_SESSION forces the static default on without a user override flag.
+		await withEnvPatch(
+			{
+				WT_SESSION: "abc",
+				PI_NO_SYNC_OUTPUT: undefined,
+				PI_FORCE_SYNC_OUTPUT: undefined,
+				PI_TUI_SYNC_OUTPUT: undefined,
+			},
+			async () => {
+				const term = new ProbingTerminal(32, 4, 100);
+				const writes = captureWrites(term);
+				const component = new MutableLines(["before probe"]);
+				const tui = new TUI(term);
+				tui.addChild(component);
+
+				try {
+					tui.start();
+					await term.waitForRender();
+					expect(tui.synchronizedOutput).toBe(true);
+
+					const mark = writes.length;
+					term.emitPrivateModeReport(2026, false);
+					expect(tui.synchronizedOutput).toBe(false);
+
+					component.lines = ["after probe"];
+					tui.requestRender();
+					await term.waitForRender();
+
+					expectNoSyncOutput(writes.slice(mark));
+				} finally {
+					tui.stop();
+				}
+			},
+		);
+	});
+
+	it("ignores a positive probe when the user opted out", async () => {
+		await withEnvPatch(
+			{ PI_NO_SYNC_OUTPUT: "1", PI_FORCE_SYNC_OUTPUT: undefined, PI_TUI_SYNC_OUTPUT: undefined },
+			async () => {
+				const term = new ProbingTerminal(32, 4, 100);
+				const writes = captureWrites(term);
+				const component = new MutableLines(["before probe"]);
+				const tui = new TUI(term);
+				tui.addChild(component);
+
+				try {
+					tui.start();
+					await term.waitForRender();
+					expect(tui.synchronizedOutput).toBe(false);
+
+					const mark = writes.length;
+					term.emitPrivateModeReport(2026, true);
+					expect(tui.synchronizedOutput).toBe(false);
+
+					component.lines = ["after probe"];
+					tui.requestRender();
+					await term.waitForRender();
+
+					expectNoSyncOutput(writes.slice(mark));
+				} finally {
+					tui.stop();
+				}
+			},
+		);
+	});
+
+	it("ignores a negative probe when the user forced sync on", async () => {
+		await withEnvPatch(
+			{ PI_FORCE_SYNC_OUTPUT: "1", PI_NO_SYNC_OUTPUT: undefined, PI_TUI_SYNC_OUTPUT: undefined },
+			async () => {
+				const term = new ProbingTerminal(32, 4, 100);
+				const writes = captureWrites(term);
+				const component = new MutableLines(["before probe"]);
+				const tui = new TUI(term);
+				tui.addChild(component);
+
+				try {
+					tui.start();
+					await term.waitForRender();
+					expect(tui.synchronizedOutput).toBe(true);
+
+					const mark = writes.length;
+					term.emitPrivateModeReport(2026, false);
+					expect(tui.synchronizedOutput).toBe(true);
+
+					component.lines = ["after probe"];
+					tui.requestRender();
+					await term.waitForRender();
+
+					const after = writes.slice(mark).join("");
+					expect(after).toContain(SYNC_BEGIN);
+					expect(after).toContain(SYNC_END);
+				} finally {
+					tui.stop();
+				}
+			},
+		);
+	});
+});
diff --git a/packages/tui/test/issue-1962-repro.test.ts b/packages/tui/test/issue-1962-repro.test.ts
new file mode 100644
index 000000000..bbceb4ca4
--- /dev/null
+++ b/packages/tui/test/issue-1962-repro.test.ts
@@ -0,0 +1,168 @@
+import { afterEach, describe, expect, it, vi } from "bun:test";
+import { type Component, type Focusable, TERMINAL, TUI } from "@oh-my-pi/pi-tui";
+import { VirtualTerminal } from "./virtual-terminal";
+
+class MutableLinesComponent implements Component {
+	#lines: string[];
+
+	constructor(lines: string[]) {
+		this.#lines = [...lines];
+	}
+
+	setLines(lines: string[]): void {
+		this.#lines = [...lines];
+	}
+
+	invalidate(): void {}
+
+	render(width: number): string[] {
+		return this.#lines.map(line => line.slice(0, width));
+	}
+}
+
+class ArrowSelectorComponent implements Component, Focusable {
+	focused = true;
+	#selectedIndex = 0;
+
+	handleInput(data: string): void {
+		if (data === "\x1b[B") this.#selectedIndex = 1;
+		if (data === "\x1b[A") this.#selectedIndex = 0;
+	}
+
+	invalidate(): void {}
+
+	render(): string[] {
+		return [this.#selectedIndex === 0 ? "> first" : "  first", this.#selectedIndex === 1 ? "> second" : "  second"];
+	}
+}
+
+class UnknownViewportTerminal extends VirtualTerminal {
+	isNativeViewportAtBottom(): undefined {
+		return undefined;
+	}
+}
+
+type MutableTerminalInfo = {
+	eagerEraseScrollbackRisk: boolean;
+};
+
+const mutableTerminalInfo = TERMINAL as unknown as MutableTerminalInfo;
+const ERASE_SCROLLBACK = /\x1b\[3J/g;
+
+async function settle(term: VirtualTerminal): Promise<void> {
+	const nextTick = Promise.withResolvers<void>();
+	process.nextTick(nextTick.resolve);
+	await nextTick.promise;
+	await Bun.sleep(40);
+	await term.flush();
+}
+
+function captureWrites(term: VirtualTerminal): string[] {
+	const writes: string[] = [];
+	const realWrite = term.write.bind(term);
+	vi.spyOn(term, "write").mockImplementation((data: string) => {
+		writes.push(data);
+		realWrite(data);
+	});
+	return writes;
+}
+
+async function withTerminalRisk<T>(risk: boolean, run: () => T | Promise<T>): Promise<T> {
+	const saved = TERMINAL.eagerEraseScrollbackRisk;
+	mutableTerminalInfo.eagerEraseScrollbackRisk = risk;
+	try {
+		return await run();
+	} finally {
+		mutableTerminalInfo.eagerEraseScrollbackRisk = saved;
+	}
+}
+
+describe("issue #1962: arrow navigation after dirty scrollback", () => {
+	afterEach(() => {
+		vi.restoreAllMocks();
+	});
+
+	it("does not clear and replay the whole transcript for a focused arrow-key frame", async () => {
+		await withTerminalRisk(true, async () => {
+			const term = new UnknownViewportTerminal(40, 6);
+			const tui = new TUI(term);
+			const transcript = new MutableLinesComponent(
+				Array.from({ length: 12 }, (_value, index) => `history-${index}`),
+			);
+			const selector = new ArrowSelectorComponent();
+			tui.addChild(transcript);
+			tui.addChild(selector);
+			tui.setFocus(selector);
+
+			try {
+				tui.start();
+				await settle(term);
+
+				tui.setEagerNativeScrollbackRebuild(true);
+				transcript.setLines([
+					"history-0 updated",
+					...Array.from({ length: 11 }, (_value, index) => `history-${index + 1}`),
+				]);
+				tui.requestRender();
+				await settle(term);
+				tui.setEagerNativeScrollbackRebuild(false);
+
+				const writes = captureWrites(term);
+				term.sendInput("\x1b[B");
+				await settle(term);
+
+				const output = writes.join("");
+				expect(output.match(ERASE_SCROLLBACK) ?? []).toHaveLength(0);
+				expect(output).not.toContain("history-0 updated");
+				expect(term.getViewport().map(line => line.trimEnd())).toEqual([
+					"history-8",
+					"history-9",
+					"history-10",
+					"history-11",
+					"  first",
+					"> second",
+				]);
+			} finally {
+				tui.stop();
+			}
+		});
+	});
+
+	it("does not clear and replay the whole transcript for a focused arrow-key frame inside an overlay", async () => {
+		await withTerminalRisk(true, async () => {
+			const term = new UnknownViewportTerminal(40, 6);
+			const tui = new TUI(term);
+			const transcript = new MutableLinesComponent(
+				Array.from({ length: 12 }, (_value, index) => `history-${index}`),
+			);
+			tui.addChild(transcript);
+			const selector = new ArrowSelectorComponent();
+			tui.showOverlay(selector);
+
+			try {
+				tui.start();
+				await settle(term);
+
+				tui.setEagerNativeScrollbackRebuild(true);
+				transcript.setLines([
+					"history-0 updated",
+					...Array.from({ length: 11 }, (_value, index) => `history-${index + 1}`),
+				]);
+				tui.requestRender();
+				await settle(term);
+				tui.setEagerNativeScrollbackRebuild(false);
+
+				const writes = captureWrites(term);
+				term.sendInput("\x1b[B");
+				await settle(term);
+
+				const output = writes.join("");
+				expect(output.match(ERASE_SCROLLBACK) ?? []).toHaveLength(0);
+				expect(output).not.toContain("history-0 updated");
+				expect(term.getViewport().map(line => line.trimEnd())).toContain("> second");
+			} finally {
+				tui.stop();
+			}
+		});
+	});
+});
diff --git a/packages/tui/test/issue-1974-repro.test.ts b/packages/tui/test/issue-1974-repro.test.ts
new file mode 100644
index 000000000..43aea302b
--- /dev/null
+++ b/packages/tui/test/issue-1974-repro.test.ts
@@ -0,0 +1,362 @@
+import { describe, expect, it } from "bun:test";
+import { type Component, CURSOR_MARKER, type NativeScrollbackLiveRegion, TERMINAL, TUI } from "@oh-my-pi/pi-tui";
+import { VirtualTerminal } from "./virtual-terminal";
+
+// Regression test for https://github.com/can1357/oh-my-pi/issues/1974
+//
+// Inside tmux (and other multiplexers), a long streamed reply that grows past
+// the viewport lost its scrolled-off head from pane history and, after a
+// later viewport repaint, the same content reappeared inside the visible
+// pane while the original streamed rows were stranded above — leaving
+// "missing sections" interleaved with "repeating chunks" when the user
+// scrolled back through the tmux pane buffer. The renderer's foreground-
+// streaming cap-to-viewport branch clipped `lines` to the visible tail and
+// reset `#scrollbackHighWater` to 0 for every streaming frame, so no rows
+// ever entered tmux pane history while the assistant reply was active.
+//
+// The `liveRegionPinned` intent already knows how to push the sealed prefix
+// of an append-only live block into native scrollback via `\r\n` without
+// emitting ED3 — exactly what tmux can accept — but it used to short-circuit
+// inside multiplexers. Enabling it (and skipping the cap when the planner
+// picks it) commits the assistant reply's head into pane history exactly
+// once while the live tail keeps repainting in place.
+
+class LineList implements Component {
+	#lines: string[];
+
+	constructor(lines: string[]) {
+		this.#lines = [...lines];
+	}
+
+	invalidate(): void {}
+
+	render(width: number): string[] {
+		return this.#lines.map(line => line.slice(0, width));
+	}
+
+	setLines(lines: string[]): void {
+		this.#lines = [...lines];
+	}
+}
+
+/**
+ * Minimal append-only live region. Models the real omp setup where
+ * `TranscriptContainer` wraps an `AssistantMessageComponent` that reports
+ * itself as `isTranscriptBlockAppendOnly() === true`.
+ */
+class StreamingLiveRegion implements Component, NativeScrollbackLiveRegion {
+	#lines: string[];
+
+	constructor(lines: string[]) {
+		this.#lines = [...lines];
+	}
+
+	invalidate(): void {}
+
+	render(width: number): string[] {
+		return this.#lines.map(line => line.slice(0, width));
+	}
+
+	setLines(lines: string[]): void {
+		this.#lines = [...lines];
+	}
+
+	getNativeScrollbackLiveRegionStart(): number | undefined {
+		return 0;
+	}
+
+	getNativeScrollbackCommitSafeEnd(): number | undefined {
+		return this.#lines.length;
+	}
+}
+
+class VolatileLiveRegion implements Component, NativeScrollbackLiveRegion {
+	#lines: string[];
+
+	constructor(lines: string[]) {
+		this.#lines = [...lines];
+	}
+
+	invalidate(): void {}
+
+	render(width: number): string[] {
+		return this.#lines.map(line => line.slice(0, width));
+	}
+
+	setLines(lines: string[]): void {
+		this.#lines = [...lines];
+	}
+
+	getNativeScrollbackLiveRegionStart(): number | undefined {
+		return 0;
+	}
+}
+
+async function settle(term: VirtualTerminal): Promise<void> {
+	const nextTick = Promise.withResolvers<void>();
+	process.nextTick(nextTick.resolve);
+	await nextTick.promise;
+	await Bun.sleep(40);
+	await term.flush();
+}
+
+async function withEnvPatch<T>(patch: Record<string, string | undefined>, run: () => T | Promise<T>): Promise<T> {
+	const saved: Record<string, string | undefined> = {};
+	for (const key in patch) {
+		saved[key] = Bun.env[key];
+		const value = patch[key];
+		if (value === undefined) {
+			delete Bun.env[key];
+		} else {
+			Bun.env[key] = value;
+		}
+	}
+	try {
+		return await run();
+	} finally {
+		for (const key in saved) {
+			const value = saved[key];
+			if (value === undefined) {
+				delete Bun.env[key];
+			} else {
+				Bun.env[key] = value;
+			}
+		}
+	}
+}
+
+type MutableTerminalInfo = { eagerEraseScrollbackRisk: boolean };
+
+async function withTerminalRisk<T>(risk: boolean, run: () => T | Promise<T>): Promise<T> {
+	const mutable = TERMINAL as unknown as MutableTerminalInfo;
+	const saved = mutable.eagerEraseScrollbackRisk;
+	mutable.eagerEraseScrollbackRisk = risk;
+	try {
+		return await run();
+	} finally {
+		mutable.eagerEraseScrollbackRisk = saved;
+	}
+}
+
+function overrideProbe(term: VirtualTerminal, answer: boolean | undefined): void {
+	(term as unknown as { isNativeViewportAtBottom: () => boolean | undefined }).isNativeViewportAtBottom = () => answer;
+}
+
+function strip(rows: string[]): string[] {
+	return rows.map(row => Bun.stripANSI(row).trimEnd());
+}
+
+const TMUX_ENV: Record<string, string | undefined> = {
+	TMUX: "1",
+	STY: undefined,
+	ZELLIJ: undefined,
+};
+
+const ERASE_SCROLLBACK = /\x1b\[3J/g;
+
+function capture(term: VirtualTerminal): string[] {
+	const writes: string[] = [];
+	const realWrite = term.write.bind(term);
+	(term as unknown as { write: (s: string) => void }).write = (data: string) => {
+		writes.push(data);
+		realWrite(data);
+	};
+	return writes;
+}
+
+function occurrencesOf(haystack: string, needle: string): number {
+	return haystack.split(needle).length - 1;
+}
+
+describe("issue #1974: tmux scrollback rendering", () => {
+	it("commits a streaming reply's scrolled-off head to pane history exactly once", async () => {
+		if (process.platform === "win32") return;
+
+		await withEnvPatch(TMUX_ENV, async () => {
+			await withTerminalRisk(true, async () => {
+				const term = new VirtualTerminal(80, 8, 10_000);
+				// Real tmux/ProcessTerminal does not implement
+				// `isNativeViewportAtBottom`, so the renderer sees `undefined`
+				// in production. Match that here.
+				overrideProbe(term, undefined);
+
+				const tui = new TUI(term);
+				const stream = new StreamingLiveRegion([]);
+				tui.addChild(stream);
+
+				const markers = Array.from({ length: 40 }, (_unused, i) => `MARK-${String(i).padStart(3, "0")}`);
+
+				try {
+					tui.start();
+					tui.setEagerNativeScrollbackRebuild(true);
+					await settle(term);
+
+					// Stream the reply in chunks. Each chunk grows the live block
+					// by 5 rows — small enough that no single frame double-
+					// overflows the 8-row viewport, large enough that the head
+					// must scroll into pane history between frames.
+					for (let chunk = 5; chunk <= markers.length; chunk += 5) {
+						stream.setLines(markers.slice(0, chunk));
+						tui.requestRender();
+						await settle(term);
+					}
+
+					// `getScrollBuffer()` returns pane history + the active grid
+					// (i.e. what tmux would show when the user scrolled all the
+					// way up). Each MARK-NNN must appear in that combined buffer
+					// exactly once — no gaps ("missing sections") and no
+					// duplicates ("repeating chunks").
+					const scrollback = strip(term.getScrollBuffer());
+					const buffer = scrollback.join("\n");
+					const missing: string[] = [];
+					const duplicated: string[] = [];
+					for (const mark of markers) {
+						const occ = occurrencesOf(buffer, mark);
+						if (occ === 0) missing.push(mark);
+						if (occ > 1) duplicated.push(mark);
+					}
+					expect(missing).toEqual([]);
+					expect(duplicated).toEqual([]);
+
+					// The visible viewport still shows the live tail.
+					const viewport = strip(term.getViewport());
+					expect(viewport.some(row => row.includes("MARK-039"))).toBe(true);
+				} finally {
+					tui.stop();
+					await term.flush();
+				}
+			});
+		});
+	});
+
+	it("never emits ED3 (CSI 3 J) inside a tmux pane during streaming", async () => {
+		if (process.platform === "win32") return;
+
+		await withEnvPatch(TMUX_ENV, async () => {
+			await withTerminalRisk(true, async () => {
+				const term = new VirtualTerminal(80, 8, 10_000);
+				overrideProbe(term, undefined);
+				const tui = new TUI(term);
+				const stream = new StreamingLiveRegion([]);
+				tui.addChild(stream);
+
+				try {
+					tui.start();
+					tui.setEagerNativeScrollbackRebuild(true);
+					await settle(term);
+
+					const writes = capture(term);
+					for (let chunk = 5; chunk <= 40; chunk += 5) {
+						stream.setLines(Array.from({ length: chunk }, (_unused, i) => `row-${String(i).padStart(3, "0")}`));
+						tui.requestRender();
+						await settle(term);
+					}
+
+					// ED3 would either be a no-op or yank a scrolled tmux reader.
+					// The tmux path must commit incrementally via \r\n.
+					expect(writes.join("").match(ERASE_SCROLLBACK)?.length ?? 0).toBe(0);
+				} finally {
+					tui.stop();
+					await term.flush();
+				}
+			});
+		});
+	});
+
+	it("does not push chrome below the live block into pane history", async () => {
+		// Models a real omp frame: streamed assistant reply on top, persistent
+		// chrome (status line / editor) below. The live region ends mid-frame,
+		// so the renderer must push only sealed rows of the live block into
+		// pane history while keeping the chrome rows transient and confined to
+		// the visible pane.
+		if (process.platform === "win32") return;
+
+		await withEnvPatch(TMUX_ENV, async () => {
+			await withTerminalRisk(true, async () => {
+				const term = new VirtualTerminal(80, 10, 10_000);
+				overrideProbe(term, undefined);
+				const tui = new TUI(term);
+				const stream = new StreamingLiveRegion([]);
+				const footer = new LineList(["── prompt ──", "> "]);
+				tui.addChild(stream);
+				tui.addChild(footer);
+
+				const markers = Array.from({ length: 30 }, (_unused, i) => `STREAM-${String(i).padStart(3, "0")}`);
+
+				try {
+					tui.start();
+					tui.setEagerNativeScrollbackRebuild(true);
+					await settle(term);
+
+					for (let chunk = 4; chunk <= markers.length; chunk += 4) {
+						stream.setLines(markers.slice(0, chunk));
+						tui.requestRender();
+						await settle(term);
+					}
+
+					// `getScrollBuffer()` returns pane history followed by the
+					// active grid (the visible viewport). For "what is in pane
+					// history alone", chop off the last `rows` entries.
+					const fullBuffer = strip(term.getScrollBuffer());
+					const viewport = strip(term.getViewport());
+					const history = fullBuffer.slice(0, Math.max(0, fullBuffer.length - viewport.length));
+
+					// Chrome must NEVER enter pane history (it sits below the live
+					// region and never sealed).
+					expect(history.some(row => row.includes("── prompt ──"))).toBe(false);
+					expect(history.some(row => row.includes("> "))).toBe(false);
+					// Chrome stays in the visible viewport.
+					expect(viewport.some(row => row.includes("── prompt ──"))).toBe(true);
+
+					// No streamed row appears twice across pane history.
+					const historyText = history.join("\n");
+					const duplicated = markers.filter(m => occurrencesOf(historyText, m) > 1);
+					expect(duplicated).toEqual([]);
+
+					// The pane-history slice runs in original streaming order so
+					// a tmux scroll-back is monotonic.
+					const historyMarks = history
+						.map(row => row.match(/STREAM-\d{3}/)?.[0] ?? null)
+						.filter((m): m is string => m !== null);
+					expect(historyMarks).toEqual(markers.slice(0, historyMarks.length));
+				} finally {
+					tui.stop();
+					await term.flush();
+				}
+			});
+		});
+	});
+
+	it("keeps the cursor anchored when a no-append live repaint shifts the viewport", async () => {
+		if (process.platform === "win32") return;
+
+		await withTerminalRisk(true, async () => {
+			const term = new VirtualTerminal(20, 5, 1_000);
+			overrideProbe(term, undefined);
+			const tui = new TUI(term, true);
+			const stream = new VolatileLiveRegion([]);
+			tui.addChild(stream);
+
+			try {
+				tui.start();
+				tui.setEagerNativeScrollbackRebuild(true);
+				await settle(term);
+
+				stream.setLines(["same", "same", "same", `same${CURSOR_MARKER}`, "same"]);
+				tui.requestRender();
+				await settle(term);
+				expect(term.getCursor()).toEqual({ row: 3, col: 4 });
+
+				stream.setLines(["same", "same", "same", "same", `same${CURSOR_MARKER}`, "same"]);
+				tui.requestRender();
+				await settle(term);
+
+				expect(strip(term.getViewport())).toEqual(["same", "same", "same", "same", "same"]);
+				expect(term.getCursor()).toEqual({ row: 3, col: 4 });
+			} finally {
+				tui.stop();
+				await term.flush();
+			}
+		});
+	});
+});
diff --git a/packages/tui/test/issue-643-repro.test.ts b/packages/tui/test/issue-643-repro.test.ts
deleted file mode 100644
index 1c125fd86..000000000
--- a/packages/tui/test/issue-643-repro.test.ts
+++ /dev/null
@@ -1,122 +0,0 @@
-import { describe, expect, it } from "bun:test";
-import { type Component, TUI, visibleWidth } from "@oh-my-pi/pi-tui";
-import { VirtualTerminal } from "./virtual-terminal";
-
-// Regression test for https://github.com/can1357/oh-my-pi/issues/643
-//
-// Arabic text with tashkeel (nonspacing diacritics, Unicode category Mn) was
-// over-measured: each mark counted as 1 column instead of 0. A line of Quranic
-// text with N marks therefore reported a width N columns larger than what the
-// terminal actually renders, and the renderer crashed with "Rendered line
-// exceeds terminal width" (13.19.0) once enough marks accumulated.
-//
-// Root cause: `Bun.stringWidth` counts Mn marks as 1 column (still true as of
-// Bun 1.3.x). The fix routes every non-ASCII line through the native width
-// engine (Rust `unicode-width`), which measures nonspacing marks as 0 — in
-// agreement with xterm.js (BMP_COMBINING tables) and real terminals.
-//
-// The contract defended here: tashkeel marks contribute zero columns, so a
-// marked-up line whose letter width exactly equals the terminal width renders
-// unclipped on a single row. Over-measurement would either truncate trailing
-// letters (today's fitting behavior) or crash (the original report).
-
-// Tashkeel samples from the issue, with their correct letter-cell widths.
-const TASHKEEL_SAMPLES: ReadonlyArray<readonly [text: string, width: number]> = [
-	["بِسْمِ", 3], // 3 letters + 3 marks
-	["سَابِقُوا", 6], // 6 letters + 3 marks
-	["فَتَبَيَّنُوا", 7], // 7 letters + 5 marks (incl. shadda)
-];
-
-class RawLinesComponent implements Component {
-	#lines: string[];
-
-	constructor(lines: string[]) {
-		this.#lines = [...lines];
-	}
-
-	setLines(lines: string[]): void {
-		this.#lines = [...lines];
-	}
-
-	invalidate(): void {}
-
-	render(): string[] {
-		return [...this.#lines];
-	}
-}
-
-async function settle(term: VirtualTerminal): Promise<void> {
-	const nextTick = Promise.withResolvers<void>();
-	process.nextTick(nextTick.resolve);
-	await nextTick.promise;
-	await Bun.sleep(20);
-	await term.flush();
-}
-
-describe("issue #643: Arabic tashkeel width measurement", () => {
-	it("measures nonspacing tashkeel marks as zero columns", () => {
-		for (const [text, width] of TASHKEEL_SAMPLES) {
-			expect(visibleWidth(text)).toBe(width);
-		}
-		// Mixed ASCII + tashkeel stays on the non-ASCII measurement path.
-		expect(visibleWidth(`ayah: ${TASHKEEL_SAMPLES[1][0]}`)).toBe(6 + TASHKEEL_SAMPLES[1][1]);
-	});
-
-	it("renders a full-width tashkeel line unclipped on a single terminal row", async () => {
-		// "quran: " (7 cells) + 6-cell word = 13 cells — an exact fit at width 13.
-		// The original bug measured the word as 9 (6 letters + 3 marks), making the
-		// line appear 3 columns too wide, so the renderer's width fitting would
-		// truncate real trailing letters before writing.
-		const word = TASHKEEL_SAMPLES[1][0];
-		const line = `quran: ${word}`;
-		const width = 13;
-		expect(visibleWidth(line)).toBe(width);
-
-		const term = new VirtualTerminal(width, 6);
-		const tui = new TUI(term);
-		const component = new RawLinesComponent(["header", line, "tail"]);
-		tui.addChild(component);
-
-		try {
-			tui.start();
-			await settle(term);
-
-			const viewport = term.getViewport().map(row => row.trimEnd());
-			// The full marked-up text survives the round trip: no truncation by the
-			// renderer, no clipping or wrapping by the terminal.
-			expect(viewport[1]).toBe(line);
-			expect(viewport[0]).toBe("header");
-			expect(viewport[2]).toBe("tail");
-			// Exactly one terminal row per logical row — the marks did not push the
-			// line across the right margin.
-			expect(term.getScrollBuffer().length).toBe(6);
-		} finally {
-			tui.stop();
-		}
-	});
-
-	it("keeps row accounting exact when tashkeel rows scroll into native history", async () => {
-		const word = TASHKEEL_SAMPLES[2][0];
-		const width = 24;
-		const height = 5;
-		const term = new VirtualTerminal(width, height);
-		const tui = new TUI(term);
-		const tashkeelRows = Array.from({ length: 8 }, (_v, i) => `ayah-${i} ${word}`);
-		const component = new RawLinesComponent(tashkeelRows);
-		tui.addChild(component);
-
-		try {
-			tui.start();
-			await settle(term);
-
-			// 8 content rows at height 5 → 3 rows pushed into scrollback.
-			const buffer = term.getScrollBuffer().map(row => row.trimEnd());
-			expect(buffer.length).toBe(tashkeelRows.length);
-			for (let i = 0; i < tashkeelRows.length; i++) {
-				expect(buffer[i]).toBe(tashkeelRows[i]);
-			}
-		} finally {
-			tui.stop();
-		}
-	});
-});
diff --git a/packages/tui/test/keybindings.test.ts b/packages/tui/test/keybindings.test.ts
index a6d616c06..10a3d93b8 100644
--- a/packages/tui/test/keybindings.test.ts
+++ b/packages/tui/test/keybindings.test.ts
@@ -20,6 +20,15 @@ describe("KeybindingsManager", () => {
 		expect(keybindings.getKeys("tui.editor.cursorUp")).toEqual(["up"]);
 	});
 
+	it("preserves Shift when matching printable uppercase letters", () => {
+		const keybindings = new KeybindingsManager(TUI_KEYBINDINGS, {
+			"tui.input.copy": "shift+a",
+		});
+
+		expect(keybindings.matches("A", "tui.input.copy")).toBe(true);
+		expect(keybindings.matches("a", "tui.input.copy")).toBe(false);
+	});
+
 	it("still reports direct user binding conflicts without evicting defaults", () => {
 		const keybindings = new KeybindingsManager(TUI_KEYBINDINGS, {
 			"tui.input.submit": "ctrl+x",
diff --git a/packages/tui/test/kitty-graphics.test.ts b/packages/tui/test/kitty-graphics.test.ts
index b2e81fb6c..bab897e04 100644
--- a/packages/tui/test/kitty-graphics.test.ts
+++ b/packages/tui/test/kitty-graphics.test.ts
@@ -1,25 +1,17 @@
 import { afterEach, describe, expect, it } from "bun:test";
-import * as fs from "node:fs";
 import { visibleWidth } from "@oh-my-pi/pi-natives";
 import {
 	detectKittyUnicodePlaceholdersSupport,
 	encodeKittyPlaceholderGrid,
-	encodeKittyTempFileProbe,
-	encodeKittyTempFileTransmit,
 	encodeKittyVirtualPlacement,
 	getKittyGraphics,
-	isPngBase64,
 	KITTY_PLACEHOLDER,
 	KITTY_PLACEHOLDER_MAX_CELLS,
 	kittyPlaceholdersFit,
-	kittyTempFileAllowed,
 	renderKittyPlaceholderLines,
 	setKittyGraphics,
 } from "@oh-my-pi/pi-tui/kitty-graphics";
 
-const ONE_PIXEL_PNG =
-	"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8z8BQDwAEhQGAhKmMIQAAAABJRU5ErkJggg==";
-
 const ORIGINAL = { ...getKittyGraphics() };
 
 afterEach(() => {
@@ -80,81 +72,12 @@ describe("kitty Unicode placeholder encoding", () => {
 	});
 });
 
-describe("kitty temp-file transmission", () => {
-	it("isPngBase64 recognizes PNG payloads only", () => {
-		expect(isPngBase64(ONE_PIXEL_PNG)).toBe(true);
-		expect(isPngBase64("/9j/4AAQSkZJRgABAQ")).toBe(false); // JPEG magic
-		expect(isPngBase64("")).toBe(false);
-	});
-
-	it("encodeKittyTempFileTransmit writes the bytes to a magic-named file and sends its path", () => {
-		const seq = encodeKittyTempFileTransmit(ONE_PIXEL_PNG, 12);
-		expect(seq).not.toBeNull();
-		const match = seq!.match(/^\x1b_Ga=t,f=100,t=t,S=(\d+),q=2,i=12;([^\x1b]+)\x1b\\$/);
-		expect(match).not.toBeNull();
-		const [, sizeRaw, encodedPath] = match!;
-		const filePath = Buffer.from(encodedPath!, "base64").toString("utf8");
-		// Kitty only deletes temp files whose path contains this substring.
-		expect(filePath).toContain("tty-graphics-protocol");
-		try {
-			const written = fs.readFileSync(filePath);
-			expect(written.equals(Buffer.from(ONE_PIXEL_PNG, "base64"))).toBe(true);
-			expect(Number(sizeRaw)).toBe(written.length);
-		} finally {
-			fs.rmSync(filePath, { force: true });
-		}
-	});
-
-	it("encodeKittyTempFileProbe builds an a=q,t=t query and cleans up its file", () => {
-		const probe = encodeKittyTempFileProbe(99);
-		expect(probe).not.toBeNull();
-		expect(probe!.sequence).toContain("\x1b_Ga=q,t=t,f=100,");
-		expect(probe!.sequence).toContain("i=99;");
-		const encodedPath = probe!.sequence.match(/;([^\x1b]+)\x1b\\$/)?.[1];
-		const filePath = Buffer.from(encodedPath!, "base64").toString("utf8");
-		expect(fs.existsSync(filePath)).toBe(true);
-		probe!.cleanup();
-		expect(fs.existsSync(filePath)).toBe(false);
-	});
-});
-
 describe("kitty graphics feature state", () => {
 	it("getKittyGraphics/setKittyGraphics round-trips overrides", () => {
-		setKittyGraphics({ unicodePlaceholders: false, transmissionMedium: "temp-file" });
-		expect(getKittyGraphics()).toEqual({ unicodePlaceholders: false, transmissionMedium: "temp-file" });
+		setKittyGraphics({ unicodePlaceholders: false });
+		expect(getKittyGraphics()).toEqual({ unicodePlaceholders: false });
 		setKittyGraphics({ unicodePlaceholders: true });
 		expect(getKittyGraphics().unicodePlaceholders).toBe(true);
-		expect(getKittyGraphics().transmissionMedium).toBe("temp-file");
-	});
-
-	it("kittyTempFileAllowed honors env force/off and the remote-session guard", () => {
-		const keys = ["PI_KITTY_IMAGE_TRANSMISSION", "SSH_CONNECTION", "SSH_CLIENT", "SSH_TTY"];
-		const saved: Record<string, string | undefined> = {};
-		for (const k of keys) saved[k] = Bun.env[k];
-		const restore = () => {
-			for (const k of keys) {
-				const v = saved[k];
-				if (v === undefined) delete Bun.env[k];
-				else Bun.env[k] = v;
-			}
-		};
-		try {
-			for (const k of keys) delete Bun.env[k];
-			// Auto + local session: allowed.
-			expect(kittyTempFileAllowed()).toBe(true);
-			// Auto + SSH session: not allowed.
-			Bun.env.SSH_CONNECTION = "1.2.3.4 5 6.7.8.9 22";
-			expect(kittyTempFileAllowed()).toBe(false);
-			// Explicit force overrides the remote guard.
-			Bun.env.PI_KITTY_IMAGE_TRANSMISSION = "temp-file";
-			expect(kittyTempFileAllowed()).toBe(true);
-			// Explicit off wins even on a local session.
-			delete Bun.env.SSH_CONNECTION;
-			Bun.env.PI_KITTY_IMAGE_TRANSMISSION = "direct";
-			expect(kittyTempFileAllowed()).toBe(false);
-		} finally {
-			restore();
-		}
 	});
 });
 
diff --git a/packages/tui/test/loader.test.ts b/packages/tui/test/loader.test.ts
index 06b35dcfa..20fc98851 100644
--- a/packages/tui/test/loader.test.ts
+++ b/packages/tui/test/loader.test.ts
@@ -1,10 +1,15 @@
-import { describe, expect, it } from "bun:test";
+import { afterEach, describe, expect, it, spyOn, vi } from "bun:test";
 import { TUI } from "@oh-my-pi/pi-tui";
-import { Loader } from "@oh-my-pi/pi-tui/components/loader";
+import { Loader, type LoaderMessageColorFn } from "@oh-my-pi/pi-tui/components/loader";
 import { visibleWidth } from "@oh-my-pi/pi-tui/utils";
 import { VirtualTerminal } from "./virtual-terminal";
 
 describe("Loader component", () => {
+	afterEach(() => {
+		vi.useRealTimers();
+		vi.restoreAllMocks();
+	});
+
 	it("clamps rendered lines to terminal width", async () => {
 		const term = new VirtualTerminal(1, 4);
 		const tui = new TUI(term);
@@ -28,4 +33,36 @@ describe("Loader component", () => {
 		loader.stop();
 		tui.stop();
 	});
+
+	it("keeps spinner cadence when animated messages repaint at 30fps", () => {
+		vi.useFakeTimers();
+		const ui = { requestRender: vi.fn() } as unknown as TUI;
+		const colorMessage = ((text: string) => text) as LoaderMessageColorFn & { animated: true };
+		colorMessage.animated = true;
+		const loader = new Loader(ui, text => text, colorMessage, "Checking", ["0", "1", "2", "3"]);
+
+		vi.advanceTimersByTime(170);
+
+		expect(loader.render(20).join("\n")).toContain("2 Checking");
+		loader.stop();
+	});
+
+	it("dispose() stops the animation so no further renders are scheduled", async () => {
+		const term = new VirtualTerminal(20, 4);
+		const tui = new TUI(term);
+		const loader = new Loader(
+			tui,
+			text => text,
+			text => text,
+			"Checking",
+			["a", "b", "c"],
+		);
+		const spy = spyOn(tui, "requestRender");
+		loader.dispose();
+		const after = spy.mock.calls.length;
+		await Bun.sleep(40); // longer than the spinner interval
+		expect(spy.mock.calls.length).toBe(after);
+		expect(() => loader.dispose()).not.toThrow(); // idempotent
+		tui.stop();
+	});
 });
diff --git a/packages/tui/test/markdown.test.ts b/packages/tui/test/markdown.test.ts
index fcba22a74..0bd139fd2 100644
--- a/packages/tui/test/markdown.test.ts
+++ b/packages/tui/test/markdown.test.ts
@@ -1,7 +1,7 @@
 import { afterAll, afterEach, beforeAll, describe, expect, it } from "bun:test";
 import { stripVTControlCharacters } from "node:util";
 import { Chalk } from "chalk";
-import { Markdown, renderInlineMarkdown } from "../src/components/markdown.js";
+import { clearRenderCache, Markdown, renderInlineMarkdown } from "../src/components/markdown.js";
 import { setTerminalTextSizing, TERMINAL } from "../src/terminal-capabilities.js";
 import { type Component, TUI } from "../src/tui.js";
 import { visibleWidth } from "../src/utils.js";
@@ -1240,6 +1240,29 @@ describe("Module-level LRU render cache", () => {
 		// Output must be byte-identical — cache is transparent to callers.
 		expect(lines2).toEqual(lines1);
 	});
+
+	it("returns caller-owned arrays from L1 and L2 cache hits", () => {
+		clearRenderCache();
+		const text = "Cache mutability sentinel";
+		const width = 80;
+		const markdown = new Markdown(text, 0, 0, defaultMarkdownTheme);
+
+		const first = markdown.render(width);
+		const expected = [...first];
+		first.push("mutated first render");
+
+		const l1Hit = markdown.render(width);
+		expect(l1Hit).toEqual(expected);
+		l1Hit.push("mutated L1 hit");
+		expect(markdown.render(width)).toEqual(expected);
+
+		const l2Markdown = new Markdown(text, 0, 0, defaultMarkdownTheme);
+		const l2Hit = l2Markdown.render(width);
+		expect(l2Hit).toEqual(expected);
+		l2Hit.push("mutated L2 hit");
+		expect(l2Markdown.render(width)).toEqual(expected);
+		expect(new Markdown(text, 0, 0, defaultMarkdownTheme).render(width)).toEqual(expected);
+	});
 });
 
 describe("OSC 66 text-sizing headings", () => {
@@ -1308,3 +1331,43 @@ describe("OSC 66 text-sizing headings", () => {
 		expect(lines.some(line => stripVTControlCharacters(line).includes("Sub"))).toBe(true);
 	});
 });
+
+describe("Markdown.render cache ownership", () => {
+	// Regression: the ask tool renderer did `md(question).push(...optionLines)`,
+	// mutating Markdown's cached array in place. render() handed out the live L1
+	// (per-instance) and L2 (module-level, shared across instances) cache arrays,
+	// so every redraw re-pushed onto the same growing array (+N lines/frame). That
+	// inflated the chat block unboundedly and cascaded into native-scrollback
+	// duplication. render() must return a caller-owned copy so push/splice can
+	// never poison the cache or a future render.
+	afterEach(() => clearRenderCache());
+
+	it("does not let a caller's mutation grow the next render (per-instance cache)", () => {
+		const md = new Markdown("Question text", 1, 0, defaultMarkdownTheme);
+		const baseline = md.render(40).length;
+		md.render(40).push("INJECTED-A", "INJECTED-B");
+		const after = md.render(40);
+		expect(after.length).toBe(baseline);
+		expect(after.some(line => line.includes("INJECTED"))).toBe(false);
+	});
+
+	it("does not let one instance's mutation leak into another via the shared L2 cache", () => {
+		const a = new Markdown("Shared markdown body", 1, 0, defaultMarkdownTheme);
+		const b = new Markdown("Shared markdown body", 1, 0, defaultMarkdownTheme);
+		const baseline = b.render(40).length;
+		// `a` populates L2; mutating its result must not corrupt the entry `b` reads.
+		a.render(40).push("LEAKED-1", "LEAKED-2", "LEAKED-3");
+		const fromB = b.render(40);
+		expect(fromB.length).toBe(baseline);
+		expect(fromB.some(line => line.includes("LEAKED"))).toBe(false);
+	});
+
+	it("stays stable across many mutate-then-render cycles (no accumulation)", () => {
+		const md = new Markdown("Pick one", 1, 0, defaultMarkdownTheme);
+		const baseline = md.render(40).length;
+		for (let i = 0; i < 25; i++) {
+			md.render(40).push(`OPT-${i}`);
+		}
+		expect(md.render(40).length).toBe(baseline);
+	});
+});
diff --git a/packages/tui/test/overlay-scroll.test.ts b/packages/tui/test/overlay-scroll.test.ts
index dfdd5d7dc..af85a123e 100644
--- a/packages/tui/test/overlay-scroll.test.ts
+++ b/packages/tui/test/overlay-scroll.test.ts
@@ -106,7 +106,7 @@ async function withEnv(name: string, value: string, run: () => Promise<void>): P
 
 async function flushRender(term: VirtualTerminal): Promise<void> {
 	await new Promise<void>(resolve => process.nextTick(resolve));
-	await Bun.sleep(17);
+	await Bun.sleep(40);
 	await term.flush();
 }
 
@@ -131,7 +131,7 @@ describe("TUI overlays", () => {
 		expect(term.getScrollBuffer().length).toBeLessThan(200);
 	});
 
-	it("clears stale viewport content on launch without clearing shell scrollback", async () => {
+	it("clears stale viewport content on launch", async () => {
 		const term = new VirtualTerminal(40, 4);
 		term.write("shell-0\r\nshell-1\r\nshell-2\r\nshell-3\r\nshell-4\r\n");
 		await flushRender(term);
@@ -143,7 +143,34 @@ describe("TUI overlays", () => {
 			await flushRender(term);
 
 			expect(term.getViewport().join("\n").includes("shell-")).toBeFalsy();
-			expect(term.getScrollBuffer().join("\n").includes("shell-0")).toBeTruthy();
+		} finally {
+			tui.stop();
+		}
+	});
+
+	it("can clear saved native scrollback on the first paint", async () => {
+		const term = new VirtualTerminal(40, 4);
+		term.write("shell-0\r\nshell-1\r\nshell-2\r\nshell-3\r\nshell-4\r\n");
+		await flushRender(term);
+		const writes: string[] = [];
+		const realWrite = term.write.bind(term);
+		(term as unknown as { write: (s: string) => void }).write = (data: string) => {
+			writes.push(data);
+			realWrite(data);
+		};
+
+		const tui = new TUI(term);
+		tui.addChild(new MutableContentComponent(buildRows(8)));
+		try {
+			tui.start({ clearScrollback: true });
+			await flushRender(term);
+
+			const output = writes.join("");
+			const scrollback = term.getScrollBuffer().join("\n");
+			expect(output).toContain("\x1b[3J");
+			expect(scrollback.includes("shell-")).toBeFalsy();
+			expect(scrollback.includes("row-0")).toBeTruthy();
+			expect(scrollback.includes("row-7")).toBeTruthy();
 		} finally {
 			tui.stop();
 		}
diff --git a/packages/tui/test/process-terminal-render-harness.ts b/packages/tui/test/process-terminal-render-harness.ts
index 56803f77a..15c2ab91a 100644
--- a/packages/tui/test/process-terminal-render-harness.ts
+++ b/packages/tui/test/process-terminal-render-harness.ts
@@ -13,10 +13,10 @@ const PRISTINE: Array<[NodeJS.Process["stdin"] | NodeJS.Process["stdout"], strin
 	[process.stdout, "rows", Object.getOwnPropertyDescriptor(process.stdout, "rows")],
 ];
 
-// One frame interval is 16ms (TUI.#MIN_RENDER_INTERVAL_MS); two frames of
+// One frame interval is ~33ms (TUI.#MIN_RENDER_INTERVAL_MS); two frames of
 // headroom keeps the scheduler-driven paint deterministic without slowing the
 // suite materially.
-const SETTLE_MS = 32;
+const SETTLE_MS = 67;
 
 /**
  * A root component that records the width it is asked to render at. The renderer
diff --git a/packages/tui/test/render-regressions.test.ts b/packages/tui/test/render-regressions.test.ts
index b05df7198..79e6385f7 100644
--- a/packages/tui/test/render-regressions.test.ts
+++ b/packages/tui/test/render-regressions.test.ts
@@ -27,6 +27,34 @@ class MutableLinesComponent implements Component {
 	}
 }
 
+// Models a component that caches its rendered output and only refreshes it when
+// `invalidate()` fires — like a transcript block that freezes a snapshot. A
+// state change behind the cache is invisible until something invalidates it,
+// which is exactly what `resetDisplay()` must do to surface a Ctrl+O expansion.
+class CachedComponent implements Component {
+	#current: string[];
+	#cache: string[] | undefined;
+
+	constructor(lines: string[]) {
+		this.#current = [...lines];
+	}
+
+	setLines(lines: string[]): void {
+		this.#current = [...lines];
+	}
+
+	invalidate(): void {
+		this.#cache = undefined;
+	}
+
+	render(width: number): string[] {
+		if (this.#cache === undefined) {
+			this.#cache = this.#current.map(line => line.slice(0, width));
+		}
+		return this.#cache;
+	}
+}
+
 class WrappingLinesComponent implements Component {
 	#lines: string[];
 
@@ -185,12 +213,12 @@ async function withTerminalRisk<T>(risk: boolean, run: () => T | Promise<T>): Pr
 
 describe("TUI terminal-state regressions", () => {
 	let monotonicNow = 0;
-	// Keep TUI's 16ms render throttle deterministic without sleeping a real frame per render.
+	// Keep TUI's ~33ms render throttle deterministic without sleeping a real frame per render.
 
 	beforeEach(() => {
 		monotonicNow = 0;
 		vi.spyOn(performance, "now").mockImplementation(() => {
-			monotonicNow += 20;
+			monotonicNow += 40;
 			return monotonicNow;
 		});
 	});
@@ -199,6 +227,25 @@ describe("TUI terminal-state regressions", () => {
 		vi.restoreAllMocks();
 	});
 
+	it("coalesces non-force render requests to 30fps", () => {
+		const delays: number[] = [];
+		const renderScheduler = {
+			now: () => 0,
+			scheduleImmediate: (callback: () => void) => callback(),
+			scheduleRender: (_callback: () => void, delayMs: number) => {
+				delays.push(delayMs);
+				return { cancel: () => {} };
+			},
+		};
+		const tui = new TUI(new VirtualTerminal(20, 4), true, { renderScheduler });
+
+		tui.requestRender();
+
+		expect(delays).toHaveLength(1);
+		expect(delays[0]!).toBeCloseTo(1000 / 30, 5);
+		tui.stop();
+	});
+
 	describe("cursor + differential stability", () => {
 		it("keeps stable output across repeated no-op renders", async () => {
 			const term = new VirtualTerminal(40, 10);
@@ -247,6 +294,33 @@ describe("TUI terminal-state regressions", () => {
 				tui.stop();
 			}
 		});
+		it("rewrites changed rows before clearing suffixes for non-synchronized hosts", async () => {
+			const term = new VirtualTerminal(40, 8);
+			const tui = new TUI(term);
+			const component = new MutableLinesComponent([
+				"assistant output already rendered",
+				"tool output already rendered",
+				"todos/status already rendered",
+			]);
+			tui.addChild(component);
+
+			try {
+				tui.start();
+				await settle(term);
+				const writes = captureWrites(term);
+
+				component.setLines(["assistant output already rendered", "tool", "todos/status already rendered"]);
+				tui.requestRender();
+				await settle(term);
+
+				const paint = writes.at(-1) ?? "";
+				expect(paint).toContain("tool\x1b[0m\x1b[K");
+				expect(paint).not.toContain("\x1b[2Ktool");
+				expect(visible(term)[1]).toBe("tool");
+			} finally {
+				tui.stop();
+			}
+		});
 
 		it("clears removed tail lines after shrink", async () => {
 			const term = new VirtualTerminal(40, 10);
@@ -344,6 +418,54 @@ describe("TUI terminal-state regressions", () => {
 			}
 		});
 
+		it("resetDisplay performs a clean redraw without a geometry change", async () => {
+			const term = new VirtualTerminal(20, 3);
+			const tui = new TUI(term);
+			const component = new MutableLinesComponent(rows("L", 8));
+			tui.addChild(component);
+
+			try {
+				tui.start();
+				await settle(term);
+				const writes = captureWrites(term);
+
+				tui.resetDisplay();
+				await settle(term);
+
+				expect(writes.some(write => write.includes("\x1b[2J\x1b[H\x1b[3J"))).toBe(true);
+				expect(term.getScrollBuffer().map(line => line.trimEnd())).toEqual(rows("L", 8));
+				expect(visible(term)).toEqual(["L5", "L6", "L7"]);
+			} finally {
+				tui.stop();
+			}
+		});
+
+		it("resetDisplay surfaces a state change hidden behind a component's render cache", async () => {
+			const term = new VirtualTerminal(20, 3);
+			const tui = new TUI(term);
+			const component = new CachedComponent(rows("L", 8));
+			tui.addChild(component);
+
+			try {
+				tui.start();
+				await settle(term);
+				expect(visible(term)).toEqual(["L5", "L6", "L7"]);
+
+				// The component's content changes, but its render stays cached (a
+				// frozen transcript snapshot). resetDisplay() must invalidate it so the
+				// forced replay reflects the new content rather than the stale cache —
+				// the Ctrl+O expansion path depends on this.
+				component.setLines(rows("M", 8));
+				tui.resetDisplay();
+				await settle(term);
+
+				expect(term.getScrollBuffer().map(line => line.trimEnd())).toEqual(rows("M", 8));
+				expect(visible(term)).toEqual(["M5", "M6", "M7"]);
+			} finally {
+				tui.stop();
+			}
+		});
+
 		it("keeps appended rows in scrollback when a forced render coalesces with content growth", async () => {
 			const term = new VirtualTerminal(20, 3);
 			const tui = new TUI(term);
@@ -1016,7 +1138,7 @@ describe("TUI terminal-state regressions", () => {
 
 		it("keeps appended rows contiguous when a height grow coincides with new content", async () => {
 			// A terminal resize fires requestRender(), and streamed content fires
-			// its own requestRender(); the 16ms throttle coalesces them into a
+			// its own requestRender(); the ~33ms throttle coalesces them into a
 			// single frame that is both taller and longer. The diff/append-tail
 			// emitters position scrolled rows against the previous viewport top and
 			// hardware cursor row, both invalidated by the reflow — so the appended
@@ -1371,7 +1493,7 @@ describe("TUI terminal-state regressions", () => {
 						await settle(term);
 
 						// SIGWINCH (height shrink) and a streamed token arrive inside the
-						// same 16ms frame budget. The TUI's own resize handler schedules a
+						// same ~33ms frame budget. The TUI's own resize handler schedules a
 						// non-forced render; the append rides along.
 						lines.push("line-40 streamed");
 						component.setLines(lines);
@@ -1681,7 +1803,7 @@ describe("TUI terminal-state regressions", () => {
 			const tui = new TUI(term);
 			const collapsedLines = [
 				"frame-top",
-				"code preview … 16 more lines ⟨(Ctrl+O for more)⟩",
+				"code preview … 16 more lines ⟨Ctrl+O: Expand⟩",
 				"output preview … 106 more lines (ctrl+o to expand)",
 				...rows("json-", 10),
 				"status",
@@ -1732,7 +1854,7 @@ describe("TUI terminal-state regressions", () => {
 			const tui = new TUI(term);
 			const component = new MutableLinesComponent([
 				"frame-top",
-				"code preview … 16 more lines ⟨(Ctrl+O for more)⟩",
+				"code preview … 16 more lines ⟨Ctrl+O: Expand⟩",
 				"output preview … 106 more lines (ctrl+o to expand)",
 				...rows("json-", 10),
 				"status",
@@ -1777,7 +1899,7 @@ describe("TUI terminal-state regressions", () => {
 			const tui = new TUI(term);
 			const component = new MutableLinesComponent([
 				"frame-top",
-				"code preview … 16 more lines ⟨(Ctrl+O for more)⟩",
+				"code preview … 16 more lines ⟨Ctrl+O: Expand⟩",
 				"output preview … 106 more lines (ctrl+o to expand)",
 				...rows("json-", 10),
 				"status",
@@ -2096,6 +2218,66 @@ describe("TUI terminal-state regressions", () => {
 				tui.stop();
 			}
 		});
+
+		it("repaints only the active-grid bottom row while unknown viewport mutation is deferred", async () => {
+			const initial = [...rows("line-", 12), "spinner-a"];
+			const updated = ["edited-0", ...rows("line-", 12).slice(1), "spinner-b"];
+
+			await withTerminalRisk(true, async () => {
+				const term = new UnknownViewportTerminal(40, 6);
+				const tui = new TUI(term);
+				const component = new MutableLinesComponent(initial);
+				tui.addChild(component);
+
+				try {
+					tui.start();
+					await settle(term);
+					const writes = captureWrites(term);
+
+					component.setLines(updated);
+					tui.requestRender();
+					await settle(term);
+
+					const viewport = visible(term).map(line => line.trim());
+					expect(viewport.at(-1)).toBe("spinner-b");
+					expect(term.getScrollBuffer().join("\n")).not.toContain("edited-0");
+					const paint = writes.at(-1) ?? "";
+					expect(paint).toContain("\rspinner-b\x1b[0m\x1b[K");
+					expect(paint).not.toContain("\x1b[H");
+					expect(paint).not.toContain("\x1b[3J");
+				} finally {
+					tui.stop();
+				}
+
+				const scrolledTerm = new UnknownViewportTerminal(40, 6);
+				const scrolledTui = new TUI(scrolledTerm);
+				const scrolledComponent = new MutableLinesComponent(initial);
+				scrolledTui.addChild(scrolledComponent);
+
+				try {
+					scrolledTui.start();
+					await settle(scrolledTerm);
+					scrolledTerm.scrollLines(-1);
+					const before = scrolledTerm.getBufferPosition();
+					const beforeViewport = visible(scrolledTerm).map(line => line.trim());
+					const writes = captureWrites(scrolledTerm);
+
+					scrolledComponent.setLines(updated);
+					scrolledTui.requestRender();
+					await settle(scrolledTerm);
+
+					expect(scrolledTerm.getBufferPosition()).toEqual(before);
+					expect(visible(scrolledTerm).map(line => line.trim())).toEqual(beforeViewport);
+					expect(scrolledTerm.getScrollBuffer().join("\n")).not.toContain("edited-0");
+					const paint = writes.at(-1) ?? "";
+					expect(paint).toContain("\rspinner-b\x1b[0m\x1b[K");
+					expect(paint).not.toContain("\x1b[H");
+					expect(paint).not.toContain("\x1b[3J");
+				} finally {
+					scrolledTui.stop();
+				}
+			});
+		});
 		it("rebuilds history when a shrink leaves no real rows above the scrollback boundary", async () => {
 			// Reviewer scenario (#1599): a large completion-style collapse (e.g. a 100-row
 			// streamed transcript shrinking to a 20-row final cell in a 10-row viewport)
@@ -3046,6 +3228,86 @@ describe("TUI terminal-state regressions", () => {
 		});
 	});
 
+	describe("fullscreen overlay alt-screen", () => {
+		it("enters the alt buffer on show, leaves it on hide, and emits no ED3 while modal", async () => {
+			const term = new VirtualTerminal(40, 8, 200);
+			const writes = captureWrites(term);
+			const tui = new TUI(term);
+			tui.addChild(new MutableLinesComponent(rows("base-", 8)));
+
+			try {
+				tui.start();
+				await settle(term);
+
+				const showFrom = writes.length;
+				const handle = tui.showOverlay(new MutableLinesComponent(["MODAL-0", "MODAL-1"]), {
+					anchor: "bottom-center",
+					width: "100%",
+					maxHeight: "100%",
+					margin: 0,
+					fullscreen: true,
+				});
+				await settle(term);
+
+				const modalWrites = writes.slice(showFrom).join("");
+				// Borrowed the alternate screen buffer …
+				expect(modalWrites).toContain("\x1b[?1049h");
+				// … enabled mouse tracking for click/scroll support …
+				expect(modalWrites).toContain("\x1b[?1000h");
+				expect(modalWrites).toContain("\x1b[?1006h");
+				// … and never erased scrollback (ED3) or otherwise touched the transcript.
+				expect(modalWrites).not.toContain("\x1b[3J");
+				expect(visible(term).some(line => line.includes("MODAL-0"))).toBeTrue();
+
+				const hideFrom = writes.length;
+				handle.hide();
+				await settle(term);
+
+				const hideWrites = writes.slice(hideFrom).join("");
+				expect(hideWrites).toContain("\x1b[?1049l");
+				// Mouse tracking is disabled again so the rest of the app keeps native
+				// terminal selection.
+				expect(hideWrites).toContain("\x1b[?1000l");
+				// Transcript is back on the normal screen after leaving the alt buffer.
+				expect(visible(term).some(line => line.includes("base-"))).toBeTrue();
+				expect(visible(term).some(line => line.includes("MODAL-0"))).toBeFalse();
+			} finally {
+				tui.stop();
+			}
+		});
+
+		it("leaves native scrollback untouched across the modal lifetime", async () => {
+			const term = new VirtualTerminal(40, 6, 200);
+			const tui = new TUI(term);
+			// Base transcript overflows the viewport, so rows land in scrollback.
+			tui.addChild(new MutableLinesComponent(rows("base-", 24)));
+
+			try {
+				tui.start();
+				await settle(term);
+				const scrollbackBefore = term.getScrollBuffer().map(line => line.trimEnd());
+
+				const handle = tui.showOverlay(new MutableLinesComponent(["MODAL"]), {
+					anchor: "bottom-center",
+					width: "100%",
+					maxHeight: "100%",
+					margin: 0,
+					fullscreen: true,
+				});
+				await settle(term);
+				handle.hide();
+				await settle(term);
+
+				// The modal borrowed/returned the alt buffer without rewriting the
+				// normal screen's scrollback — the transcript a reader scrolled up to
+				// see is identical before and after.
+				expect(term.getScrollBuffer().map(line => line.trimEnd())).toEqual(scrollbackBefore);
+			} finally {
+				tui.stop();
+			}
+		});
+	});
+
 	describe("stress scenarios", () => {
 		it("rapid content mutations converge to final expected screen", async () => {
 			const term = new VirtualTerminal(30, 8);
@@ -3344,8 +3606,8 @@ describe("TUI terminal-state regressions", () => {
 				// Initial paint: only the styled row carries background cells.
 				expect(backgroundRows(term, height)).toEqual([1]);
 
-				// Diff path: rewriting the row below starts with \x1b[2K — with leaked
-				// background, BCE would paint that whole row red.
+				// Diff path: rewriting the row below clears only after the row reset;
+				// with leaked background, BCE would otherwise paint that row red.
 				component.setLines(["plain-0", UNRESET_BG_ROW, "EDITED-2"]);
 				tui.requestRender();
 				await settle(term);
@@ -3377,8 +3639,8 @@ describe("TUI terminal-state regressions", () => {
 				expect(foregroundRows(term, height)).toEqual([1]);
 				expect(underlineRows(term, height)).toEqual([1]);
 
-				// Rewriting the next row starts with an erase; leaked SGR would make
-				// the edited row green/underlined despite containing plain text.
+				// Rewriting the next row clears only after the row reset; leaked SGR
+				// would make the edited row green/underlined despite containing plain text.
 				component.setLines(["plain-0", UNRESET_FG_UNDERLINE_ROW, "EDITED-2"]);
 				tui.requestRender();
 				await settle(term);
@@ -3413,7 +3675,7 @@ describe("TUI terminal-state regressions", () => {
 				tui.start();
 				await settle(term);
 
-				// Force a full repaint (viewport rewrite path emits \x1b[2K per row).
+				// Force a full repaint (viewport rewrite path suffix-clears each text row).
 				tui.requestRender(true);
 				await settle(term);
 
@@ -3938,7 +4200,7 @@ describe("foreground-tool streaming on ED3-risk terminals", () => {
 				];
 				component.setLines(frameB);
 				tui.requestRender();
-				await settle(term);
+				await term.waitForRender();
 				expect(visible(term)).toEqual(["chip-1", "chip-2", "chip-3", "loader", "todos", "editor"]);
 
 				// Frame C: a visible chip collapses (a shrink whose first change lands in
@@ -3964,7 +4226,7 @@ describe("foreground-tool streaming on ED3-risk terminals", () => {
 				];
 				component.setLines(frameC);
 				tui.requestRender();
-				await settle(term);
+				await term.waitForRender();
 				expect(visible(term)).toEqual(["chip-0", "chip-1", "chip-2", "loader", "todos", "editor"]);
 			} finally {
 				tui.stop();
@@ -3972,6 +4234,23 @@ describe("foreground-tool streaming on ED3-risk terminals", () => {
 		});
 	});
 
+	it("honors a clear-scrollback replay queued before the initial paint", async () => {
+		const term = new VirtualTerminal(40, 6);
+		const writes = captureWrites(term);
+		const tui = new TUI(term);
+		tui.addChild(new MutableLinesComponent(["resumed-message", "prompt>"]));
+
+		try {
+			tui.start();
+			tui.requestRender(true, { clearScrollback: true });
+			await settle(term);
+
+			expect(writes.join("")).toContain("\x1b[3J");
+		} finally {
+			tui.stop();
+		}
+	});
+
 	// Repro of the drag-resize line-duplication: dragging the terminal smaller
 	// fires a stream of height shrinks. While the transcript FITS the viewport,
 	// each shrink used to scroll live rows into native scrollback — the in-place
diff --git a/packages/tui/test/render-stress-harness.ts b/packages/tui/test/render-stress-harness.ts
index b80d93277..79066ee75 100644
--- a/packages/tui/test/render-stress-harness.ts
+++ b/packages/tui/test/render-stress-harness.ts
@@ -327,12 +327,14 @@ interface AppliedOperation {
 	// them. Defaults to the net frame growth when absent.
 	transientFrameGrowth?: number;
 	// The periodic prompt-submit checkpoint pins the viewport to the bottom and
-	// runs the real reconciliation (`refreshNativeScrollbackIfDirty` outside
-	// `normal`, a `/clear`-style forced rebuild for `normal`), so native
-	// scrollback must equal the transcript afterward. Plain `scrollToBottom` /
-	// forced-render ops also set `checkpoint`, but on Windows hosts a forced
-	// render cannot rebuild ConPTY-hidden history (it defers to the next submit),
-	// so the clean-buffer oracle keys on this flag for non-`normal` scenarios.
+	// attempts the real reconciliation (`refreshNativeScrollbackIfDirty` outside
+	// `normal`, a `/clear`-style forced rebuild for `normal`). Native scrollback
+	// must equal the transcript only when that reconciliation actually ran:
+	// ConPTY/Windows and other unobservable host-scrollback paths deliberately
+	// keep dirty history deferred until the renderer gets a positive at-tail probe.
+	// Plain `scrollToBottom` / forced-render ops also set `checkpoint`, but on
+	// Windows hosts a forced render cannot rebuild ConPTY-hidden history, so the
+	// clean-buffer oracle keys on this flag for non-`normal` scenarios.
 	reconcilesNativeScrollback?: boolean;
 }
 
@@ -2004,8 +2006,10 @@ class StressDriver {
 	async #checkpoint(index: number, kind: "periodicCheckpoint"): Promise<void> {
 		const before = this.#snapshot();
 		// Model a prompt submit: the editor keystroke pins the terminal to the
-		// bottom, then the app reconciles any deferred native-scrollback rewrite.
+		// bottom, then the app reconciles any deferred native-scrollback rewrite
+		// only if the renderer can prove the native host viewport is at the tail.
 		this.#term.scrollLines(LARGE_SCROLL);
+		let reconcilesNativeScrollback = false;
 		if (this.#traits.strictNativeScrollback || this.#traits.preservesPaneHistory) {
 			// Normal POSIX uses a /clear-style forced rebuild; tmux keeps its forced
 			// repaint (its pane history cannot be destructively reconciled).
@@ -2013,13 +2017,13 @@ class StressDriver {
 				allowUnknownViewportMutation: true,
 				clearScrollback: this.#traits.strictNativeScrollback,
 			});
+			reconcilesNativeScrollback = this.#traits.strictNativeScrollback;
 		} else {
 			// Unknown-viewport / ED3-risk / Windows hosts take the real prompt-submit
-			// path: refreshNativeScrollbackIfDirty rebuilds the deferred history now
-			// that the keystroke has pinned the viewport to the bottom. This is where
-			// the streaming turn's dirty/lagged scrollback must reconcile to an exact
-			// copy of the transcript.
-			this.#tui.refreshNativeScrollbackIfDirty({ allowUnknownViewport: true });
+			// path. `refreshNativeScrollbackIfDirty` returns false for permanently
+			// unobservable hosts such as ConPTY, where a submit key is not proof that
+			// hidden host scrollback is at the tail and ED3 would still yank readers.
+			reconcilesNativeScrollback = this.#tui.refreshNativeScrollbackIfDirty({ allowUnknownViewport: true });
 		}
 		await this.#settle();
 		const after = this.#snapshot();
@@ -2034,7 +2038,7 @@ class StressDriver {
 				forcedRender: true,
 				mutatesViewport: true,
 				checkpoint: true,
-				reconcilesNativeScrollback: true,
+				reconcilesNativeScrollback,
 			},
 			before,
 			after,
@@ -2082,10 +2086,12 @@ class StressDriver {
 		this.#assertUniqueContentNoUnexpectedDuplicates(op, before, after, index);
 		this.#assertNoBackgroundBleed(op, before, after, index);
 		// Native scrollback must reconcile to an exact bottom-anchored copy of the
-		// transcript at every checkpoint — including the unknown-viewport / ED3-risk
-		// / Windows hosts whose live oracles are relaxed (they defer history rewrites
-		// mid-stream and only reconcile here). tmux is excluded: its pane history is
-		// preserved, not rebuilt, so the buffer snapshot is the view, not history.
+		// transcript at checkpoints where the renderer actually performed a
+		// destructive/native-history rebuild. Unknown ConPTY host scrollback and
+		// ED3-risk terminals with no positive at-tail probe intentionally keep dirty
+		// history deferred; asserting a clean buffer there would contradict the
+		// anti-yank contract. tmux is excluded: its pane history is preserved, not
+		// rebuilt, so the buffer snapshot is the view, not history.
 		if (
 			op.checkpoint &&
 			!this.#traits.preservesPaneHistory &&
diff --git a/packages/tui/test/scroll-view.test.ts b/packages/tui/test/scroll-view.test.ts
new file mode 100644
index 000000000..6e2c2af67
--- /dev/null
+++ b/packages/tui/test/scroll-view.test.ts
@@ -0,0 +1,103 @@
+import { describe, expect, it } from "bun:test";
+import { ScrollView } from "../src/components/scroll-view";
+import { Ellipsis, visibleWidth } from "../src/utils";
+
+const theme = {
+	track: () => "T",
+	thumb: () => "B",
+};
+
+describe("ScrollView", () => {
+	it("renders a fixed-height viewport and omits auto scrollbar when content fits", () => {
+		const view = new ScrollView(["one", "two"], { height: 3, theme });
+
+		expect(view.render(10)).toEqual(["one", "two", ""]);
+	});
+
+	it("renders a right-edge scrollbar when content overflows", () => {
+		const view = new ScrollView(["alpha", "beta", "gamma", "delta", "omega"], { height: 3, theme });
+
+		expect(view.render(6)).toEqual(["alphaB", "beta T", "gammaT"]);
+	});
+
+	it("scrolls and clamps offsets", () => {
+		const view = new ScrollView(["one", "two", "three", "four", "five"], { height: 3, theme });
+
+		view.scroll(10);
+
+		expect(view.getScrollOffset()).toBe(2);
+		expect(view.render(6)).toEqual(["threeT", "four T", "five B"]);
+
+		view.scroll(-10);
+
+		expect(view.getScrollOffset()).toBe(0);
+	});
+
+	it("reserves a scrollbar column in always mode", () => {
+		const view = new ScrollView(["one"], { height: 2, scrollbar: "always", theme });
+
+		expect(view.render(5)).toEqual(["one B", "    B"]);
+	});
+
+	it("does not reserve a scrollbar column in never mode", () => {
+		const view = new ScrollView(["alpha", "beta", "gamma"], { height: 2, scrollbar: "never", theme });
+
+		expect(view.render(6)).toEqual(["alpha", "beta"]);
+	});
+
+	it("renders scrollbar geometry for pre-windowed lines", () => {
+		const view = new ScrollView(["gamma", "delta"], { height: 2, totalRows: 4, theme });
+		view.setScrollOffset(2);
+
+		expect(view.render(6)).toEqual(["gammaT", "deltaB"]);
+	});
+
+	it("does not render a scrollbar when width is zero", () => {
+		const view = new ScrollView(["one", "two"], { height: 1, theme });
+
+		expect(view.render(0)).toEqual([""]);
+	});
+
+	it("clamps scroll offset when content shrinks", () => {
+		const view = new ScrollView(["one", "two", "three", "four"], { height: 2, theme });
+		view.scrollToBottom();
+
+		view.setLines(["one"]);
+
+		expect(view.getScrollOffset()).toBe(0);
+		expect(view.render(10)).toEqual(["one", ""]);
+	});
+
+	it("keeps rendered rows within requested width with ANSI input", () => {
+		const view = new ScrollView(["\x1b[31malphabet\x1b[0m", "plain", "tail"], { height: 2, theme });
+		const rendered = view.render(5);
+
+		expect(rendered).toHaveLength(2);
+		expect(rendered.every(line => visibleWidth(line) <= 5)).toBe(true);
+		expect(rendered[0]).toContain("B");
+	});
+
+	it("appends an overflow ellipsis by default and omits it when configured", () => {
+		const long = ["abcdefghij"];
+		const def = new ScrollView(long, { height: 1, scrollbar: "never", theme });
+		expect(def.render(5)[0]).toContain("…");
+
+		const omit = new ScrollView(long, { height: 1, scrollbar: "never", ellipsis: Ellipsis.Omit, theme });
+		expect(omit.render(5)[0]).toBe("abcde");
+	});
+
+	it("handles navigation keys, with Shift+Arrow scrolling by fastScrollLines", () => {
+		const view = new ScrollView(
+			Array.from({ length: 50 }, (_, i) => String(i)),
+			{ height: 5, fastScrollLines: 7, theme },
+		);
+
+		expect(view.handleScrollKey("\x1b[B")).toBe(true); // down
+		expect(view.getScrollOffset()).toBe(1);
+		expect(view.handleScrollKey("\x1b[1;2B")).toBe(true); // shift+down
+		expect(view.getScrollOffset()).toBe(8);
+		expect(view.handleScrollKey("\x1b[1;2A")).toBe(true); // shift+up
+		expect(view.getScrollOffset()).toBe(1);
+		expect(view.handleScrollKey("x")).toBe(false);
+	});
+});
diff --git a/packages/tui/test/select-list.test.ts b/packages/tui/test/select-list.test.ts
index a7026074b..f7be4ec02 100644
--- a/packages/tui/test/select-list.test.ts
+++ b/packages/tui/test/select-list.test.ts
@@ -203,4 +203,26 @@ describe("SelectList", () => {
 		expect(rendered).not.toContain("Search:");
 		expect(list.getSelectedItem()?.value).toBe("alpha");
 	});
+
+	it("renders a right-edge scrollbar when the list overflows maxVisible", () => {
+		const items = Array.from({ length: 8 }, (_, i) => ({ value: `v${i}`, label: `Item ${i}` }));
+		const list = new SelectList(items, 3, testTheme);
+
+		const rendered = list.render(40);
+
+		// Default ScrollView glyphs: track │, thumb █. Overflow must surface the bar
+		// and drop the old (N/M) text indicator.
+		expect(rendered.join("\n")).toContain("█");
+		expect(rendered.join("\n")).not.toContain("(1/8)");
+	});
+
+	it("omits the scrollbar when every item fits", () => {
+		const items = [
+			{ value: "alpha", label: "Alpha" },
+			{ value: "beta", label: "Beta" },
+		];
+		const list = new SelectList(items, 5, testTheme);
+
+		expect(list.render(40).join("\n")).not.toContain("█");
+	});
 });
diff --git a/packages/tui/test/settings-list.test.ts b/packages/tui/test/settings-list.test.ts
index e75d769a1..c69094be3 100644
--- a/packages/tui/test/settings-list.test.ts
+++ b/packages/tui/test/settings-list.test.ts
@@ -70,4 +70,41 @@ describe("SettingsList", () => {
 		expect(output).toContain("[changed-value]on");
 		expect(output).not.toContain("[changed-label]Default");
 	});
+
+	it("renders long settings tabs through a scrollbar viewport", () => {
+		const list = new SettingsList(
+			Array.from({ length: 6 }, (_, i) => ({
+				id: `item-${i}`,
+				label: `Item ${i}`,
+				currentValue: `value-${i}`,
+				values: [`value-${i}`],
+			})),
+			3,
+			{
+				...testTheme,
+				label: (text: string, selected: boolean) => (selected ? `[selected]${text}` : text),
+				hint: (text: string) => `[dim]${text}`,
+			},
+			() => {},
+			() => {},
+		);
+
+		const output = list.render(32);
+
+		expect(output.slice(0, 3).join("\n")).toContain("[selected]");
+		expect(output.slice(0, 3).join("\n")).toContain("[dim]");
+		expect(output).not.toContain("(1/6)");
+	});
+
+	it("does not reserve a scrollbar column when all settings fit", () => {
+		const list = new SettingsList(
+			[{ id: "mode", label: "Mode", currentValue: "123456", values: ["123456"] }],
+			3,
+			testTheme,
+			() => {},
+			() => {},
+		);
+
+		expect(list.render(16)[0]).toBe("→ Mode  123456");
+	});
 });
diff --git a/packages/tui/test/slash-autocomplete-viewport.test.ts b/packages/tui/test/slash-autocomplete-viewport.test.ts
index dd9b8596d..71d4e1b9c 100644
--- a/packages/tui/test/slash-autocomplete-viewport.test.ts
+++ b/packages/tui/test/slash-autocomplete-viewport.test.ts
@@ -29,14 +29,32 @@ class SlashProvider implements AutocompleteProvider {
 }
 
 class UnknownViewportTerminal extends VirtualTerminal {
+	#eagerEraseScrollbackRisk: boolean | undefined;
+
+	constructor(columns: number, rows: number, eagerEraseScrollbackRisk?: boolean) {
+		super(columns, rows);
+		this.#eagerEraseScrollbackRisk = eagerEraseScrollbackRisk;
+	}
+
 	isNativeViewportAtBottom(): undefined {
 		return undefined;
 	}
+
+	hasEagerEraseScrollbackRisk(): boolean | undefined {
+		return this.#eagerEraseScrollbackRisk;
+	}
 }
 
 async function settle(term: VirtualTerminal): Promise<void> {
 	await new Promise<void>(resolve => process.nextTick(resolve));
-	await Bun.sleep(120);
+	// Each keystroke arms Editor's autocomplete debounce (100ms) before the
+	// provider is re-queried, and the resulting onAutocompleteUpdate render is
+	// throttled by the TUI's MIN_RENDER_INTERVAL_MS (~33ms). 120ms no longer
+	// covers debounce plus render jitter, so capture could still see the stale
+	// (unfiltered) menu and push the live editor row out of a 6-row viewport
+	// (issue #1979). Stay well above 100+33ms so the post-debounce render lands
+	// before getViewport() runs.
+	await Bun.sleep(250);
 	await term.flush();
 }
 
@@ -75,6 +93,87 @@ describe("slash command autocomplete with unknown native viewport state", () =>
 		}
 	});
 
+	it("repaints direct autocomplete shrink on ED3-risk POSIX terminals", async () => {
+		const originalPlatform = process.platform;
+		Object.defineProperty(process, "platform", { configurable: true, value: "darwin" });
+		let tui: TUI | undefined;
+		try {
+			const term = new UnknownViewportTerminal(40, 8, true);
+			tui = new TUI(term);
+			const root = new Container();
+			root.addChild({
+				invalidate() {},
+				render: () => ["chat-0", "chat-1", "chat-2", "chat-3", "chat-4", "chat-5", "chat-6"],
+			});
+			const editor = new Editor(defaultEditorTheme);
+			let submitted: string | undefined;
+			editor.setAutocompleteProvider(new SlashProvider());
+			editor.onAutocompleteUpdate = () => {
+				tui?.requestRender(false, { allowUnknownViewportMutation: true });
+			};
+			editor.onSubmit = text => {
+				submitted = text;
+			};
+			root.addChild(editor);
+			tui.addChild(root);
+			tui.setFocus(editor);
+
+			tui.start();
+			await settle(term);
+			for (const char of "/st") {
+				term.sendInput(char);
+				await settle(term);
+			}
+			let viewport = term.getViewport().join("\n");
+			expect(editor.getText()).toBe("/st");
+			expect(viewport).toContain("/st");
+			expect(viewport).not.toContain("/s\n");
+			expect(term.getScrollBuffer()).toEqual([
+				"chat-0",
+				"chat-1",
+				"chat-2",
+				"chat-3",
+				"chat-4",
+				"chat-5",
+				"chat-6",
+				"+--------------------------------------+",
+				"+- /st|                               -+",
+				"> status",
+				"  stats",
+				"  stop",
+				"",
+				"",
+			]);
+
+			term.sendInput("\r");
+			await settle(term);
+			viewport = term.getViewport().join("\n");
+			expect(submitted).toBe("/status");
+			expect(editor.getText()).toBe("");
+			expect(viewport).not.toContain("status");
+			expect(viewport).not.toContain("/st");
+			expect(term.getScrollBuffer()).toEqual([
+				"chat-0",
+				"chat-1",
+				"chat-2",
+				"chat-3",
+				"chat-4",
+				"chat-5",
+				"chat-6",
+				"+--------------------------------------+",
+				"+- |                                  -+",
+				"",
+				"",
+				"",
+				"",
+				"",
+			]);
+		} finally {
+			tui?.stop();
+			Object.defineProperty(process, "platform", { configurable: true, value: originalPlatform });
+		}
+	});
+
 	it("repaints autocomplete updates coalesced with offscreen background mutations", async () => {
 		const originalPlatform = process.platform;
 		const originalWtSession = Bun.env.WT_SESSION;
diff --git a/packages/tui/test/start-listener.test.ts b/packages/tui/test/start-listener.test.ts
new file mode 100644
index 000000000..3e5f372b1
--- /dev/null
+++ b/packages/tui/test/start-listener.test.ts
@@ -0,0 +1,24 @@
+import { describe, expect, it } from "bun:test";
+import { TUI } from "@oh-my-pi/pi-tui";
+import { VirtualTerminal } from "./virtual-terminal";
+
+describe("TUI start listeners", () => {
+	it("fires registered hooks on initial start and restart", () => {
+		const tui = new TUI(new VirtualTerminal(80, 24));
+		let starts = 0;
+		tui.addStartListener(() => {
+			starts++;
+		});
+
+		try {
+			tui.start();
+			expect(starts).toBe(1);
+
+			tui.stop();
+			tui.start();
+			expect(starts).toBe(2);
+		} finally {
+			tui.stop();
+		}
+	});
+});
diff --git a/packages/tui/test/streaming-scrollback-defer.test.ts b/packages/tui/test/streaming-scrollback-defer.test.ts
index 84919cd63..e1723c57d 100644
--- a/packages/tui/test/streaming-scrollback-defer.test.ts
+++ b/packages/tui/test/streaming-scrollback-defer.test.ts
@@ -39,7 +39,10 @@ class AppendOnlyLiveLineList extends LiveLineList {
 }
 
 async function settle(term: VirtualTerminal): Promise<void> {
-	await Bun.sleep(20);
+	const nextTick = Promise.withResolvers<void>();
+	process.nextTick(nextTick.resolve);
+	await nextTick.promise;
+	await Bun.sleep(40);
 	await term.flush();
 }
 
@@ -372,6 +375,48 @@ describe("streaming scrollback defer", () => {
 		});
 	});
 
+	it("keeps committed prefix accounting after a capped streaming frame", async () => {
+		if (process.platform === "win32") return;
+		await withTerminalRisk(true, async () => {
+			const term = new VirtualTerminal(24, 4);
+			overrideProbe(term, undefined);
+			const tui = new TUI(term);
+			const sealed = new LineList(rows("base-", 12));
+
+			try {
+				tui.addChild(sealed);
+				tui.start();
+				await settle(term);
+
+				const writes = capture(term);
+				tui.setEagerNativeScrollbackRebuild(true);
+
+				// No live-region marker yet: ED3-risk streaming caps this transient
+				// frame to the viewport. The already-committed base-0..base-7 rows
+				// remain physically in native scrollback and must stay accounted.
+				sealed.setLines([...rows("base-", 12), ...rows("transient-", 30)]);
+				tui.requestRender();
+				await settle(term);
+
+				expect(eraseScrollbackCount(writes)).toBe(0);
+
+				// A later frame introduces a live region after the same sealed prefix.
+				// If the cap zeroed the high-water mark, liveRegionPinned would append
+				// base-0..base-11 again, duplicating base-0..base-7 in native history.
+				const live = new LiveLineList(rows("live-", 20));
+				sealed.setLines(rows("base-", 12));
+				tui.addChild(live);
+				tui.requestRender();
+				await settle(term);
+
+				expect(eraseScrollbackCount(writes)).toBe(0);
+				expect(term.getScrollBuffer().filter(line => line.startsWith("base-"))).toEqual(rows("base-", 12));
+			} finally {
+				tui.stop();
+			}
+		});
+	});
+
 	it("erases mis-wrapped native scrollback on resize even mid-stream on ED3-risk terminals", async () => {
 		if (process.platform === "win32") return;
 		await withTerminalRisk(true, async () => {
diff --git a/packages/tui/test/submit-checkpoint-reconcile.test.ts b/packages/tui/test/submit-checkpoint-reconcile.test.ts
new file mode 100644
index 000000000..7c10e1ef9
--- /dev/null
+++ b/packages/tui/test/submit-checkpoint-reconcile.test.ts
@@ -0,0 +1,117 @@
+import { describe, expect, it } from "bun:test";
+import { type Component, setTerminalSubmitPinsViewportToTail, TERMINAL, TUI } from "@oh-my-pi/pi-tui";
+import { VirtualTerminal } from "./virtual-terminal";
+
+// The prompt-submit reconciliation checkpoint (`refreshNativeScrollbackIfDirty`)
+// must ED3-rebuild deferred-dirty native scrollback on genuine local terminals,
+// where the submit keystroke pins the host to its tail, even though their viewport
+// position is unprobeable (ghostty/kitty/iTerm report `undefined`). Without this,
+// every offscreen shrink/edit that defers during streaming leaves stale rows above
+// the viewport that never clear until Ctrl+L or a resize. Hosts that cannot prove
+// at-tail (Windows console/Terminal, SSH, multiplexers — modeled here by
+// submitPinsViewportToTail=false) keep deferring so a scrolled reader is never
+// yanked by ED3 (#1610/#1682/#1746).
+
+class LineList implements Component {
+	#lines: string[];
+	constructor(lines: string[]) {
+		this.#lines = [...lines];
+	}
+	invalidate(): void {}
+	render(width: number): string[] {
+		return this.#lines.map(line => line.slice(0, width));
+	}
+	setLines(lines: string[]): void {
+		this.#lines = [...lines];
+	}
+}
+
+async function settle(term: VirtualTerminal): Promise<void> {
+	const tick = Promise.withResolvers<void>();
+	process.nextTick(tick.resolve);
+	await tick.promise;
+	await Bun.sleep(20);
+	await term.flush();
+}
+
+function capture(term: VirtualTerminal): string[] {
+	const writes: string[] = [];
+	const realWrite = term.write.bind(term);
+	(term as unknown as { write: (s: string) => void }).write = (data: string) => {
+		writes.push(data);
+		realWrite(data);
+	};
+	return writes;
+}
+
+function overrideProbe(term: VirtualTerminal, answer: boolean | undefined): void {
+	(term as unknown as { isNativeViewportAtBottom: () => boolean | undefined }).isNativeViewportAtBottom = () => answer;
+}
+
+const eraseScrollbackCount = (writes: string[]): number => (writes.join("").match(/\x1b\[3J/g) ?? []).length;
+
+interface CheckpointResult {
+	deferredErases: number;
+	reconciled: boolean;
+	checkpointErases: number;
+	viewport: string[];
+}
+
+// Drive an ED3-risk terminal with an unprobeable viewport through an eager-stream
+// offscreen shrink (which defers, marking native scrollback dirty without erasing)
+// and then the prompt-submit checkpoint, returning what each step emitted.
+async function deferThenCheckpoint(submitPinsViewportToTail: boolean): Promise<CheckpointResult> {
+	const savedRisk = TERMINAL.eagerEraseScrollbackRisk;
+	const savedPins = TERMINAL.submitPinsViewportToTail;
+	// RuntimeTerminal exposes these as writable — no cast needed.
+	TERMINAL.eagerEraseScrollbackRisk = true;
+	setTerminalSubmitPinsViewportToTail(submitPinsViewportToTail);
+	const term = new VirtualTerminal(80, 12);
+	overrideProbe(term, undefined);
+	const tui = new TUI(term);
+	const component = new LineList(Array.from({ length: 60 }, (_value, index) => `init-${index}`));
+	tui.addChild(component);
+	try {
+		tui.start();
+		await settle(term);
+		const writes = capture(term);
+		tui.setEagerNativeScrollbackRebuild(true);
+
+		// Offscreen shrink: repaints the visible window in place and marks native
+		// scrollback dirty instead of erasing (the deferral that strands stale rows).
+		component.setLines(Array.from({ length: 4 }, (_value, index) => `done-${index}`));
+		tui.requestRender();
+		await settle(term);
+		const deferredErases = eraseScrollbackCount(writes);
+
+		const reconciled = tui.refreshNativeScrollbackIfDirty();
+		await settle(term);
+		return {
+			deferredErases,
+			reconciled,
+			checkpointErases: eraseScrollbackCount(writes),
+			viewport: term.getViewport().map(line => line.trim()),
+		};
+	} finally {
+		tui.stop();
+		TERMINAL.eagerEraseScrollbackRisk = savedRisk;
+		setTerminalSubmitPinsViewportToTail(savedPins);
+	}
+}
+
+describe("submit-checkpoint native scrollback reconciliation", () => {
+	it("reconciles deferred scrollback at the checkpoint when submit pins the host to its tail", async () => {
+		const result = await deferThenCheckpoint(true);
+		expect(result.deferredErases).toBe(0);
+		expect(result.reconciled).toBe(true);
+		expect(result.checkpointErases).toBeGreaterThan(0);
+		expect(result.viewport).toContain("done-3");
+	});
+
+	it("keeps deferring at the checkpoint when the host cannot prove at-tail", async () => {
+		const result = await deferThenCheckpoint(false);
+		expect(result.deferredErases).toBe(0);
+		expect(result.reconciled).toBe(false);
+		expect(result.checkpointErases).toBe(0);
+	});
+});
diff --git a/packages/tui/test/terminal-appearance.test.ts b/packages/tui/test/terminal-appearance.test.ts
index 03bd9bbee..88712e02c 100644
--- a/packages/tui/test/terminal-appearance.test.ts
+++ b/packages/tui/test/terminal-appearance.test.ts
@@ -1,14 +1,10 @@
 import { afterEach, beforeEach, describe, expect, it, vi } from "bun:test";
-import { getKittyGraphics, type KittyGraphicsFeatures, setKittyGraphics } from "@oh-my-pi/pi-tui/kitty-graphics";
 import { ProcessTerminal } from "@oh-my-pi/pi-tui/terminal";
 import {
 	type CellDimensions,
 	getCellDimensions,
 	getTerminalInfo,
-	ImageProtocol,
 	setCellDimensions,
-	setTerminalImageProtocol,
-	TERMINAL,
 } from "@oh-my-pi/pi-tui/terminal-capabilities";
 
 const stdinIsTtyDescriptor = Object.getOwnPropertyDescriptor(process.stdin, "isTTY");
@@ -618,70 +614,6 @@ describe("ProcessTerminal DECRQM + in-band resize (DEC 2026/2048)", () => {
 	});
 });
 
-describe("ProcessTerminal Kitty graphics temp-file probe", () => {
-	const originalProtocol = TERMINAL.imageProtocol;
-	let originalGraphics: KittyGraphicsFeatures;
-
-	beforeEach(() => {
-		Object.defineProperty(process.stdin, "isTTY", { value: true, configurable: true });
-		Object.defineProperty(process.stdout, "isTTY", { value: true, configurable: true });
-		Object.defineProperty(process.stdin, "setRawMode", { value: vi.fn(), configurable: true });
-		originalGraphics = { ...getKittyGraphics() };
-		Bun.env.PI_TUI_KITTY_GRAPHICS_PROBE = "1";
-		setTerminalImageProtocol(ImageProtocol.Kitty);
-		setKittyGraphics({ transmissionMedium: "direct" });
-	});
-
-	afterEach(() => {
-		vi.restoreAllMocks();
-		restoreProperty(process.stdin, "isTTY", stdinIsTtyDescriptor);
-		restoreProperty(process.stdout, "isTTY", stdoutIsTtyDescriptor);
-		restoreProperty(process.stdin, "setRawMode", stdinSetRawModeDescriptor);
-		delete Bun.env.PI_TUI_KITTY_GRAPHICS_PROBE;
-		setTerminalImageProtocol(originalProtocol);
-		setKittyGraphics(originalGraphics);
-	});
-
-	function startProbed() {
-		const writes: string[] = [];
-		vi.spyOn(process, "kill").mockReturnValue(true);
-		vi.spyOn(process.stdin, "resume").mockImplementation(() => process.stdin);
-		vi.spyOn(process.stdin, "pause").mockImplementation(() => process.stdin);
-		vi.spyOn(process.stdin, "setEncoding").mockImplementation(() => process.stdin);
-		vi.spyOn(process.stdout, "write").mockImplementation(chunk => {
-			writes.push(typeof chunk === "string" ? chunk : chunk.toString());
-			return true;
-		});
-		const terminal = new ProcessTerminal();
-		terminal.start(
-			() => {},
-			() => {},
-		);
-		return { terminal, writes };
-	}
-
-	it("emits an a=q,t=t probe and promotes to temp-file transmission on OK", () => {
-		const { terminal, writes } = startProbed();
-		const probe = writes.find(w => w.includes("\x1b_Ga=q,t=t"));
-		expect(probe).toBeDefined();
-		const id = probe?.match(/i=(\d+)/)?.[1];
-		expect(id).toBeDefined();
-		expect(getKittyGraphics().transmissionMedium).toBe("direct");
-		process.stdin.emit("data", `\x1b_Gi=${id};OK\x1b\\`);
-		expect(getKittyGraphics().transmissionMedium).toBe("temp-file");
-		terminal.stop();
-	});
-
-	it("stays on direct transmission when the probe reports an error", () => {
-		const { terminal, writes } = startProbed();
-		const id = writes.find(w => w.includes("\x1b_Ga=q,t=t"))?.match(/i=(\d+)/)?.[1];
-		expect(id).toBeDefined();
-		process.stdin.emit("data", `\x1b_Gi=${id};ENOTSUP:bad\x1b\\`);
-		expect(getKittyGraphics().transmissionMedium).toBe("direct");
-		terminal.stop();
-	});
-});
-
 describe("OSC 66 text-sizing capability", () => {
 	it("advertises text sizing only for Kitty", () => {
 		// OSC 66 is a Kitty-only protocol; any other terminal must report the
diff --git a/packages/tui/test/terminal-capabilities.test.ts b/packages/tui/test/terminal-capabilities.test.ts
index 384b0a2cc..3d4d835f7 100644
--- a/packages/tui/test/terminal-capabilities.test.ts
+++ b/packages/tui/test/terminal-capabilities.test.ts
@@ -2,6 +2,7 @@ import { describe, expect, it } from "bun:test";
 import {
 	detectTerminalEagerEraseScrollbackRisk,
 	shouldEnableSynchronizedOutputByDefault,
+	synchronizedOutputUserOverride,
 } from "@oh-my-pi/pi-tui/terminal-capabilities";
 
 describe("terminal capability defaults", () => {
@@ -22,20 +23,93 @@ describe("terminal capability defaults", () => {
 	it("keeps native win32 on the dedicated ConPTY deferral path", () => {
 		expect(detectTerminalEagerEraseScrollbackRisk({ WT_SESSION: "abc" }, "win32")).toBe(false);
 	});
+});
 
-	it("disables synchronized output by default for remote, VTE, and unknown profiles", () => {
-		expect(shouldEnableSynchronizedOutputByDefault({ SSH_CONNECTION: "1 2 3 4" }, "linux", "base")).toBe(false);
-		expect(shouldEnableSynchronizedOutputByDefault({ VTE_VERSION: "6800" }, "linux", "base")).toBe(false);
-		expect(shouldEnableSynchronizedOutputByDefault({ TERM: "xterm-256color" }, "linux", "base")).toBe(false);
+describe("synchronizedOutputUserOverride", () => {
+	it("returns null when the user expresses no preference", () => {
+		expect(synchronizedOutputUserOverride({})).toBeNull();
+		expect(synchronizedOutputUserOverride({ TERM: "xterm-256color" })).toBeNull();
 	});
 
-	it("allows explicit synchronized-output force-on for diagnostics", () => {
+	it("returns false for either opt-out flag", () => {
+		expect(synchronizedOutputUserOverride({ PI_NO_SYNC_OUTPUT: "1" })).toBe(false);
+		expect(synchronizedOutputUserOverride({ PI_TUI_SYNC_OUTPUT: "0" })).toBe(false);
+	});
+
+	it("returns true for either force-on flag", () => {
+		expect(synchronizedOutputUserOverride({ PI_FORCE_SYNC_OUTPUT: "1" })).toBe(true);
+		expect(synchronizedOutputUserOverride({ PI_TUI_SYNC_OUTPUT: "1" })).toBe(true);
+	});
+
+	it("resolves opt-out ahead of force-on when both are set", () => {
+		expect(synchronizedOutputUserOverride({ PI_NO_SYNC_OUTPUT: "1", PI_FORCE_SYNC_OUTPUT: "1" })).toBe(false);
+		expect(synchronizedOutputUserOverride({ PI_TUI_SYNC_OUTPUT: "0", PI_FORCE_SYNC_OUTPUT: "1" })).toBe(false);
+	});
+});
+
+describe("shouldEnableSynchronizedOutputByDefault", () => {
+	it("enables sync for every known direct terminal, including Alacritty and VS Code", () => {
+		for (const id of ["kitty", "ghostty", "wezterm", "iterm2", "alacritty", "vscode"] as const) {
+			expect(shouldEnableSynchronizedOutputByDefault({}, id)).toBe(true);
+		}
+	});
+
+	it("enables sync in Windows Terminal / WSL via WT_SESSION regardless of terminal id", () => {
+		expect(shouldEnableSynchronizedOutputByDefault({ WT_SESSION: "abc" }, "trueColor")).toBe(true);
+		// WSL shape: Linux + WT_SESSION + COLORTERM=truecolor collapses to trueColor id.
+		expect(shouldEnableSynchronizedOutputByDefault({ WT_SESSION: "abc", COLORTERM: "truecolor" }, "trueColor")).toBe(
+			true,
+		);
+	});
+
+	it("enables sync when TERM_FEATURES advertises the Sy capability, even through SSH/mux", () => {
+		expect(shouldEnableSynchronizedOutputByDefault({ TERM_FEATURES: "ClSyTc" }, "base")).toBe(true);
+		expect(
+			shouldEnableSynchronizedOutputByDefault({ TERM_FEATURES: "ClSyTc", SSH_CONNECTION: "1 2 3 4" }, "base"),
+		).toBe(true);
+		expect(shouldEnableSynchronizedOutputByDefault({ TERM_FEATURES: "ClSyTc", TMUX: "1" }, "base")).toBe(true);
+	});
+
+	it("does not treat a TERM_FEATURES list without the Sy token as advertising support", () => {
+		expect(shouldEnableSynchronizedOutputByDefault({ TERM_FEATURES: "ClTc" }, "base")).toBe(false);
+	});
+
+	it("no longer blanket-disables SSH for recognized terminals", () => {
+		expect(shouldEnableSynchronizedOutputByDefault({ SSH_CONNECTION: "1 2 3 4" }, "iterm2")).toBe(true);
+		expect(shouldEnableSynchronizedOutputByDefault({ SSH_TTY: "/dev/pts/3" }, "kitty")).toBe(true);
+	});
+
+	it("keeps risky multiplexers off by default even when an inner terminal id leaks", () => {
+		expect(shouldEnableSynchronizedOutputByDefault({ TMUX: "1" }, "kitty")).toBe(false);
+		expect(shouldEnableSynchronizedOutputByDefault({ ZELLIJ: "0" }, "ghostty")).toBe(false);
+		expect(shouldEnableSynchronizedOutputByDefault({ STY: "x" }, "wezterm")).toBe(false);
+		expect(shouldEnableSynchronizedOutputByDefault({ TERM: "tmux-256color" }, "iterm2")).toBe(false);
+		expect(shouldEnableSynchronizedOutputByDefault({ TERM: "screen-256color" }, "kitty")).toBe(false);
+	});
+
+	it("keeps known-unsupported and unknown profiles off", () => {
+		expect(shouldEnableSynchronizedOutputByDefault({ VTE_VERSION: "6800" }, "base")).toBe(false);
+		expect(shouldEnableSynchronizedOutputByDefault({ TERM: "xterm-256color" }, "base")).toBe(false);
+		expect(shouldEnableSynchronizedOutputByDefault({}, "base")).toBe(false);
+		expect(shouldEnableSynchronizedOutputByDefault({}, "trueColor")).toBe(false);
+	});
+
+	it("lets a user opt-out beat every positive heuristic", () => {
+		expect(shouldEnableSynchronizedOutputByDefault({ PI_NO_SYNC_OUTPUT: "1" }, "kitty")).toBe(false);
+		expect(shouldEnableSynchronizedOutputByDefault({ PI_TUI_SYNC_OUTPUT: "0" }, "ghostty")).toBe(false);
 		expect(
 			shouldEnableSynchronizedOutputByDefault(
-				{ PI_FORCE_SYNC_OUTPUT: "1", SSH_CONNECTION: "1 2 3 4" },
-				"linux",
-				"base",
+				{ PI_NO_SYNC_OUTPUT: "1", WT_SESSION: "abc", TERM_FEATURES: "Sy" },
+				"kitty",
 			),
+		).toBe(false);
+	});
+
+	it("lets a user force-on beat the conservative defaults", () => {
+		expect(shouldEnableSynchronizedOutputByDefault({ PI_FORCE_SYNC_OUTPUT: "1" }, "base")).toBe(true);
+		expect(shouldEnableSynchronizedOutputByDefault({ PI_TUI_SYNC_OUTPUT: "1", TMUX: "1" }, "base")).toBe(true);
+		expect(
+			shouldEnableSynchronizedOutputByDefault({ PI_FORCE_SYNC_OUTPUT: "1", SSH_CONNECTION: "1 2 3 4" }, "base"),
 		).toBe(true);
 	});
 });
diff --git a/packages/tui/test/text-utils.test.ts b/packages/tui/test/text-utils.test.ts
index e118afde2..3f7c820e1 100644
--- a/packages/tui/test/text-utils.test.ts
+++ b/packages/tui/test/text-utils.test.ts
@@ -17,9 +17,6 @@ describe("text utils", () => {
 		expect(visibleWidth("a\tb")).toBe(1 + 3 + 1);
 	});
 
-	it("treats Arabic combining marks as zero-width", () => {
-		expect(visibleWidth("بَسِمَ")).toBe(3);
-	});
 	it("ignores OSC hyperlinks in visible width", () => {
 		const text = "\x1b]8;;https://example.com\x07link\x1b]8;;\x07";
 		expect(visibleWidth(text)).toBe(4);
diff --git a/packages/tui/test/virtual-terminal.ts b/packages/tui/test/virtual-terminal.ts
index f4c6d8945..b356d980a 100644
--- a/packages/tui/test/virtual-terminal.ts
+++ b/packages/tui/test/virtual-terminal.ts
@@ -134,7 +134,7 @@ export class VirtualTerminal implements Terminal {
 	}
 
 	stop(): void {
-		this.#engineWrite("\x1b[?2004l");
+		this.#engineWrite("\x1b[?2004l\x1b[?5522l");
 		this.#inputHandler = undefined;
 		this.#resizeHandler = undefined;
 	}
@@ -219,12 +219,12 @@ export class VirtualTerminal implements Terminal {
 
 	// --- Test-only helpers ---------------------------------------------------
 
-	/** Wait for TUI's throttled render pipeline to settle (matches the 16ms frame budget). */
+	/** Wait for TUI's throttled render pipeline to settle (matches the ~33ms frame budget). */
 	async waitForRender(): Promise<void> {
 		const nextTick = Promise.withResolvers<void>();
 		process.nextTick(nextTick.resolve);
 		await nextTick.promise;
-		await Bun.sleep(20);
+		await Bun.sleep(40);
 		await this.flush();
 	}
 
diff --git a/packages/tui/test/visible-width-jamo.test.ts b/packages/tui/test/visible-width-jamo.test.ts
deleted file mode 100644
index 818a544f1..000000000
--- a/packages/tui/test/visible-width-jamo.test.ts
+++ /dev/null
@@ -1,146 +0,0 @@
-/**
- * Regression guard for the Hangul Compatibility Jamo width correction in
- * `visibleWidthRaw`.
- *
- * `Bun.stringWidth` (and the underlying UAX#11 EAW tables) classify Hangul
- * Compatibility Jamo (U+3131..U+318E) as Wide (2 cells), but every macOS
- * terminal we ship to (Ghostty, Terminal.app, iTerm2) actually renders them
- * as a single cell. Without the correction, `#extractCursorPosition` doubles
- * the column count for every jamo emitted by a Korean IME during
- * composition, displacing the hardware cursor (and therefore the IME
- * candidate window) `N_jamo` cells past the actual glyph.
- *
- * Hangul Syllables (U+AC00..U+D7A3, e.g. `안`) are correctly 2 cells in both
- * Bun and the terminal — make sure the fix did NOT regress that. The
- * Halfwidth Hangul block (U+FFA0..U+FFDC) is already classified as Narrow
- * by Bun, so it does not appear in the correction and the test below is a
- * regression sanity check.
- */
-import { describe, expect, it } from "bun:test";
-import { Ellipsis, sliceWithWidth, truncateToWidth, visibleWidth } from "@oh-my-pi/pi-tui/utils";
-
-// The macOS-only correction (see PR #1410) keeps jamo at 1 cell on darwin;
-// every other platform follows UAX#11 and reports 2 cells per jamo.
-const JAMO_CELLS = process.platform === "darwin" ? 1 : 2;
-
-describe("visibleWidth — Hangul Compatibility Jamo correction", () => {
-	it("single compatibility jamo is 1 cell on darwin, 2 elsewhere", () => {
-		// U+3141 HANGUL LETTER MIEUM
-		expect(visibleWidth("ㅁ")).toBe(JAMO_CELLS);
-		// U+3134 HANGUL LETTER NIEUN
-		expect(visibleWidth("ㄴ")).toBe(JAMO_CELLS);
-		// U+3147 HANGUL LETTER IEUNG
-		expect(visibleWidth("ㅇ")).toBe(JAMO_CELLS);
-		// U+3142 HANGUL LETTER PIEUP
-		expect(visibleWidth("ㅂ")).toBe(JAMO_CELLS);
-		// U+3148 HANGUL LETTER JIEUJ
-		expect(visibleWidth("ㅈ")).toBe(JAMO_CELLS);
-	});
-
-	it("range edges U+3131 and U+318E follow platform width", () => {
-		// U+3131 HANGUL LETTER KIYEOK — first jamo in the block
-		expect(visibleWidth("\u3131")).toBe(JAMO_CELLS);
-		// U+318E HANGUL LETTER ARAEAE — last jamo in the block
-		expect(visibleWidth("\u318e")).toBe(JAMO_CELLS);
-	});
-
-	it("U+3164 HANGUL FILLER (inside the corrected range) follows platform width", () => {
-		// Often emitted by IME for empty-syllable placeholders. The filler is the
-		// one code point in the block that UAX#11 / `unicode-width` classify as
-		// zero-width, so off-darwin it measures 0 cells. On darwin the blanket
-		// jamo correction (U+3131..U+318E → 1) forces it to a single cell.
-		const fillerCells = process.platform === "darwin" ? 1 : 0;
-		expect(visibleWidth("\u3164")).toBe(fillerCells);
-	});
-
-	it("combining marks on compatibility jamo keep the platform base width", () => {
-		// Combining marks add no cells. The native scanner must keep
-		// UnicodeWidthStr's sequence rules, then apply the same local jamo
-		// correction used for standalone code points.
-		expect(visibleWidth("\u3141\u0301")).toBe(JAMO_CELLS);
-		const fillerCells = process.platform === "darwin" ? 1 : 0;
-		expect(visibleWidth("\u3164\u0301")).toBe(fillerCells);
-	});
-
-	it("string of 8 consecutive jamo is 8 cells on darwin, 16 elsewhere", () => {
-		// Matches the user-typed sequence in the v2 screen recording —
-		// before the macOS fix this returned 16 and produced an 8-cell gap.
-		expect(visibleWidth("ㅁㄴㅁㄴㅇㅂㄴㅂ")).toBe(8 * JAMO_CELLS);
-	});
-
-	it("Hangul Syllables (U+AC00..U+D7A3) stay at 2 cells", () => {
-		// `안` U+C548 — composed syllable, must remain 2 cells
-		expect(visibleWidth("안")).toBe(2);
-		// `녕` U+B155 — composed syllable
-		expect(visibleWidth("녕")).toBe(2);
-		// Whole word: 안녕 = 4 cells
-		expect(visibleWidth("안녕")).toBe(4);
-		// First & last in the block, for boundary coverage
-		expect(visibleWidth("\uac00")).toBe(2); // 가
-		expect(visibleWidth("\ud7a3")).toBe(2); // 힣
-	});
-
-	it("mixed ASCII + syllable + jamo strings add correctly", () => {
-		// a (1) + 안 (2) + ㅂ (J) + b (1) = 4 + J
-		expect(visibleWidth("a안ㅂb")).toBe(4 + JAMO_CELLS);
-		// 11 ASCII letters + 1 syllable + 4 jamo = 11 + 2 + 4*J
-		expect(visibleWidth("hello world안ㅁㄴㅇㅂ")).toBe(11 + 2 + 4 * JAMO_CELLS);
-	});
-
-	it("does not regress ASCII fast path or empty input", () => {
-		expect(visibleWidth("")).toBe(0);
-		expect(visibleWidth("hello")).toBe(5);
-		expect(visibleWidth("a")).toBe(1);
-		// Tab character (ASCII 0x09) inside the fast path expands to >2
-		expect(visibleWidth("a\tb")).toBeGreaterThan(2);
-	});
-
-	it("does not change width for other CJK characters", () => {
-		// Chinese: 漢字 (each 2 cells)
-		expect(visibleWidth("漢字")).toBe(4);
-		// Japanese hiragana: あい (each 2 cells)
-		expect(visibleWidth("あい")).toBe(4);
-		// Japanese katakana: アイ (each 2 cells)
-		expect(visibleWidth("アイ")).toBe(4);
-	});
-
-	it("Halfwidth Hangul block is unaffected (already Narrow in Bun)", () => {
-		// U+FFA1 HALFWIDTH HANGUL LETTER KIYEOK — Bun reports 1, untouched.
-		expect(visibleWidth("\uffa1")).toBe(1);
-		// U+FFDC HALFWIDTH HANGUL LETTER I
-		expect(visibleWidth("\uffdc")).toBe(1);
-	});
-});
-
-describe("native text helpers — Hangul Compatibility Jamo correction", () => {
-	// These exercise the Rust-side `char_width_corrected` wrapper in
-	// crates/pi-natives/src/text.rs. They will fail until the native
-	// binding is rebuilt (`bun run build:native`); CI rebuilds natives so
-	// they pass there. Mirrors the TS-side range U+3131..=U+318E.
-
-	it("sliceWithWidth treats jamo per platform width", () => {
-		// 8 jamo at JAMO_CELLS cells each must fit fully within the
-		// platform's natural cell count.
-		const input = "ㅁ".repeat(8);
-		const { text, width } = sliceWithWidth(input, 0, 8 * JAMO_CELLS, true);
-		expect(text).toBe(input);
-		expect(width).toBe(8 * JAMO_CELLS);
-	});
-
-	it("truncateToWidth keeps 8 jamo within their platform budget", () => {
-		// On darwin (1 cell/jamo) 8 jamo fit in 8 cells; off-darwin they need 16.
-		const result = truncateToWidth("ㅁ".repeat(20), 8 * JAMO_CELLS, Ellipsis.Omit);
-		// Strip any trailing pad to count jamo content.
-		const jamo = result.replaceAll(/[^\u3131-\u318E]/g, "");
-		expect(jamo.length).toBe(8);
-	});
-
-	it("native and TS visibleWidth agree on a jamo run", () => {
-		// Cross-layer parity guard: without the native fix, the TS path
-		// (Bun.stringWidth + manual correction) and the native path
-		// (unicode_width) disagreed by a factor of 2.
-		const input = "ㅁㄴㅇㅂㅈㄷㄱㅅ";
-		expect(visibleWidth(input)).toBe(8 * JAMO_CELLS);
-		expect(sliceWithWidth(input, 0, 8 * JAMO_CELLS, true).width).toBe(8 * JAMO_CELLS);
-	});
-});
diff --git a/packages/tui/test/visible-width.test.ts b/packages/tui/test/visible-width.test.ts
new file mode 100644
index 000000000..d1bfd8faa
--- /dev/null
+++ b/packages/tui/test/visible-width.test.ts
@@ -0,0 +1,77 @@
+/**
+ * `visibleWidth` measures terminal column width via `Bun.stringWidth` (a JSC
+ * builtin) instead of the native scanner, to keep the render loop off the
+ * N-API number-boxing path that traps under Bun 1.3.x GC pressure.
+ *
+ * Correctness contract: the result MUST equal the native engine's width for the
+ * same input, because `truncateToWidth` / `sliceWithWidth` / `wrapTextWithAnsi`
+ * cut text using that native model — any divergence makes padding / cursor math
+ * (`width - visibleWidth(...)`) drift. This guards the two corrections layered
+ * on top of `Bun.stringWidth` (tabs, OSC 66 scaling) and catches silent
+ * `Bun.stringWidth` width-table drift across Bun upgrades.
+ */
+import { describe, expect, it } from "bun:test";
+import { visibleWidth as nativeVisibleWidth } from "@oh-my-pi/pi-natives";
+import { getDefaultTabWidth, visibleWidth } from "@oh-my-pi/pi-tui/utils";
+
+const ESC = "\x1b";
+const ST = "\x1b\\";
+const BEL = "\x07";
+const TAB = getDefaultTabWidth();
+
+describe("visibleWidth — parity with the native width engine", () => {
+	const corpus: [string, string][] = [
+		["empty", ""],
+		["ascii", "Pending run: passed"],
+		["styled", `${ESC}[31mred${ESC}[0m text`],
+		["styled-truecolor", `${ESC}[38;2;1;2;3mx${ESC}[0m`],
+		["nested-sgr", `${ESC}[1m${ESC}[31mbold${ESC}[0m${ESC}[0m`],
+		["osc8-st", `${ESC}]8;;https://x.com${ST}link${ESC}]8;;${ST}`],
+		["osc8-bel", `${ESC}]8;;u${BEL}t${ESC}]8;;${BEL}`],
+		["cjk", "日本語のテキスト"],
+		["cjk-mixed", "abc中文def"],
+		["hangul-syllables", "안녕하세요"],
+		["styled-cjk", `${ESC}[1m漢字${ESC}[0m`],
+		["emoji", "👍 done"],
+		["emoji-zwj", "👨‍👩‍👧‍👦"],
+		["emoji-flag", "🇯🇵"],
+		["styled-zwj", `${ESC}[31m👨‍👩‍👧‍👦${ESC}[0m`],
+		["variation-selector", "▶️"],
+		["combining", "e\u0301"],
+		["ambiguous", "§±×→①②③"],
+		["box-drawing", "─│┌┐└┘"],
+		["fullwidth", "１２３"],
+		["halfwidth-kana", "ｱｲｳ"],
+		["rtl-arabic", "مرحبا"],
+		["thai", "สวัสดี"],
+		["tabs", "name\tvalue\tstatus"],
+		["leading-tabs", "\t\tindented"],
+		["osc66-scale", `${ESC}]66;s=2;big${ST}`],
+		["osc66-explicit-w", `${ESC}]66;w=5;Hi${BEL}`],
+		["osc66-scale-and-w", `${ESC}]66;s=3:w=4;X${ST}`],
+		["osc66-cjk", `${ESC}]66;s=2;日本${ST}`],
+		["osc66-inline", `pre ${ESC}]66;s=2;AB${ST} post`],
+		["osc66-multi", `${ESC}]66;s=2;A${ST} ${ESC}]66;s=3;B${ST}`],
+		["osc66-with-tabs", `\t${ESC}]66;s=2;X${ST}\t`],
+	];
+	for (const [name, input] of corpus) {
+		it(name, () => {
+			expect(visibleWidth(input)).toBe(nativeVisibleWidth(input, TAB));
+		});
+	}
+
+	it("strips ANSI (styled text measures as its plain content)", () => {
+		expect(visibleWidth(`${ESC}[31mhello${ESC}[0m`)).toBe(5);
+	});
+
+	it("expands each tab to the configured tab width", () => {
+		expect(visibleWidth("a\tb")).toBe(2 + TAB);
+		expect(visibleWidth("\t\t")).toBe(2 * TAB);
+	});
+
+	it("scales OSC 66 text-sizing payloads by `s=`", () => {
+		expect(visibleWidth(`${ESC}]66;s=2;big${ST}`)).toBe(6); // 2 * width("big")
+		expect(visibleWidth(`${ESC}]66;w=5;Hi${BEL}`)).toBe(5); // explicit width, scale 1
+		expect(visibleWidth(`${ESC}]66;s=3:w=4;X${ST}`)).toBe(12); // 3 * 4
+	});
+});
diff --git a/packages/utils/CHANGELOG.md b/packages/utils/CHANGELOG.md
index 23eee711b..5bb2baa09 100644
--- a/packages/utils/CHANGELOG.md
+++ b/packages/utils/CHANGELOG.md
@@ -2,6 +2,12 @@
 
 ## [Unreleased]
 
+## [15.10.0] - 2026-06-06
+
+### Changed
+
+- `logger.printTimings()` (the `PI_TIMING` startup tree) now surfaces two previously-invisible regions: a `(before instrumentation)` line for runtime init / uncaptured pre-marker work, and an `(unattributed self)` line for the root span's own untimed work so the gap between visible top-level spans and `Total` is no longer swallowed. `Total` is now labelled `(since first marker)` to make the window explicit. The restored `module-timer.ts` preload can feed module spans into the report: each module records `onLoad` → final top-level marker as `total`, a prepended body marker → final marker as `body/TLA`, and resolved static imports as a bounded dependency tree so the report separates graph wait from actual top-level module work.
+
 ## [15.9.2] - 2026-06-05
 
 ### Added
diff --git a/packages/utils/package.json b/packages/utils/package.json
index 4dc27e182..1368e3552 100644
--- a/packages/utils/package.json
+++ b/packages/utils/package.json
@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-utils",
-	"version": "15.9.4",
+	"version": "15.10.1",
 	"description": "Shared utilities for pi packages",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
diff --git a/packages/utils/src/logger.ts b/packages/utils/src/logger.ts
index ae0f7fd36..1591e1620 100644
--- a/packages/utils/src/logger.ts
+++ b/packages/utils/src/logger.ts
@@ -15,6 +15,7 @@ import { isPromise } from "node:util/types";
 import winston from "winston";
 import DailyRotateFile from "winston-daily-rotate-file";
 import { getLogsDir } from "./dirs";
+import { drainModuleLoadEvents } from "./timing-buffer";
 
 /** Ensure a logs directory exists; return the resolved path. */
 function ensureDir(dir: string): string {
@@ -166,12 +167,36 @@ interface Span {
 	children: Span[];
 	/** Marker / point event without a duration. */
 	point?: boolean;
+	/** Absolute module path for module-load spans. */
+	modulePath?: string;
+	/** Own top-level module body / TLA duration for module-load spans. */
+	moduleBodyMs?: number;
+	/** Resolved static imports for module-load spans. */
+	moduleImports?: string[];
 }
-
 const spanStorage = new AsyncLocalStorage<Span>();
 let gRootSpan: Span | undefined;
 let gRecordTimings = false;
 
+export function timingModeIncludes(option: "full" | "x"): boolean {
+	const value = process.env.PI_TIMING;
+	if (!value) return false;
+	if (value === option) return true;
+	let start = 0;
+	for (let i = 0; i <= value.length; i++) {
+		const code = i === value.length ? 44 : value.charCodeAt(i);
+		const separator = code === 44 || code === 58 || code === 59 || code === 43 || code <= 32;
+		if (!separator) continue;
+		if (i > start && value.slice(start, i) === option) return true;
+		start = i + 1;
+	}
+	return false;
+}
+
+export function shouldExitAfterTimings(): boolean {
+	return timingModeIncludes("x") || timingModeIncludes("full");
+}
+
 /**
  * Print collected timings as an indented tree.
  * Each span shows wall duration; parents with children also show "(self)" for unattributed time.
@@ -184,9 +209,22 @@ export function printTimings(): void {
 	}
 
 	gRootSpan.end = performance.now();
+	// Splice any preload-captured module-load events into the tree as root
+	// children and back-extend the root window over them, so the static-import
+	// phase that ran before the first explicit marker becomes visible (the
+	// `(modules)` summary below) instead of being lumped into the opaque
+	// `(before instrumentation)` figure.
+	spliceModuleLoadBuffer();
 	const lines: string[] = [];
 	lines.push("");
 	lines.push("--- Startup timings (hierarchical) ---");
+	// performance.now() shares the process-start origin, so the root span's start
+	// is the wall time before the first marker — runtime init plus any module
+	// loads not captured below. With the module-load preload active this shrinks
+	// to ~runtime init because the load phase is back-folded into the window.
+	if (gRootSpan.start > LOGGED_TIMING_THRESHOLD_MS) {
+		lines.push(`(before instrumentation): ${fmtMs(gRootSpan.start)} [runtime init + module load]`);
+	}
 	const work: Span[] = [];
 	const loads: Span[] = [];
 	for (const child of gRootSpan.children) {
@@ -199,8 +237,14 @@ export function printTimings(): void {
 	if (loads.length > 0) {
 		printModuleLoadSummary(loads, 0, lines);
 	}
+	// Surface the root's own unattributed time so the gap between the visible
+	// top-level spans and Total isn't silently swallowed.
+	const rootSelf = selfTimeOf(gRootSpan);
+	if (gRootSpan.children.length > 0 && rootSelf > LOGGED_TIMING_THRESHOLD_MS) {
+		lines.push(`(unattributed self): ${fmtMs(rootSelf)}`);
+	}
 	const totalMs = (gRootSpan.end - gRootSpan.start).toFixed(1);
-	lines.push(`Total: ${totalMs}ms`);
+	lines.push(`Total: ${totalMs}ms (since first marker)`);
 	lines.push("--------------------------------------");
 	lines.push("");
 	console.error(lines.join("\n"));
@@ -209,8 +253,8 @@ export function printTimings(): void {
 
 /**
  * Begin recording startup timings under a new root span.
- * Idempotent: a second call while already recording is a no-op so that side-effect
- * starters (see module-timer.ts) and explicit starters (main.ts) can coexist.
+ * Idempotent: a second call while already recording is a no-op, so an explicit
+ * starter (main.ts) and any future early starter can coexist.
  */
 export function startTiming(): void {
 	if (gRecordTimings) return;
@@ -225,10 +269,16 @@ export function startTiming(): void {
 
 /**
  * Record an externally-measured span as a leaf child of the active span (or root
- * when no span is active). Used by the module-load timing plugin to splice load
- * events into the tree retroactively.
+ * when no span is active). Used by {@link spliceModuleLoadBuffer} to fold
+ * preload-captured module windows into the tree.
  */
-export function recordModuleLoadSpan(path: string, start: number, durationMs: number): void {
+export function recordModuleLoadSpan(
+	path: string,
+	start: number,
+	durationMs: number,
+	bodyMs?: number,
+	imports: string[] = [],
+): void {
 	if (!gRecordTimings || !gRootSpan) return;
 	const parent = spanStorage.getStore() ?? gRootSpan;
 	const span: Span = {
@@ -237,10 +287,32 @@ export function recordModuleLoadSpan(path: string, start: number, durationMs: nu
 		end: start + durationMs,
 		parent,
 		children: [],
+		modulePath: path,
+		moduleBodyMs: bodyMs,
+		moduleImports: imports,
 	};
 	parent.children.push(span);
 }
 
+/**
+ * Drain the preload's module-load buffer (see module-timer.ts) into the tree as
+ * `load:` children of the root, then back-extend the root window to the earliest
+ * captured read so the pre-marker load phase is counted in Total rather than
+ * hidden as `(before instrumentation)`. No-op when nothing was captured (e.g. no
+ * `--preload`, or a compiled binary where module reads are not interceptable).
+ */
+function spliceModuleLoadBuffer(): void {
+	if (!gRootSpan) return;
+	const events = drainModuleLoadEvents();
+	if (events.length === 0) return;
+	let earliest = gRootSpan.start;
+	for (const event of events) {
+		recordModuleLoadSpan(event.path, event.start, event.durationMs, event.bodyMs, event.imports);
+		if (event.start < earliest) earliest = event.start;
+	}
+	gRootSpan.start = earliest;
+}
+
 function shortenLoadPath(p: string): string {
 	const cwd = process.cwd();
 	if (p.startsWith(`${cwd}/`)) return p.slice(cwd.length + 1);
@@ -296,6 +368,16 @@ function fmtMs(ms: number): string {
 
 const MODULE_LOAD_PREFIX = "load:";
 const MODULE_LOAD_VERBOSE_TOP = 10;
+const MODULE_TREE_MAX_DEPTH = 5;
+const MODULE_TREE_ROOT_TOP = 5;
+const MODULE_TREE_CHILD_TOP = 8;
+
+interface ModuleTimingNode {
+	span: Span;
+	children: ModuleTimingNode[];
+	parents: number;
+	body: number;
+}
 
 function isModuleLoadSpan(span: Span): boolean {
 	return span.op.startsWith(MODULE_LOAD_PREFIX);
@@ -330,33 +412,118 @@ function printSpan(span: Span, depth: number, lines: string[]): void {
 	}
 }
 
-/** Collapse the (typically hundreds of) module-load spans into one summary line. */
+/** Render module-load spans as a dependency-aware DAG/tree. */
 function printModuleLoadSummary(loads: Span[], depth: number, lines: string[]): void {
 	const childIndent = "  ".repeat(depth);
 	const grandIndent = "  ".repeat(depth + 1);
 	let unionStart = Number.POSITIVE_INFINITY;
 	let unionEnd = 0;
-	let totalSelf = 0;
 	for (const span of loads) {
 		if (span.end === undefined) continue;
 		if (span.start < unionStart) unionStart = span.start;
 		if (span.end > unionEnd) unionEnd = span.end;
-		totalSelf += span.end - span.start;
 	}
 	const wall = unionEnd > unionStart ? unionEnd - unionStart : 0;
-	lines.push(`${childIndent}(modules): ${loads.length} loaded, wall ${fmtMs(wall)}, sum ${fmtMs(totalSelf)}`);
-	const showAll = process.env.PI_TIMING === "full";
-	const sorted = [...loads].sort((a, b) => durationOf(b) - durationOf(a));
-	const visible = showAll ? sorted : sorted.slice(0, MODULE_LOAD_VERBOSE_TOP);
-	for (const span of visible) {
-		const dur = durationOf(span);
-		if (dur < LOGGED_TIMING_THRESHOLD_MS) break;
-		const tag = isParallel(span) ? " [parallel]" : "";
-		lines.push(`${grandIndent}${span.op}: ${fmtMs(dur)}${tag}`);
+	const nodes = buildModuleTimingGraph(loads);
+	lines.push(`${childIndent}(modules): ${loads.length} loaded, wall ${fmtMs(wall)}`);
+	if (nodes.length === 0) return;
+
+	const showAll = timingModeIncludes("full");
+	const byBody = [...nodes].sort(compareModuleNodes);
+	const topBody = showAll ? byBody : byBody.slice(0, MODULE_LOAD_VERBOSE_TOP);
+	lines.push(`${grandIndent}top body/TLA:`);
+	for (const node of topBody) {
+		if (!showAll && node.body < LOGGED_TIMING_THRESHOLD_MS) break;
+		lines.push(`${grandIndent}  ${node.span.op}: body ${fmtMs(node.body)} (total ${fmtMs(durationOf(node.span))})`);
 	}
-	if (!showAll && sorted.length > MODULE_LOAD_VERBOSE_TOP) {
-		lines.push(`${grandIndent}… ${sorted.length - MODULE_LOAD_VERBOSE_TOP} more (PI_TIMING=full to show all)`);
+	if (!showAll && byBody.length > MODULE_LOAD_VERBOSE_TOP) {
+		lines.push(`${grandIndent}  … ${byBody.length - MODULE_LOAD_VERBOSE_TOP} more (PI_TIMING=full to show all)`);
 	}
+
+	const roots = nodes.filter(node => node.parents === 0);
+	const treeRoots = (roots.length > 0 ? roots : nodes).sort((a, b) => durationOf(b.span) - durationOf(a.span));
+	const visibleRoots = showAll ? treeRoots : treeRoots.slice(0, MODULE_TREE_ROOT_TOP);
+	lines.push(`${grandIndent}tree:`);
+	const rendered = new Set<string>();
+	for (const node of visibleRoots) {
+		renderModuleTimingNode(node, depth + 2, lines, rendered, new Set<string>(), showAll);
+	}
+	if (!showAll && treeRoots.length > MODULE_TREE_ROOT_TOP) {
+		lines.push(
+			`${grandIndent}  … ${treeRoots.length - MODULE_TREE_ROOT_TOP} more roots (PI_TIMING=full to show all)`,
+		);
+	}
+}
+
+function buildModuleTimingGraph(loads: Span[]): ModuleTimingNode[] {
+	const nodes = new Map<string, ModuleTimingNode>();
+	for (const span of loads) {
+		if (!span.modulePath || span.end === undefined) continue;
+		nodes.set(span.modulePath, { span, children: [], parents: 0, body: span.moduleBodyMs ?? 0 });
+	}
+	for (const node of nodes.values()) {
+		for (const childPath of node.span.moduleImports ?? []) {
+			const child = nodes.get(childPath);
+			if (!child || child === node) continue;
+			node.children.push(child);
+			child.parents++;
+		}
+	}
+	for (const node of nodes.values()) {
+		node.children.sort(compareModuleNodes);
+	}
+	return [...nodes.values()];
+}
+
+function compareModuleNodes(a: ModuleTimingNode, b: ModuleTimingNode): number {
+	const bodyDiff = b.body - a.body;
+	if (Math.abs(bodyDiff) > 0.001) return bodyDiff;
+	return durationOf(b.span) - durationOf(a.span);
+}
+
+function renderModuleTimingNode(
+	node: ModuleTimingNode,
+	depth: number,
+	lines: string[],
+	rendered: Set<string>,
+	ancestors: Set<string>,
+	showAll: boolean,
+): void {
+	const path = node.span.modulePath;
+	if (!path) return;
+	const indent = "  ".repeat(depth);
+	const total = durationOf(node.span);
+	if (!showAll && total < LOGGED_TIMING_THRESHOLD_MS && node.children.length === 0) return;
+	const wait = Math.max(0, total - node.body);
+	const shared = node.parents > 1 ? " [shared]" : "";
+	const timing =
+		node.body > LOGGED_TIMING_THRESHOLD_MS || node.children.length > 0
+			? ` (body ${fmtMs(node.body)}, wait ${fmtMs(wait)})`
+			: "";
+	const alreadyRendered = rendered.has(path);
+	const cycle = ancestors.has(path);
+	const suffix = cycle ? " [cycle]" : alreadyRendered ? " [already shown]" : "";
+	lines.push(`${indent}${node.span.op}: ${fmtMs(total)}${timing}${shared}${suffix}`);
+	if (cycle || alreadyRendered) return;
+	rendered.add(path);
+	ancestors.add(path);
+	if (!showAll && ancestors.size >= MODULE_TREE_MAX_DEPTH) {
+		if (node.children.length > 0) {
+			lines.push(`${indent}  … ${node.children.length} imports deeper (PI_TIMING=full to show all)`);
+		}
+		ancestors.delete(path);
+		return;
+	}
+	const visibleChildren = showAll ? node.children : node.children.slice(0, MODULE_TREE_CHILD_TOP);
+	for (const child of visibleChildren) {
+		renderModuleTimingNode(child, depth + 1, lines, rendered, ancestors, showAll);
+	}
+	if (!showAll && node.children.length > MODULE_TREE_CHILD_TOP) {
+		lines.push(
+			`${indent}  … ${node.children.length - MODULE_TREE_CHILD_TOP} more imports (PI_TIMING=full to show all)`,
+		);
+	}
+	ancestors.delete(path);
 }
 
 /** A span is parallel if it overlaps a sibling that started before it. */
diff --git a/packages/utils/src/module-timer.ts b/packages/utils/src/module-timer.ts
new file mode 100644
index 000000000..b0ed91d39
--- /dev/null
+++ b/packages/utils/src/module-timer.ts
@@ -0,0 +1,148 @@
+/**
+ * Module-load timing preload.
+ *
+ * `bun --preload .../module-timer.ts <entry>` installs Bun plugin hooks (only
+ * when `PI_TIMING` is set) that record an inclusive module window plus resolved
+ * static child edges:
+ *
+ *   onLoad start → appended end marker after the module's top-level body
+ *
+ * Events are pushed into a process-global buffer that {@link logger.printTimings}
+ * drains and renders as a module DAG/tree. Each module row can therefore show
+ * both total time and `self` time after subtracting child module intervals.
+ *
+ * Why a preload (and not a normal import): Bun reads the *entire* statically
+ * reachable graph before evaluating any module, so hooks installed from inside
+ * that graph cannot observe its own loading — they only catch later dynamically
+ * loaded modules. A preload runs first, so it sees the static-import phase that
+ * dominates startup.
+ *
+ * Kept dependency-free on purpose: the sole import is Bun's `plugin`, so this is
+ * cheap to preload before pi-utils (and winston) exist. The buffer is shared with
+ * the logger via a registry Symbol so neither side needs to import the other.
+ *
+ * **What is measured:** an inclusive per-module window. `onLoad` stamps the
+ * start before reading source; the returned source has a tiny marker appended at
+ * the end of the module. That marker runs after Bun parses/transpiles the module
+ * and after any top-level await in that module completes, so the duration
+ * includes read + parse/transpile + dependency wait + top-level execution/TLA.
+ * If a module throws before its final statement, no end marker is recorded.
+ *
+ * **Tree shape:** `onResolve` observes importer → specifier edges and resolves
+ * them with `Bun.resolveSync` without taking over Bun's real resolution. The
+ * logger renders these edges as a DAG/tree and computes module `self` time by
+ * subtracting the union of child intervals, avoiding misleading flat inclusive
+ * totals.
+ *
+ * **Coverage limits:**
+ * - TS/TSX only — intercepting `node_modules` CJS `.js`/`.cjs` and forcing ESM
+ *   breaks their default-export detection, so they are left to Bun's default path.
+ * - **Dev runs only.** In the compiled `omp` binary every module is pre-bundled
+ *   into bunfs, so `onLoad` never fires; profile with a `bun --preload` dev run.
+ */
+import { plugin } from "bun";
+import { moduleLoadBuffer } from "./timing-buffer";
+
+// Restrict to TS/TSX only. node_modules ships CommonJS `.js`/`.cjs` that Bun
+// auto-detects when loaded via its default path; if we intercept and return
+// `{ contents, loader: "js" }`, Bun forces ESM and CJS modules fail to load
+// (e.g. `Missing 'default' export`). Our own source tree (where the interesting
+// timing lives) is uniformly TypeScript, so a TS-only filter is both safe and
+// sufficient.
+const MODULE_LOADER_FILTER = /\.[mc]?tsx?$/;
+const MODULE_COMPLETE_KEY: symbol = Symbol.for("omp.moduleLoadComplete");
+const MODULE_BODY_START_KEY: symbol = Symbol.for("omp.moduleBodyStart");
+const STATIC_IMPORT_PATTERN =
+	/\b(?:import|export)\s+(?:type\s+)?(?:[^"']*?\s+from\s+)?["']([^"']+)["']|\bimport\s*\(\s*["']([^"']+)["']\s*\)/g;
+
+type CompleteStore = Record<symbol, ((path: string) => void) | undefined>;
+
+function bodyStartMarker(path: string): string {
+	return `;globalThis[Symbol.for("omp.moduleBodyStart")]?.(${JSON.stringify(path)});\n`;
+}
+
+function completionMarker(path: string): string {
+	return `\n;globalThis[Symbol.for("omp.moduleLoadComplete")]?.(${JSON.stringify(path)});\n`;
+}
+
+function instrumentContents(path: string, contents: string): string {
+	const start = bodyStartMarker(path);
+	const end = completionMarker(path);
+	if (!contents.startsWith("#!")) return `${start}${contents}${end}`;
+	const newline = contents.indexOf("\n");
+	if (newline === -1) return `${contents}\n${start}${end}`;
+	return `${contents.slice(0, newline + 1)}${start}${contents.slice(newline + 1)}${end}`;
+}
+function importerDir(importer: string): string {
+	const slash = importer.lastIndexOf("/");
+	if (slash === -1) return ".";
+	return importer.slice(0, slash);
+}
+
+function childSetFor(importsByPath: Map<string, Set<string>>, path: string): Set<string> {
+	let children = importsByPath.get(path);
+	if (!children) {
+		children = new Set<string>();
+		importsByPath.set(path, children);
+	}
+	return children;
+}
+
+function addImportEdges(importsByPath: Map<string, Set<string>>, importer: string, contents: string): void {
+	STATIC_IMPORT_PATTERN.lastIndex = 0;
+	for (const match of contents.matchAll(STATIC_IMPORT_PATTERN)) {
+		const specifier = match[1] ?? match[2];
+		if (!specifier) continue;
+		try {
+			const resolved = Bun.resolveSync(specifier, importerDir(importer));
+			if (MODULE_LOADER_FILTER.test(resolved) && resolved !== importer) {
+				childSetFor(importsByPath, importer).add(resolved);
+			}
+		} catch {
+			// Leave Bun's real resolver/runtime to surface any error. This scanner is only an observer.
+		}
+	}
+}
+
+if (process.env.PI_TIMING) {
+	const buffer = moduleLoadBuffer();
+	const starts = new Map<string, number>();
+	const bodyStarts = new Map<string, number>();
+	const importsByPath = new Map<string, Set<string>>();
+	const store = globalThis as unknown as CompleteStore;
+	store[MODULE_BODY_START_KEY] = (path: string): void => {
+		bodyStarts.set(path, performance.now());
+	};
+	store[MODULE_COMPLETE_KEY] = (path: string): void => {
+		const start = starts.get(path);
+		if (start === undefined) return;
+		starts.delete(path);
+		const end = performance.now();
+		const bodyStart = bodyStarts.get(path);
+		bodyStarts.delete(path);
+		const imports = importsByPath.get(path);
+		buffer.push({
+			path,
+			start,
+			durationMs: end - start,
+			bodyMs: bodyStart === undefined ? undefined : end - bodyStart,
+			imports: imports ? [...imports] : [],
+		});
+	};
+
+	plugin({
+		name: "pi-module-load-timer",
+		setup(build) {
+			build.onLoad({ filter: MODULE_LOADER_FILTER }, async args => {
+				starts.set(args.path, performance.now());
+				childSetFor(importsByPath, args.path);
+				const contents = await Bun.file(args.path).text();
+				addImportEdges(importsByPath, args.path, contents);
+				return {
+					contents: instrumentContents(args.path, contents),
+					loader: args.path.endsWith(".tsx") ? "tsx" : "ts",
+				};
+			});
+		},
+	});
+}
diff --git a/packages/utils/src/timing-buffer.ts b/packages/utils/src/timing-buffer.ts
new file mode 100644
index 000000000..4208f9298
--- /dev/null
+++ b/packages/utils/src/timing-buffer.ts
@@ -0,0 +1,47 @@
+/**
+ * Shared contract between the {@link module-timer} preload and {@link logger}'s
+ * timing tree. Kept in its own dependency-free module so the preload can import
+ * it without pulling in winston (via logger) and the logger can drain the buffer
+ * without importing the Bun-plugin preload.
+ */
+
+export interface ModuleLoadEvent {
+	/** Absolute or Bun-resolved module path. */
+	path: string;
+	/** `performance.now()` timestamp captured at Bun `onLoad` entry. */
+	start: number;
+	/** Inclusive module window: `onLoad` entry → appended final marker. */
+	durationMs: number;
+	/** Own top-level body / TLA time: prepended body marker → appended final marker. */
+	bodyMs?: number;
+	/** Resolved static children imported by this module. */
+	imports: string[];
+}
+
+/**
+ * Registry-global key under which the preload accumulates module-load events.
+ * `Symbol.for` so both modules resolve the same symbol independently.
+ */
+const KEY: symbol = Symbol.for("omp.moduleLoadBuffer");
+
+type Store = Record<symbol, ModuleLoadEvent[] | undefined>;
+
+/** The append-only buffer the preload pushes into (created on first access). */
+export function moduleLoadBuffer(): ModuleLoadEvent[] {
+	const store = globalThis as unknown as Store;
+	let buffer = store[KEY];
+	if (!buffer) {
+		buffer = [];
+		store[KEY] = buffer;
+	}
+	return buffer;
+}
+
+/** Drain and return all buffered events, leaving the buffer empty. */
+export function drainModuleLoadEvents(): ModuleLoadEvent[] {
+	const store = globalThis as unknown as Store;
+	const buffer = store[KEY];
+	if (!buffer || buffer.length === 0) return [];
+	store[KEY] = [];
+	return buffer;
+}
diff --git a/scripts/release.ts b/scripts/release.ts
index 95fb46b6a..12ecfff5c 100755
--- a/scripts/release.ts
+++ b/scripts/release.ts
@@ -328,38 +328,37 @@ async function cmdRelease(version: string): Promise<void> {
 	await git(["commit", "-m", `chore: bump version to ${version}`]);
 	console.log();
 
-	// 8. Tag + push atomically.
+	// 8. Tag, then push branch + tag atomically — pushing the tag by object id.
 	//
-	// Background `git maintenance run` (scheduled via the global `[maintenance]
-	// repo = …` list) fetches origin with `fetch.pruneTags=true` set globally,
-	// which deletes any local tag that does not yet exist on the remote — i.e.
-	// the brand-new release tag. The `-c fetch.pruneTags=false` we pass to our
-	// git wrapper only applies to our git invocations, not to the concurrent
-	// maintenance process, so we have to defend against the race ourselves:
-	// (re)create the tag immediately before the push and retry on the specific
-	// "src refspec … does not match any" symptom that means it got pruned.
+	// This repo is in the global `[maintenance] repo = …` list, so a scheduled
+	// `git maintenance run` fetches origin with `fetch.pruneTags=true` (set
+	// globally) and deletes any local tag not yet on the remote — i.e. the
+	// brand-new release tag. The `-c fetch.pruneTags=false` on our git wrapper
+	// only governs our own git calls, not the concurrent maintenance process, so
+	// a local tag ref may vanish before or while the push resolves it.
+	//
+	// A bare push refspec (`refs/tags/v…` with no `:dst`) re-resolves the tag on
+	// disk during refspec matching (git's remote.c:match_explicit); if the prune
+	// lands in that window git dies with
+	// "refs/tags/v… cannot be resolved to branch", and if it lands before the
+	// push it dies with "src refspec … does not match any". We sidestep both by
+	// pushing the HEAD commit object id straight into the remote tag ref
+	// (`<sha>:refs/tags/v…`): the push has no dependency on a local tag, and the
+	// commit is reachable from main so maintenance cannot prune it. The local
+	// tag we still create is only for `git describe`; losing it is harmless. The
+	// default Git LFS pre-push hook uploads the branch's LFS objects as part of
+	// this same atomic push — no separate `git lfs push` is needed.
 	console.log("Tagging and pushing to remote...");
 	const tagRef = `v${version}`;
-	for (let attempt = 1; ; attempt++) {
-		await git(["tag", "-f", tagRef]);
-		const result = await git([
-			"push",
-			"--atomic",
-			"origin",
-			"main",
-			`refs/tags/${tagRef}`,
-		]).nothrow();
-		if (result.exitCode === 0) break;
-		const stderr = result.stderr.toString();
-		process.stderr.write(stderr);
-		const pruned = /src refspec .* does not match any/.test(stderr);
-		if (!pruned || attempt >= 3) {
-			throw new Error(`git push failed for ${tagRef} (attempt ${attempt})`);
-		}
-		console.warn(
-			`  Tag ${tagRef} pruned by background maintenance, retrying (${attempt + 1}/3)...`,
-		);
-	}
+	const sha = (await git(["rev-parse", "HEAD"]).text()).trim();
+	await git(["tag", "-f", tagRef]);
+	await git([
+		"push",
+		"--atomic",
+		"origin",
+		"refs/heads/main:refs/heads/main",
+		`${sha}:refs/tags/${tagRef}`,
+	]);
 	console.log();
 
 	// 9. Watch CI
@@ -372,7 +371,9 @@ async function cmdRelease(version: string): Promise<void> {
 		console.log("\nTo retry after fixing (repeat until CI passes):");
 		console.log("  git commit -m \"fix: <brief description>\"");
 		console.log(`  git tag -f v${version}`);
-		console.log(`  git push --atomic origin main +refs/tags/v${version}`);
+		console.log(
+			`  git push --atomic origin refs/heads/main:refs/heads/main "+$(git rev-parse HEAD):refs/tags/v${version}"`,
+		);
 		console.log("  bun scripts/release.ts watch");
 		process.exit(1);
 	}
diff --git a/tsconfig.base.json b/tsconfig.base.json
index c7f4cfcbf..0639244e8 100644
--- a/tsconfig.base.json
+++ b/tsconfig.base.json
@@ -10,20 +10,10 @@
 		"allowArbitraryExtensions": true,
 		"verbatimModuleSyntax": true,
 		"noEmit": true,
-		"emitDeclarationOnly": false,
-		"declaration": true,
-		"declarationMap": true,
-		"sourceMap": true,
-		"inlineSources": true,
-		"inlineSourceMap": false,
 		"resolveJsonModule": true,
 		"esModuleInterop": true,
 		"forceConsistentCasingInFileNames": true,
-		"experimentalDecorators": true,
-		"emitDecoratorMetadata": true,
-		"useDefineForClassFields": false,
 		"types": ["bun", "assets"],
-		"typeRoots": ["./types", "./node_modules/@types"],
-		"assumeChangesOnlyAffectDirectDependencies": true
+		"typeRoots": ["./types", "./node_modules/@types"]
 	}
 }