From b191e6c5ff73ee3ba73ea14cbb0d033080e19df8 Mon Sep 17 00:00:00 2001
From: can1357 <me@can.ac>
Date: Tue, 19 May 2026 19:18:44 +0900
Subject: [PATCH] fix(coding-agent): reject task-N ID confusion, error on
 partial todo apply

- todo-write.md adds an explicit note that tasks are referenced by
  verbatim content text; the tool never emits task-N IDs.
- resolveTaskOrError rejects ^task-\d+$ inputs with a clarifying error.
- execute sets isError:true when any op failed.
- appendItems short-circuits on the first "already exists" error so the
  call no longer applies the prefix of a doomed batch.
---
 .../coding-agent/src/prompts/tools/todo-write.md  |  2 ++
 packages/coding-agent/src/tools/todo-write.ts     | 15 +++++++++++----
 2 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/packages/coding-agent/src/prompts/tools/todo-write.md b/packages/coding-agent/src/prompts/tools/todo-write.md
index cfe67f542..34221d17c 100644
--- a/packages/coding-agent/src/prompts/tools/todo-write.md
+++ b/packages/coding-agent/src/prompts/tools/todo-write.md
@@ -1,3 +1,5 @@
+**Tasks are referenced by their verbatim content string, not by any auto-generated ID. There is no "task-1"/"task-N" identifier — the tool never emits one. Pass the task's content text in the `task` field.**
+
 Manages a phased task list. Pass `ops`: a flat array of operations.
 The next pending task is auto-promoted to `in_progress` after each completion.
 Allowed `op` values are only `init`, `start`, `done`, `drop`, `rm`, `append`, and `note`. `pending` is a task status, not an `op`; leave not-yet-started tasks implicit in `init`/`append` lists.
diff --git a/packages/coding-agent/src/tools/todo-write.ts b/packages/coding-agent/src/tools/todo-write.ts
index ba9f8e56c..5541af834 100644
--- a/packages/coding-agent/src/tools/todo-write.ts
+++ b/packages/coding-agent/src/tools/todo-write.ts
@@ -150,9 +150,15 @@ function resolveTaskOrError(
 	}
 	const hit = findTaskByContent(phases, content);
 	if (!hit) {
-		const totalTasks = phases.reduce((sum, phase) => sum + phase.tasks.length, 0);
-		const hint = totalTasks === 0 ? " (todo list is empty — was it replaced or not yet created?)" : "";
-		errors.push(`Task "${content}" not found${hint}`);
+		if (/^task-\d+$/.test(content)) {
+			errors.push(
+				`Task "${content}" not found. Tasks are referenced by content, not by IDs — pass the task's full text from the previous result.`,
+			);
+		} else {
+			const totalTasks = phases.reduce((sum, phase) => sum + phase.tasks.length, 0);
+			const hint = totalTasks === 0 ? " (todo list is empty — was it replaced or not yet created?)" : "";
+			errors.push(`Task "${content}" not found${hint}`);
+		}
 	}
 	return hit;
 }
@@ -209,7 +215,7 @@ function appendItems(phases: TodoPhase[], entry: TodoOpEntryValue, errors: strin
 	for (const content of entry.items) {
 		if (findTaskByContent(phases, content)) {
 			errors.push(`Task "${content}" already exists`);
-			continue;
+			return phases;
 		}
 		phase.tasks.push({ content, status: "pending" });
 	}
@@ -513,6 +519,7 @@ export class TodoWriteTool implements AgentTool<typeof todoWriteSchema, TodoWrit
 		return {
 			content: [{ type: "text", text: formatSummary(updated, errors) }],
 			details: { phases: updated, storage },
+			isError: errors.length > 0 ? true : undefined,
 		};
 	}
 }