feat(coding-agent): introduced fine grained per agent advisor configuration
- Replaced blanket subagent advisor global settings with fine-grained per-agent configuration and frontmatter support. - Added dashboard keybindings and inline override editors for managing agent advisor patterns. - Implemented settings migration logic to convert legacy global options into per-agent settings. - Updated session persistence and execution layers to restore and enforce per-agent advisor behaviors.
This commit is contained in:
can1357
@@ -294,12 +294,14 @@ Fields:
|
||||
|
||||
## Subagents
|
||||
|
||||
`advisor.subagents` controls whether spawned task/eval subagents also get an advisor runtime.
|
||||
Subagents run unadvised by default; advisors are opted in **per agent** instead of via a blanket toggle:
|
||||
|
||||
- `false` (default): only the main session can run an advisor.
|
||||
- `true`: eligible subagent sessions build their own advisor subsystem with the same settings/model-role resolution, then rerun both `WATCHDOG.md` and `WATCHDOG.yml` discovery for that subagent session's `cwd` and agent directory.
|
||||
- Agent definition frontmatter `advisor`: `true` advises spawned sessions of that agent with the model resolved for the `advisor` role; a string (e.g. `advisor: "deepseek/deepseek-v4-flash"` or `advisor: "@smol:high"`) sets an explicit advisor model pattern with an optional `:level` thinking suffix.
|
||||
- The `task.agentAdvisor` settings record (agent name → `"on"` / `"off"` / model pattern) overrides the frontmatter, and is edited per agent from `/agents` with `A` (an inline editor with model suggestions and a live resolution preview).
|
||||
|
||||
Subagent advisors remain isolated from the subagent's primary tool session in the same way the main advisor is isolated from the main agent.
|
||||
The legacy `advisor.subagents: true` setting migrates to `task.agentAdvisor: { task: "on" }` — the bundled generic `task` agent keeps its advisor, other agents start unadvised.
|
||||
|
||||
An advised subagent session builds its own advisor subsystem with the same settings/model-role resolution (an explicit pattern lands on the spawned session's `modelRoles.advisor`), then reruns both `WATCHDOG.md` and `WATCHDOG.yml` discovery for that subagent session's `cwd` and agent directory. Subagent advisors remain isolated from the subagent's primary tool session in the same way the main advisor is isolated from the main agent.
|
||||
|
||||
## Cost and context behavior
|
||||
|
||||
@@ -319,7 +321,7 @@ The advisor is a passive reviewer with its own model usage, so — like a task s
|
||||
|
||||
- legacy/default advisor: `<session>/__advisor.jsonl`
|
||||
- named advisor: `<session>/__advisor.<slug>.jsonl`
|
||||
- subagent advisor (`advisor.subagents: true`): `<session>/<SubId>/__advisor[.<slug>].jsonl`
|
||||
- subagent advisor (frontmatter `advisor` / `task.agentAdvisor`): `<session>/<SubId>/__advisor[.<slug>].jsonl`
|
||||
|
||||
Paths derive from the owning session file (not the shared artifacts root), so each primary/subagent advisor writes a distinct file. The reserved `__advisor` stem cannot collide with a task subagent id.
|
||||
|
||||
|
||||
+1
-1
@@ -374,7 +374,7 @@ See [Advisor and WATCHDOG.md](./advisor-watchdog.md) for runtime behavior, `WATC
|
||||
| Key | Type | Default | Notes |
|
||||
| --------------------- | ------- | ------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- |
|
||||
| `advisor.enabled` | boolean | `false` | Enable the advisor runtime when `modelRoles.advisor` resolves to an available model. |
|
||||
| `advisor.subagents` | boolean | `false` | Also enable advisor runtimes for spawned task/eval subagents. |
|
||||
| `task.agentAdvisor` | record | `{}` | Per-agent subagent advisor: agent name → `"on"` / `"off"` / advisor model pattern. Overrides agent frontmatter `advisor`; edited from `/agents` with `A`. |
|
||||
| `advisor.syncBacklog` | enum | `off` | Bounded advisor catch-up delay: `off`, `1`, `3`, or `5`. The primary waits up to 30 seconds only while advisor backlog is at or above the threshold. |
|
||||
| `advisor.immuneTurns` | number | `3` | After a `concern`/`blocker` interrupts, route further concerns/blockers as non-interrupting asides for this many completed primary turns. |
|
||||
|
||||
|
||||
@@ -27,7 +27,7 @@ It covers runtime behavior as implemented today, including precedence, invalid-d
|
||||
Task agents normalize into `AgentDefinition` (`src/task/types.ts`):
|
||||
|
||||
- required `name`, `description`, and `systemPrompt`
|
||||
- optional `tools`, `spawns`, prioritized `model` list, `thinkingLevel`, `output`, `blocking`, `autoloadSkills`, `readSummarize`, `prewalk`
|
||||
- optional `tools`, `spawns`, prioritized `model` list, `thinkingLevel`, `output`, `blocking`, `autoloadSkills`, `readSummarize`, `prewalk`, `advisor`
|
||||
- `source`: `"bundled" | "user" | "project"` (extension agents are tagged with their extension root's project/user level)
|
||||
- optional `filePath`
|
||||
|
||||
@@ -43,7 +43,8 @@ Parsing comes from frontmatter via `parseAgentFields()` (`src/discovery/helpers.
|
||||
- `thinking-level` / `thinking` selects the agent's configured effort. When `task.enableEffort` (default `false`) exposes it, a task item's coarse `effort` (`lo`, `med`, `hi`) takes precedence at launch. OMP maps that hint to the selected model's lowest, middle, or highest supported effort, then clamps it to `task.maxEffort` (default `max`). The ceiling is carried across retry-fallback model switches. If the selected model has no supported effort at or below the ceiling, the spawn fails; models without a controllable effort surface instead fall back to their normal selector.
|
||||
- `blocking: true` makes the parent wait for that agent even when async task execution is enabled
|
||||
- `autoloadSkills` names skills from the parent session to inject before the first child prompt; unknown names are ignored
|
||||
- `prewalk: true` starts the subagent on its resolved model and hands off to the default prewalk target (the `smol` role) at its first edit/write, exactly like the session-level `--prewalk`; a string value (e.g. `prewalk: "@smol"` or `prewalk: "openai/gpt-5-mini"`) picks a custom target. The `task.agentPrewalk` settings record (agent name → `"on"` / `"off"` / pattern, toggled per agent from `/agents` with `P`) overrides the frontmatter. Resolution happens in `runSubprocess` (`src/task/executor.ts`). An unavailable target is skipped instead of failing the spawn. A resolved target is skipped only when both its model identity and its effective thinking mode/level match the starting selection after model clamping; a same-model effort downgrade is a real hand-off and still arms and switches at the first edit/write.
|
||||
- `prewalk: true` starts the subagent on its resolved model and hands off to the default prewalk target (the `smol` role) at its first edit/write, exactly like the session-level `--prewalk`; a string value (e.g. `prewalk: "@smol"` or `prewalk: "openai/gpt-5-mini"`) picks a custom target. The `task.agentPrewalk` settings record (agent name → `"on"` / `"off"` / pattern, edited per agent from `/agents` with `P`) overrides the frontmatter. Resolution happens in `runSubprocess` (`src/task/executor.ts`). An unavailable target is skipped instead of failing the spawn. A resolved target is skipped only when both its model identity and its effective thinking mode/level match the starting selection after model clamping; a same-model effort downgrade is a real hand-off and still arms and switches at the first edit/write.
|
||||
- `advisor: true` pairs spawned sessions of the agent with an advisor running the model resolved for the `advisor` role; a string value (e.g. `advisor: "deepseek/deepseek-v4-flash"` or `advisor: "@smol:high"`) sets an explicit advisor model pattern (optional `:level` suffix), applied as the spawned session's `modelRoles.advisor`. The `task.agentAdvisor` settings record (agent name → `"on"` / `"off"` / pattern, edited per agent from `/agents` with `A`) overrides the frontmatter. Resolution happens in `runSubprocess` (`src/task/executor.ts`); subagents default to no advisor, and the effective opt-in is persisted in `session_init` so cold revival restores it.
|
||||
|
||||
## Role-backed custom agents
|
||||
|
||||
|
||||
+2
-1
@@ -94,7 +94,7 @@ Artifacts and side channels:
|
||||
9. If `isolated`, it requires a git repo (`getRepoRoot(...)` / `captureBaseline(...)`), maps `task.isolation.mode` to a backend-kind hint (`parseIsolationMode`), and materializes the workspace via the natives PAL (`ensureIsolation` → `isoResolve`/`isoStart`), walking the candidate list when a backend is unavailable.
|
||||
10. Artifacts dir comes from the parent session file when available, otherwise a temp dir. When the session is executing an approved plan, the plan reference is handed to the subagent.
|
||||
11. Non-isolated spawns call `runSubprocess(...)` directly with parent cwd; isolated spawns run inside the isolation workspace, then commit to a branch (`mergeMode === "branch"`) or capture a patch, and always clean up the workspace.
|
||||
12. `runSubprocess(...)` creates a child agent session with an isolated settings snapshot (parent settings inherited — `async.enabled` and `bash.autoBackground.enabled` are **inherited** from the parent, not force-disabled; `tier.openai`/`tier.anthropic`/`tier.google` are re-resolved through `tier.subagent`; `tools.approvalMode` is forced to `yolo` because headless subagents have no UI to confirm prompts against; per-spawn overrides may disable read summarization and clear extra workspace roots for isolated runs), child `agentId` equal to the allocated id, child internal URL router/`AgentOutputManager`, output schema, the shared `context` (batch calls) in the system prompt's `CONTEXT` section, and the IRC peer roster in the system prompt.
|
||||
12. `runSubprocess(...)` creates a child agent session with an isolated settings snapshot (parent settings inherited — `async.enabled` and `bash.autoBackground.enabled` are **inherited** from the parent, not force-disabled; `tier.openai`/`tier.anthropic`/`tier.google` are re-resolved through `tier.subagent`; `tools.approvalMode` is forced to `yolo` because headless subagents have no UI to confirm prompts against; `advisor.enabled` is forced off unless the spawn opts in per agent; per-spawn overrides may disable read summarization and clear extra workspace roots for isolated runs), child `agentId` equal to the allocated id, child internal URL router/`AgentOutputManager`, output schema, the shared `context` (batch calls) in the system prompt's `CONTEXT` section, and the IRC peer roster in the system prompt.
|
||||
13. Child tool availability: explicit `agent.tools` if provided; auto-add `task` when the agent has `spawns` and depth allows; strip `task` at `task.maxRecursionDepth`; ensure `hub` is present in explicit tool lists; expand `exec` to `eval` + `bash`; strip parent-owned `todo` — unless the spawn is prewalk-armed, whose plan nudge + todo gate need the child to commit its own todo list before the model hand-off.
|
||||
14. The child must finish through the hidden `yield` tool; up to 3 reminder prompts, the last forcing `toolChoice = yield` when supported. `finalizeSubprocessOutput(...)` reconciles raw text, `yield` payloads, structured schemas, and abort states.
|
||||
15. End-of-run lifecycle (keep-alive, in the run finalizer):
|
||||
@@ -115,6 +115,7 @@ Artifacts and side channels:
|
||||
- Isolation merge strategy: patch mode (capture/apply root patches) or branch mode (commit to `omp/task/<id>`, cherry-pick into parent).
|
||||
- Agent source precedence is first-wins by exact name: project `.omp/agents`; user `.omp/agent/agents`; OMP extension-package `agents/` roots in CLI → project settings → user settings → installed npm/link plugin order; Claude marketplace plugin agents (project before user); then bundled (`scout`, `designer`, `reviewer`, `security-reviewer`, `librarian`, `task`, `sonic`).
|
||||
- Prewalk: agent frontmatter `prewalk` or `task.agentPrewalk[agentName]` can start on the normal model and hand off to a cheaper resolved model at the first edit/write. `task.prewalk` (default off) arms this behavior for the bundled generic `task` agent. Missing/unconfigured targets and exact model+effort no-ops skip the handoff rather than failing the spawn.
|
||||
- Advisor: agent frontmatter `advisor` or `task.agentAdvisor[agentName]` (`"on"` / `"off"` / model pattern) pairs the child session with an advisor; an explicit pattern lands on the child's `modelRoles.advisor`. Subagents default to no advisor.
|
||||
|
||||
## Side Effects
|
||||
- Filesystem
|
||||
|
||||
Reference in New Issue
Block a user