@ferris1225/pi-subagents 2.3.1 → 4.0.0

package/README.md

@@ -1,468 +1,475 @@
-# pi-subagents
-
-[![npm version](https://img.shields.io/npm/v/@ferris1225/pi-subagents?color=blue)](https://www.npmjs.com/package/@ferris1225/pi-subagents)
-[![downloads](https://img.shields.io/npm/dm/@ferris1225/pi-subagents)](https://www.npmjs.com/package/@ferris1225/pi-subagents)
-[![license](https://img.shields.io/npm/l/@ferris1225/pi-subagents)](./LICENSE)
-![platform](https://img.shields.io/badge/platform-Windows%20%7C%20macOS%20%7C%20Linux-lightgrey)
-![pi](https://img.shields.io/badge/pi-extension-orange)
-
-Focused background delegation for [pi](https://pi.dev): `explore` / `worker` /
-`cleaner` / `reviewer` agents run in **isolated child processes** and hand their
-results back to the main agent automatically. Install it, and the main model
-starts using it on its own — no prompt engineering, no babysitting.
-
-## 2.0.0 — direct model handoff and capability-aware thinking
-
-Version 2 removes backup pools and global thinking strength. Every agent now has
-one optional selected model; any model/provider failure hands its retained session
-directly to the current main model, while ordinary tool/task failures stay put.
-Thinking defaults to Auto and is clamped through Pi's real model capability map.
-The setup menu and normalized config drop the obsolete options instead of carrying
-compatibility aliases.
-
-This release also adds semantic routing to a dedicated evidence-first `cleaner`
-and makes the active widget show each run's task, effective model/thinking,
-activity, and elapsed time. Every dispatch has a stable run id, so work can be
-steered while it runs, parked without losing context, resumed after settlement,
-retargeted, or forked into another path. Generation ownership keeps startup-race
-retries and stale child processes from corrupting the logical thread without
-retaining a duplicate event history.
-
-The common quality loop now runs end to end without waking the main agent between
-steps:
-
-```text
-reviewer (find issues) → worker (fix every finding) → reviewer (verify) → final PASS/FAIL
-```
-
-Reviews use a single flat findings list — no severity triage. Every reported
-finding is fixed before the change is accepted, and each re-review converges on
-an open-finding set: the worker's explicit rejections are adjudicated once, only
-defects a fix round introduced or exposed are added, and resolved items never
-re-open. `maxFixRounds` stays the hard cap, so a chain always settles and wakes
-the main agent with the full picture.
-
-Cleanup stays a separate lifecycle: explicit cleanup intent can dispatch the
-evidence-first `cleaner`; any edits still go through the independent `reviewer`
-gate.
-
-Each chain is delivered as one concise completion group whose footer totals the
-aggregate token usage and cost of every included run, while full per-run reports
-remain available through `subagent_status`. Its parent stays `running`
-until the whole chain settles; completed internal rounds leave active status
-immediately, so no `done` row keeps accumulating elapsed time. Selected-to-main
-model handoffs keep the same retained context, and isolated parallel workers use
-detached Git worktrees whose changes are applied back without touching the parent
-index.
-
-## Highlights
-
-- **Zero-setup proactive dispatch** — the extension injects a delegation directive
-  into the main system prompt, so the main model sends broad searches to `explore`,
-  self-contained implementations to `worker`, explicit cleanup intent to `cleaner`,
-  and pre-commit reviews to `reviewer`. You just use pi; delegation happens by itself.
-- **Vision-capable image tasks** — flag screenshot/mockup/design work with
-  `vision: true`. The configured vision model hands directly to the current main
-  model on model/provider failure. Setup lists only in-scope, image-capable models
-  from providers with configured authentication, and failures never silently
-  rewrite your configuration.
-- **Results come back on their own** — completions are delivered as messages that
-  wake the main agent automatically, even mid-turn. No polling, no `sleep`, no
-  "go check" step. `subagent_wait` is a **non-blocking** in-turn lookup by default
-  (pass `timeoutMs` to block); `subagent_status` inspects runs; `subagent_stop`
-  cancels one and delivers its partial output.
-- **Active-only live widget, as a tree** — each queued or running sub-agent gets one compact
-  width-aware primary line with task, effective model/thinking, and elapsed time; current
-  activity appears only when present on an indented second line. Auto-fix rounds nest under
-  the triggering reviewer row that owns the chain, so it is always visible who dispatched
-  what; no run ids appear here — the tree and the task label identify each row:
-  ```text
-  ● reviewer · review diff of src/foo.ts · claude-sonnet-4-5/high · 42s
-    ├ ● worker · fix round 1 · src/foo.ts · claude-sonnet-4-5/high · 10s
-    │    grep cacheKey
-    └ ○ reviewer · re-review round 1 · claude-sonnet-4-5/high · 3s
-  ```
-  Long tasks and activity paths truncate first (preserving a useful path tail when
-  possible), groups have no blank rows, and settled/parked runs disappear immediately.
-- **Results are not re-narrated** — a sub-agent's completion is shown to you
-  verbatim, and the main agent is told not to paraphrase it back. It replies with
-  only its own conclusion or next step, so the same findings are never paid for
-  twice in tokens.
-- **Evidence-first cleanup, not deletion by guesswork** — `cleaner` distinguishes
-  read-only audit wording from explicit apply wording, proves consumers and dynamic
-  entrypoints before cutting, and treats finding nothing safe as a valid outcome.
-  It is periodic/intent-driven, never PR-count-driven or an automatic commit gate.
-- **A quality gate that closes the loop** — when a reviewer returns `REVIEW_FAIL`,
-  the extension dispatches a worker briefed with the concrete findings, then a
-  re-review, up to `maxFixRounds` times — and only then wakes the main agent.
-  Every reported finding gets fixed (no severity triage), and re-reviews converge
-  on an open-finding set instead of ping-ponging: worker rejections are adjudicated
-  once, only defects the fix round introduced are added, and resolved items never
-  re-open. Every round stays in the triggering reviewer's cwd, and chains that target the
-  same repository are serialized so shared-checkout edits cannot race.
-- **Direct fallback with real thinking capabilities** — each agent has at most
-  one selected model. An unavailable selection, rate limit, invalid key, quota,
-  missing model, or provider failure hands directly to the current main model.
-  A child-only provider adapter forces inner request retries to zero; transient
-  stream drops still use Pi's outer turn retry, and only a settled model-level
-  failure hands off, without changing user settings. Auto thinking clamps the
-  agent preference to the
-  effective model's real `thinkingLevelMap`; manual setup shows only levels that
-  model supports.
-- **Resumes, retargets, and forks preserve context** — every run is session-backed.
-  `subagent_control` can steer active work, retarget it after a stable abort,
-  park/resume it under the same run id, or fork a parked/settled checkpoint into
-  a new independent run. Concurrent resume calls are serialized.
-- **Concise but honest completions** — group completions end with aggregate token
-  and cost totals across every included run; failed-tool diagnostics stay out of the
-  delivered message and remain one `subagent_status` call away. Actual process,
-  model, and integration failures still surface as failures.
-- **Parallel fan-out with filesystem isolation** — independent tasks run up to a
-  configurable limit (default 4). Parallel workers default to detached Git
-  worktrees; tracked, deleted, untracked, and binary changes are applied back
-  without touching the parent index. Failed integration keeps recovery artifacts.
-- **Recursion is structurally impossible** — children are leaf processes; the
-  `subagent` tool is excluded from their toolset.
-- **Zero runtime dependencies** — agents are plain Markdown files; overriding or
-  adding one is writing a file.
-- **Update announcements** — when a new configurable feature ships, you are told
-  about it once (a persisted marker stops the notice from nagging).
-
-## What this adds beyond generic subagent dispatch
-
-This package combines several concrete runtime behaviors rather than only exposing
-an undifferentiated child-agent launcher:
-
-- language-agnostic semantic role guidance for cleanup intent;
-- a dedicated evidence-first cleaner, with cleanup kept separate from the
-  independent reviewer gate;
-- isolated, retained threads that can be steered, parked, resumed, retargeted, or
-  forked under stable run ids;
-- the reviewer → worker auto-fix → reviewer loop, fixing every finding under a
-  convergence contract with a hard round cap;
-- failed-tool diagnostics available by run id through `subagent_status`;
-- direct selected→main fallback plus capability-aware Auto thinking;
-- detached Git worktree isolation for parallel workers and opt-in write-capable
-  cleaner runs.
-
-## Install
-
-```bash
-pi install npm:@ferris1225/pi-subagents
-```
-
-Requires pi **>= 0.83.0**. After installation, open the setup wizard in an
-interactive TUI session:
-
-```text
-/subagents-setup
-```
-
-Fresh installs enable `explore`, `worker`, `cleaner`, and `reviewer` — you can
-start delegating immediately. Existing explicit `enabledAgents` lists are never
-silently extended; users upgrading with an existing explicit list get a one-time
-notice to opt into `cleaner` with `/subagents-setup`.
-
-## The agents
-
-| Agent | Access | Purpose |
-| --- | --- | --- |
-| `explore` | Read-only | Fast codebase reconnaissance: broad/open-ended search, multi-file lookups, mapping unfamiliar code. Returns compressed, structured findings. |
-| `worker` | Full | Implements, fixes, refactors, and tests a self-contained task end to end, then reports honest verification. |
-| `cleaner` | Full | Evidence-first cleanup. Audits and ranks candidates read-only, or applies the smallest proven cuts when removal is explicit; supports worktree isolation. |
-| `reviewer` | Read-only | Adversarial pre-commit quality gate: independently reviews worker and cleaner edits, plus plans, proposed solutions, codebase health, and PR/issue validation. |
-
-Each agent runs in its own isolated `pi` process with a clean context window; it
-has no memory of your conversation, so briefs must be self-contained (goal, exact
-paths, constraints, expected output).
-
-## Usage
-
-### Single task
-
-```ts
-subagent({ agent: "explore", task: "Map the test setup: which files run what, and how is CI wired? Report exact paths." });
-subagent({ agent: "worker", task: "Implement X in src/foo.ts, add tests, run npm test." });
-subagent({ agent: "cleaner", task: "Audit src/cache for dead code and redundant state; report ranked evidence only." });
-subagent({ agent: "cleaner", task: "Remove the proven dead cache adapter, update its tests/docs, and verify the smallest then broad checks." });
-subagent({ agent: "reviewer", task: "Review the diff of src/index.ts and tests/load.test.ts for correctness and edge cases." });
-```
-
-### Parallel tasks
-
-```ts
-subagent({
-  tasks: [
-    { agent: "explore", task: "Where is the selected-to-main handoff logic?" },
-    { agent: "worker", task: "Add unit tests for models.ts." },
-  ],
-});
-```
-
-### Cleanup routing and lifecycle
-
-The injected guidance routes `cleaner` by explicit semantic intent: **code cleanup**,
-**dead code**, **代码清理**, **精简**, **清理冗余**, **简化**, or **去除过度设计**.
-Requested periodic maintenance passes also qualify; PR counts do not, and cleaner is
-never run automatically as the pre-commit gate.
-
-- **Audit mode:** audit/find/report/review or 审计/查找/检查/报告 wording produces
-  read-only ranked evidence.
-- **Apply mode:** explicit remove/clean/simplify/refactor or 清理/删除/移除/精简/简化/重构
-  wording permits the smallest proven edits plus narrow-then-broad verification.
-
-```text
-explicit cleanup intent → cleaner (audit or apply)
-cleaner apply → reviewer gate → worker auto-fix (on REVIEW_FAIL) → reviewer
-```
-
-`reviewer` remains the independent gate for non-trivial cleaner edits. The auto-fix
-portion runs only when enabled by `maxFixRounds`; cleaner itself is not a pre-commit
-hook or a PR-count scheduler.
-
-### Vision tasks (screenshots / mockups / designs)
-
-When a task may require viewing images — frontend work, UI review, design
-comparisons — set `vision: true` and give the sub-agent the exact image paths:
-
-```ts
-subagent({
-  agent: "reviewer",
-  task: "Compare the UI in screenshots/settings.png against the mockup design.png; list every visual mismatch.",
-  vision: true,
-});
-```
-
-The sub-agent reads images with its `read` tool. Runtime order for a
-vision-flagged run is `visionModel` → current main-window model. A stale or
-currently unavailable vision selection is skipped immediately without rewriting
-it. A vision-flagged auto-fix chain keeps the flag for worker/re-review rounds
-because they may need to inspect the same images.
-
-As a deterministic fallback, a brief that names an image file (an extension such
-as `.png`, `.jpg`, `.webp`) or describes frontend/UI work — Vue/React/Svelte
-pages, mockups, screenshots, page styling — is dispatched as
-vision even without the flag: the model cannot be swapped mid-run, so a worker
-that screenshots and visually verifies what it builds must start on a
-vision-capable model. How do you verify which model ran? The live widget line,
-dispatch result row, and `subagent_status` all show each run's effective model
-id, and a selected→main handoff is labeled with its origin.
-
-### Controlling and stopping
-
-Dispatch confirmations, tool result rows, and completion blocks all show the
-stable `#id`, so a thread remains directly controllable after its live UI is gone
-(the widget itself identifies rows by tree position and task instead of ids).
-
-- `subagent_control` — `steer`, `retarget`, `park`, `resume`, or `fork` a logical
-  thread by stable run id. Resume accepts an optional replacement objective;
-  fork creates a new id and leaves the source unchanged. Park active work before
-  forking it.
-- `subagent_wait` — in-turn result lookup. **Non-blocking by default**: a settled
-  run returns immediately; an active run tells the model to end its turn. Pass
-  `timeoutMs` only when you must stay in the turn.
-- `subagent_status` — active/parked/finished runs and full result by run id.
-- `subagent_stop` — destructive cancellation. It retires that thread's retained
-  session (independent forks survive) and delivers exactly one aborted partial
-  result after the run and any worktree integration have quiesced.
-
-Examples:
-
-```ts
-subagent_control({ action: "steer", id: 7, instruction: "Check the Windows path too." });
-subagent_control({ action: "park", id: 7 });
-subagent_control({ action: "resume", id: 7, objective: "Finish the tests." });
-subagent_control({ action: "fork", id: 7, objective: "Try the smaller alternative." });
-```
-
-### Worktree isolation
-
-Single tasks default to `isolation: "shared"`. Parallel `worker` tasks default
-to `isolation: "worktree"`; opt into shared mode only when a worker must see the
-caller's live uncommitted tree. `cleaner` is also write-capable and supports
-worktree mode when explicitly requested (its default remains shared). Worktree
-mode requires a Git repository with a committed `HEAD` and is rejected for the
-read-only `explore` and `reviewer` agents.
-
-A parked isolated thread keeps its current worktree. Resume it there; fork is
-available after that isolated checkpoint settles and its seed is integrated.
-Resuming or forking a settled isolated thread creates a fresh worktree, merges a
-recorded checkpoint onto the current `HEAD` (including when the seed was already
-committed), and clones the Pi session with the new cwd. Forks then integrate only
-their unique follow-on edits, so a shared seed is applied once. A run remains
-active while final Git integration is in progress and becomes `done` only after
-that boundary finishes.
-
-Every Git operation has a 120-second deadline and process-tree cleanup; captured
-Git output and binary patches are capped at 64 MiB. Setup/bound failures surface
-instead of hanging. Finalization failures retain the patch/worktree when
-available and are recorded in `~/.pi/agent/pi-subagents-recovery.json`; later
-sessions show the recovery paths again until the artifacts are removed.
-
-## Configuration
-
-Stored at `~/.pi/agent/pi-subagents.json` (follows `PI_CODING_AGENT_DIR` when
-set). `/subagents-setup` has five top-level choices: enable agents, configure one
-agent's model/thinking, choose a vision model, runtime settings, or full setup.
-After one agent's model + thinking picks, the wizard returns to the agent picker
-so several agents can be configured in one pass; Esc at any step ends the pass
-and keeps every agent already configured. There is no backup pool or global thinking menu. Model pickers show only in-scope
-models with configured authentication and display their real supported thinking
-levels. Thinking defaults to **Auto**; manual overrides show only levels supported
-by that agent's effective model. `notifyOnReviewPass` and `maxResultLines` remain
-direct-file settings.
-
-```json
-{
-  "enabledAgents": ["explore", "worker", "cleaner", "reviewer"],
-  "agentModels": {
-    "explore": "anthropic/claude-haiku-4-5"
-  },
-  "agentThinkingLevels": {
-    "reviewer": "high"
-  },
-  "visionModel": "anthropic/claude-sonnet-4-5",
-  "notifyOnReviewPass": false,
-  "maxResultLines": 80,
-  "proactiveInjection": true,
-  "agentScope": "user",
-  "maxConcurrency": 4,
-  "maxFixRounds": 2,
-  "idleTimeoutSec": 90
-}
-```
-
-| Field | Description |
-| --- | --- |
-| `enabledAgents` | Agent names exposed to discovery and prompt injection. An empty array disables all agents. |
-| `agentModels` | Optional selected `provider/model-id` per agent. Missing = current main model. Model-level failure hands directly to current main. |
-| `agentThinkingLevels` | Optional manual preference per agent. Missing = Auto (agent frontmatter preference, or `high`, clamped to the effective model's supported levels). |
-| `visionModel` | Optional vision-capable model for `vision: true` tasks (screenshots/mockups/designs). Unset = falls back to the main session's current model. |
-| `notifyOnReviewPass` | When `true`, a passing reviewer result is delivered without waking the main agent (default `false`). |
-| `maxResultLines` | Max lines of a sub-agent result carried in the completion message (default `80`). Longer results are truncated; full text is written to an extension-named temporary `.md`. At session start and on each write, only recognized result files older than 7 days are removed; each canonical project path has its own newest-50 bucket. |
-| `proactiveInjection` | Whether to add the delegation directive to the main system prompt. |
-| `agentScope` | `user`, `project`, or `both`; controls which user/project agent directories are discovered. |
-| `maxConcurrency` | Max sub-agent processes running at once (1–16, default 4), and the max tasks one parallel `subagent` call accepts. Extra work waits in the queue. |
-| `maxFixRounds` | Auto-fix rounds when a reviewer returns `REVIEW_FAIL` (default 2; `0` disables the loop). Hard cap: the chain always settles, delivers its condensed summary, and wakes the main agent. |
-| `idleTimeoutSec` | Idle watchdog: a sub-agent whose stdout goes silent for this long is terminated; a selected model then hands to current main. `0` disables it. Default 90. |
-
-### Model routing and thinking
-
-```text
-selected agent/vision model → current main-window model
-```
-
-Without a selected model, current main runs immediately; agent frontmatter model
-is used only when no main model exists. A selection missing from Pi's live
-available catalog is skipped. Any model-level runtime failure — rate limit,
-quota, invalid key/auth, missing model, provider error, or idle model stream —
-hands directly to current main, including stream errors that retain partial text.
-A child-only Pi extension wraps the selected provider's registered API stream
-with `maxRetries: 0` so a deterministic auth/quota miss fails fast. Transient
-stream drops such as xAI `terminated` still use Pi's outer turn retry — the
-parent does not `abort_retry` them — and only a settled model-level failure
-hands off to current main. This uses supported extension/RPC surfaces in Node
-and standalone/Bun builds, never rewrites global or project settings, and does
-not alter descendant tool environments. Tool/test failures stay on the same
-model because they are task failures, not model availability failures. A child is
-probed with RPC `get_state` before the first prompt so the 30s command ACK clock
-does not include process boot. Only a zero-activity startup miss can retry — a
-silent fast exit, a `get_state` handshake timeout, or an initial prompt ACK
-timeout before any agent/turn/stream/tool activity. Those transport misses are
-not model-level failures and do not hand the task to the main window. An accepted
-prompt or any activity forbids replay.
-
-Auto thinking starts from the Agent's declared preference (`low` for `explore`,
-`high` for the other built-ins) and uses Pi's capability map to clamp it to the
-actual model. Non-reasoning models resolve to `off`; `xhigh`/`max` appear in setup
-only when that model explicitly supports them. A selected→main handoff re-clamps
-thinking for the main model.
-
-### Choosing an explore model
-
-Choose a competent fast code model for `explore`, not automatically the cheapest
-model. Cheap reconnaissance is useful for mechanical symbol/path discovery, but
-a missed dynamic entrypoint or ownership edge can cost more through downstream
-rework. Direct main-model handoff handles provider/runtime failure; it cannot
-detect a plausible but incomplete answer.
-
-`explore` therefore returns an index of exact paths, lines, symbols, and explicit
-uncertainty. The main agent, worker, or cleaner must re-read load-bearing files
-before editing or deciding deletion, security, compatibility, persistence, or
-dynamic reachability. Prefer a stronger model or direct specialist for complex
-dynamic loading, concurrency, migrations, and security-sensitive code.
-
-### Resuming retained context
-
-Every run stores its Pi session in a private temp directory. A selected→main
-handoff resumes that same session, so searches, reads, reasoning, and edits remain
-in context. A parked, completed, or failed thread can later be resumed under its
-stable id:
-
-```ts
-subagent_control({ action: "resume", id: 7 });
-subagent_control({ action: "resume", id: 7, objective: "Continue with the repaired credentials." });
-```
-
-Use `fork` when both paths should remain available. `subagent_stop` is the
-explicit destructive operation that retires a retained session; otherwise
-sessions live until the parent Pi session shuts down.
-
-### Configuration migration
-
-The config file normalizes itself on load — no manual steps after an upgrade.
-Configured non-empty agent names are preserved, while invalid and obsolete keys
-are removed. This release deletes `agentBackupModels` and global `thinkingLevel`
-in addition to older `maxParallelTasks` / `maxSubagentDepth`; per-agent thinking
-preferences remain and are capability-clamped at runtime. A pre-existing explicit
-`enabledAgents` array is still preserved without appending `cleaner`. Existing
-configs without it receive a one-time `/subagents-setup` notice, tracked internally
-in `announcedFeatures`.
-
-## Agent discovery and overrides
-
-- Built-in agents ship with the package; user agents live in `~/.pi/agent/agents/`;
-  project agents in the nearest `.pi/agents/` directory are loaded only when Pi
-  trusts that project.
-- For duplicate names: project overrides user overrides built-in. Keep the
-  matching filename and `name` field to replace a built-in agent.
-- Optional frontmatter: `model` (default model reference), `thinking` (default
-  thinking strength), `tools` (comma-separated tool allow-list; absent = all
-  tools). Config overrides win at spawn.
-
-## How it stays reliable
-
-- **Direct model recovery** — unavailable selections skip immediately; any
-  selected-model provider/auth/quota/rate-limit failure hands directly to current
-  main with thinking re-clamped to the main model.
-- **Startup-race retries** — a silent zero-activity child exit (concurrent pi
-  startup lock contention) is relaunched with backoff; only clean silent exits
-  qualify, so real work is never duplicated.
-- **Idle watchdog** — a stalled selected-model stream (no output for
-  `idleTimeoutSec`) terminates the child and hands the retained session to current
-  main.
-- **Dispatch failures surface** — partial parallel startup reports every failed
-  item and reason; if none start, the tool throws so Pi records a real tool error.
-  Dispatch crashes likewise produce a failed result instead of a silent hang.
-- **Safe live status** — tool activity is credential-redacted and stripped of terminal control sequences before `subagent_status` can return it.
-- **Leaf children** — no nested delegation, no runaway trees.
-
-## Development
-
-```bash
-npm install
-npm run check
-npm test
-```
-
-The source is modular: `dispatch.ts` (public dispatch contract + auto-fix),
-`thread-lifecycle.ts` (queued generations, resume/fork, and isolation settlement),
-`rpc-run.ts` / `spawn.ts` (persistent child transport + selected→main handoff),
-`worktree.ts` / `session-fork.ts` (filesystem/session branching), `tools.ts`
-(wait/status/control/stop), `widget.ts` (active-only TUI status), `announcements.ts`
-(recovery and feature notices), and `runtime.ts` (session-scoped ownership). No runtime
-dependencies beyond pi peer dependencies.
-
-## License
-
-MIT
+# pi-subagents
+
+[![npm version](https://img.shields.io/npm/v/@ferris1225/pi-subagents?color=blue)](https://www.npmjs.com/package/@ferris1225/pi-subagents)
+[![downloads](https://img.shields.io/npm/dm/@ferris1225/pi-subagents)](https://www.npmjs.com/package/@ferris1225/pi-subagents)
+[![license](https://img.shields.io/npm/l/@ferris1225/pi-subagents)](./LICENSE)
+![platform](https://img.shields.io/badge/platform-Windows%20%7C%20macOS%20%7C%20Linux-lightgrey)
+![pi](https://img.shields.io/badge/pi-extension-orange)
+
+Focused background delegation for [pi](https://pi.dev): `explorer` / `worker` /
+`cleaner` / `reviewer` agents run in **isolated child processes** and hand their
+results back to the main agent automatically. Install it, and the main model
+starts using it on its own — no prompt engineering, no babysitting.
+
+## 4.0.0 — consistent agent names and direct cleanup
+
+Version 4 renames the built-in reconnaissance role from `explore` to `explorer`
+and deliberately removes the old alias. Existing explicit configuration must use
+the new key. `cleaner` is now apply-only: explicit cleanup intent authorizes it to
+prove and perform every safe in-scope cut, while generic or read-only assessments
+go to `reviewer` without triggering auto-fix.
+
+The main delegation directive is now the single authoritative routing policy;
+duplicated tool guidelines were removed to cut the default parent injection by
+more than half without changing process isolation, worktree rules, retained-session
+fallback, or result handoff. Agent frontmatter model defaults remain because they
+still select a model when no current main model exists; frontmatter comments do not
+enter model context.
+
+Every dispatch retains its stable run id and can be steered, parked, resumed,
+retargeted, or forked. The active widget continues to show task, effective model
+and thinking level, activity, and elapsed time.
+
+The common quality loop now runs end to end without waking the main agent between
+steps:
+
+```text
+reviewer (find issues) → worker (fix every finding) → reviewer (verify) → final PASS/FAIL
+```
+
+Gate reviews use a single flat findings list — no severity triage. Every reported
+finding is fixed before the change is accepted, and each re-review converges on
+an open-finding set: the worker's explicit rejections are adjudicated once, only
+defects a fix round introduced or exposed are added, and resolved items never
+re-open. `maxFixRounds` stays the hard cap, so a chain always settles and wakes
+the main agent with the full picture. Advisory reviewer requests (generic audits,
+code health, plans, and proposed solutions) return evidence without a machine
+verdict, so they never start auto-fix.
+
+Cleanup stays a separate lifecycle: only an explicit request authorizing cleanup,
+removal, or simplification edits dispatches the evidence-first `cleaner`. It proves
+candidates, applies every safe in-scope cut end to end, and may validly make zero
+edits; non-trivial changes still go through the independent `reviewer` gate.
+
+Each chain is delivered as one concise completion group whose footer totals the
+aggregate token usage and cost of every included run, while full per-run reports
+remain available through `subagent_status`. Its parent stays `running`
+until the whole chain settles; completed internal rounds leave active status
+immediately, so no `done` row keeps accumulating elapsed time. Selected-to-main
+model handoffs keep the same retained context, and isolated parallel workers use
+detached Git worktrees whose changes are applied back without touching the parent
+index.
+
+## Highlights
+
+- **Zero-setup proactive dispatch** — the extension injects a delegation directive
+  into the main system prompt, so the main model sends broad searches to `explorer`,
+  self-contained implementations to `worker`, edit-authorizing cleanup to `cleaner`,
+  and generic assessments or pre-commit gates to `reviewer`. You just use pi;
+  delegation happens by itself.
+- **Multimodal work is a model choice, not a mode** — an agent that should see
+  screenshots, mockups, or its own rendered pages simply gets a multimodal model
+  through `/subagents-setup` (the picker labels each model `vision` or
+  `text-only`). The agent reads images with its `read` tool on whatever model it
+  runs; no per-task flag, no separate vision override.
+- **Results come back on their own** — completions are delivered as messages that
+  wake the main agent automatically, even mid-turn. No polling, no `sleep`, no
+  "go check" step. `subagent_wait` is a **non-blocking** in-turn lookup by default
+  (pass `timeoutMs` to block); `subagent_status` inspects runs; `subagent_stop`
+  cancels one and delivers its partial output.
+- **Active-only live widget, as a tree** — each queued or running sub-agent gets one compact
+  width-aware primary line with task, effective model/thinking, and elapsed time; current
+  activity appears only when present on an indented second line. Auto-fix rounds nest under
+  the triggering reviewer row that owns the chain, so it is always visible who dispatched
+  what; no run ids appear here — the tree and the task label identify each row:
+  ```text
+  ● reviewer · review diff of src/foo.ts · claude-sonnet-4-5/high · 42s
+    ├ ● worker · fix round 1 · src/foo.ts · claude-sonnet-4-5/high · 10s
+    │    grep cacheKey
+    └ ○ reviewer · re-review round 1 · claude-sonnet-4-5/high · 3s
+  ```
+  Long tasks and activity paths truncate first (preserving a useful path tail when
+  possible), groups have no blank rows, and settled/parked runs disappear immediately.
+- **Results are not re-narrated** — a sub-agent's completion is shown to you
+  verbatim, and the main agent is told not to paraphrase it back. It replies with
+  only its own conclusion or next step, so the same findings are never paid for
+  twice in tokens.
+- **Evidence-first cleanup, not deletion by guesswork** — `cleaner` is apply-only:
+  an explicit cleanup request authorizes edits, but each candidate must be proved
+  before every safe in-scope cut is applied and verified. Finding nothing safe and
+  making zero edits remains valid. Generic/read-only assessments go to `reviewer`;
+  cleaner is periodic/intent-driven, never PR-count-driven or an automatic gate.
+- **A quality gate that closes the loop** — when a gate reviewer returns
+  `REVIEW_FAIL`, the extension dispatches a worker briefed with the concrete
+  findings, then a re-review, up to `maxFixRounds` times — and only then wakes the
+  main agent. Advisory reviewer reports omit that verdict and never trigger edits.
+  Every reported finding gets fixed (no severity triage), and re-reviews converge
+  on an open-finding set instead of ping-ponging: worker rejections are adjudicated
+  once, only defects the fix round introduced are added, and resolved items never
+  re-open. Every round stays in the triggering reviewer's cwd, and chains that target the
+  same repository are serialized so shared-checkout edits cannot race.
+- **Direct fallback with real thinking capabilities** — each agent has at most
+  one selected model. An unavailable selection, rate limit, invalid key, quota,
+  missing model, or provider failure hands directly to the current main model.
+  A child-only provider adapter forces inner request retries to zero; transient
+  stream drops still use Pi's outer turn retry, and only a settled model-level
+  failure hands off, without changing user settings. Auto thinking clamps the
+  agent preference to the
+  effective model's real `thinkingLevelMap`; manual setup shows only levels that
+  model supports.
+- **Resumes, retargets, and forks preserve context** — every run is session-backed.
+  `subagent_control` can steer active work, retarget it after a stable abort,
+  park/resume it under the same run id, or fork a parked/settled checkpoint into
+  a new independent run. Concurrent resume calls are serialized.
+- **Concise but honest completions** — group completions end with aggregate token
+  and cost totals across every included run; failed-tool diagnostics stay out of the
+  delivered message and remain one `subagent_status` call away. Actual process,
+  model, and integration failures still surface as failures.
+- **Parallel fan-out with filesystem isolation** — independent tasks run up to a
+  configurable limit (default 4). Parallel workers default to detached Git
+  worktrees; tracked, deleted, untracked, and binary changes are applied back
+  without touching the parent index. Failed integration keeps recovery artifacts.
+- **Recursion is structurally impossible** — children are leaf processes; the
+  `subagent` tool is excluded from their toolset.
+- **Zero runtime dependencies** — agents are plain Markdown files; overriding or
+  adding one is writing a file.
+- **Update announcements** — when a new configurable feature ships, you are told
+  about it once (a persisted marker stops the notice from nagging).
+
+## What this adds beyond generic subagent dispatch
+
+This package combines several concrete runtime behaviors rather than only exposing
+an undifferentiated child-agent launcher:
+
+- language-agnostic semantic role guidance for cleanup intent;
+- a dedicated evidence-first cleaner, with cleanup kept separate from the
+  independent reviewer gate;
+- isolated, retained threads that can be steered, parked, resumed, retargeted, or
+  forked under stable run ids;
+- the reviewer → worker auto-fix → reviewer loop, fixing every finding under a
+  convergence contract with a hard round cap;
+- failed-tool diagnostics available by run id through `subagent_status`;
+- direct selected→main fallback plus capability-aware Auto thinking;
+- detached Git worktree isolation for parallel workers and opt-in write-capable
+  cleaner runs.
+
+## Install
+
+```bash
+pi install npm:@ferris1225/pi-subagents
+```
+
+Requires pi **>= 0.83.0**. After installation, open the setup wizard in an
+interactive TUI session:
+
+```text
+/subagents-setup
+```
+
+Fresh installs enable `explorer`, `worker`, `cleaner`, and `reviewer` — you can
+start delegating immediately. Existing explicit `enabledAgents` lists are never
+silently extended; users upgrading with an existing explicit list get a one-time
+notice to opt into `cleaner` with `/subagents-setup`.
+
+## The agents
+
+| Agent | Access | Purpose |
+| --- | --- | --- |
+| `explorer` | Read-only | Fast codebase reconnaissance: broad/open-ended search, multi-file lookups, mapping unfamiliar code. Returns compressed, structured retrieval leads. |
+| `worker` | Full | Implements, fixes, refactors, and tests a self-contained task end to end, then reports honest verification. |
+| `cleaner` | Full | Proves and applies every safe in-scope cleanup authorized by an explicit cleanup/removal/simplification request; zero edits is valid. Supports worktree isolation. |
+| `reviewer` | Read-only | Handles generic audits, code health, plans, proposed solutions, PR/issue validation, and independent pre-commit gates. Advisory reports do not trigger auto-fix. |
+
+Each agent runs in its own isolated `pi` process with a clean context window; it
+has no memory of your conversation, so briefs must be self-contained (goal, exact
+paths, constraints, expected output).
+
+## Usage
+
+### Single task
+
+```ts
+subagent({ agent: "explorer", task: "Map the test setup: which files run what, and how is CI wired? Report exact paths." });
+subagent({ agent: "worker", task: "Implement X in src/foo.ts, add tests, run npm test." });
+subagent({ agent: "reviewer", task: "Audit src/cache for dead-code candidates and redundant state; report evidence only." });
+subagent({ agent: "cleaner", task: "Clean up src/cache: prove and apply every safe dead-code or redundancy cut, update tests/docs, and verify." });
+subagent({ agent: "reviewer", task: "Gate the diff of src/index.ts and tests/load.test.ts for correctness and edge cases." });
+```
+
+### Parallel tasks
+
+```ts
+subagent({
+  tasks: [
+    { agent: "explorer", task: "Where is the selected-to-main handoff logic?" },
+    { agent: "worker", task: "Add unit tests for models.ts." },
+  ],
+});
+```
+
+### Cleanup routing and lifecycle
+
+The injected guidance sends only explicit, edit-authorizing cleanup intent to
+`cleaner` in whatever language the conversation uses: clean up/remove dead code,
+reduce redundancy, simplify, remove over-engineering, or run a maintenance cleanup
+pass. Cleaner first proves reachability, ownership, history, and boundaries, then
+applies every safe in-scope cut end to end and verifies it. No proven safe cut means
+zero edits, not a forced deletion.
+
+Generic or explicitly read-only **audit**, **inspect**, **report**, **review**,
+**code-health**, **plan**, **proposed-solution**, or cleanup-candidate assessment
+requests go to `reviewer`. Those are advisory reviews: they omit the machine
+`REVIEW_PASS` / `REVIEW_FAIL` marker, cannot start auto-fix, and do not authorize
+the main agent to edit. A follow-up change needs an explicit user request. A
+reviewer emits the marker only for an explicit diff/pre-commit acceptance gate.
+
+```text
+explicit edit-authorizing cleanup → cleaner → reviewer gate
+read-only/generic assessment → reviewer advisory report (no auto-fix)
+reviewer gate REVIEW_FAIL → worker auto-fix → reviewer gate
+```
+
+Cleaner is never dispatched by PR count and never acts as the commit gate. The
+auto-fix portion runs only for gate verdicts and only when enabled by
+`maxFixRounds`.
+
+### Image work (screenshots / mockups / designs)
+
+There is no vision flag or separate vision model. Give the agent a multimodal
+model in `/subagents-setup` and name the exact image paths in the task:
+
+```ts
+subagent({
+  agent: "reviewer",
+  task: "Compare the UI in screenshots/settings.png against the mockup design.png; list every visual mismatch.",
+});
+```
+
+The sub-agent reads images with its `read` tool on its configured model; the
+setup picker labels each model `vision` or `text-only` so the choice is visible.
+The live widget line, dispatch result row, and `subagent_status` all show each
+run's effective model id, and a selected→main handoff is labeled with its
+origin.
+
+### Controlling and stopping
+
+Dispatch confirmations, tool result rows, and completion blocks all show the
+stable `#id`, so a thread remains directly controllable after its live UI is gone
+(the widget itself identifies rows by tree position and task instead of ids).
+
+- `subagent_control` — `steer`, `retarget`, `park`, `resume`, or `fork` a logical
+  thread by stable run id. Resume accepts an optional replacement objective;
+  fork creates a new id and leaves the source unchanged. Park active work before
+  forking it.
+- `subagent_wait` — in-turn result lookup. **Non-blocking by default**: a settled
+  run returns immediately; an active run tells the model to end its turn. Pass
+  `timeoutMs` only when you must stay in the turn.
+- `subagent_status` — active/parked/finished runs and full result by run id.
+- `subagent_stop` — destructive cancellation. It retires that thread's retained
+  session (independent forks survive) and delivers exactly one aborted partial
+  result after the run and any worktree integration have quiesced.
+
+Examples:
+
+```ts
+subagent_control({ action: "steer", id: 7, instruction: "Check the Windows path too." });
+subagent_control({ action: "park", id: 7 });
+subagent_control({ action: "resume", id: 7, objective: "Finish the tests." });
+subagent_control({ action: "fork", id: 7, objective: "Try the smaller alternative." });
+```
+
+### Worktree isolation
+
+Single tasks default to `isolation: "shared"`. Parallel `worker` tasks default
+to `isolation: "worktree"`; opt into shared mode only when a worker must see the
+caller's live uncommitted tree. `cleaner` is also write-capable and supports
+worktree mode when explicitly requested (its default remains shared). Worktree
+mode requires a Git repository with a committed `HEAD` and is rejected for the
+read-only `explorer` and `reviewer` agents.
+
+A parked isolated thread keeps its current worktree. Resume it there; fork is
+available after that isolated checkpoint settles and its seed is integrated.
+Resuming or forking a settled isolated thread creates a fresh worktree, merges a
+recorded checkpoint onto the current `HEAD` (including when the seed was already
+committed), and clones the Pi session with the new cwd. Forks then integrate only
+their unique follow-on edits, so a shared seed is applied once. A run remains
+active while final Git integration is in progress and becomes `done` only after
+that boundary finishes.
+
+Every Git operation has a 120-second deadline and process-tree cleanup; captured
+Git output and binary patches are capped at 64 MiB. Setup/bound failures surface
+instead of hanging. Finalization failures retain the patch/worktree when
+available and are recorded in `~/.pi/agent/pi-subagents-recovery.json`; later
+sessions show the recovery paths again until the artifacts are removed.
+
+## Configuration
+
+Stored at `~/.pi/agent/pi-subagents.json` (follows `PI_CODING_AGENT_DIR` when
+set). `/subagents-setup` has four top-level choices: enable agents, configure one
+agent's model/thinking, runtime settings, or full setup.
+After one agent's model + thinking picks, the wizard returns to the agent picker
+so several agents can be configured in one pass; Esc at any step ends the pass
+and keeps every agent already configured. There is no backup pool or global thinking menu. Model pickers show only in-scope
+models with configured authentication and display their real supported thinking
+levels. Thinking defaults to **Auto**; manual overrides show only levels supported
+by that agent's effective model. `notifyOnReviewPass` and `maxResultLines` remain
+direct-file settings.
+
+```json
+{
+  "enabledAgents": ["explorer", "worker", "cleaner", "reviewer"],
+  "agentModels": {
+    "explorer": "anthropic/claude-haiku-4-5"
+  },
+  "agentThinkingLevels": {
+    "reviewer": "high"
+  },
+  "notifyOnReviewPass": false,
+  "maxResultLines": 80,
+  "proactiveInjection": true,
+  "agentScope": "user",
+  "maxConcurrency": 4,
+  "maxFixRounds": 2,
+  "idleTimeoutSec": 90
+}
+```
+
+| Field | Description |
+| --- | --- |
+| `enabledAgents` | Agent names exposed to discovery and prompt injection. An empty array disables all agents. |
+| `agentModels` | Optional selected `provider/model-id` per agent. Missing = current main model. Model-level failure hands directly to current main. |
+| `agentThinkingLevels` | Optional manual preference per agent. Missing = Auto (agent frontmatter preference, or `high`, clamped to the effective model's supported levels). |
+| `notifyOnReviewPass` | When `true`, a passing reviewer result is delivered without waking the main agent (default `false`). |
+| `maxResultLines` | Max lines of a sub-agent result carried in the completion message (default `80`). Longer results are truncated; full text is written to an extension-named temporary `.md`. At session start and on each write, only recognized result files older than 7 days are removed; each canonical project path has its own newest-50 bucket. |
+| `proactiveInjection` | Whether to add the delegation directive to the main system prompt. |
+| `agentScope` | `user`, `project`, or `both`; controls which user/project agent directories are discovered. |
+| `maxConcurrency` | Max sub-agent processes running at once (1–16, default 4), and the max tasks one parallel `subagent` call accepts. Extra work waits in the queue. |
+| `maxFixRounds` | Auto-fix rounds when a reviewer returns `REVIEW_FAIL` (default 2; `0` disables the loop). Hard cap: the chain always settles, delivers its condensed summary, and wakes the main agent. |
+| `idleTimeoutSec` | Idle watchdog: a sub-agent whose stdout goes silent for this long is terminated; a selected model then hands to current main. `0` disables it. Default 90. |
+
+### Model routing and thinking
+
+```text
+selected agent model → current main-window model
+```
+
+Without a selected model, current main runs immediately; agent frontmatter `model`
+is used only when no main model exists, so the shipped defaults remain behaviorally
+load-bearing. From an agent Markdown file, only the body after frontmatter becomes
+the child's appended system prompt; model-selection comments inside YAML
+frontmatter are parser comments, not model prompt tokens. A selection missing from Pi's live
+available catalog is skipped. Any model-level runtime failure — rate limit,
+quota, invalid key/auth, missing model, provider error, or idle model stream —
+hands directly to current main, including stream errors that retain partial text.
+A child-only Pi extension wraps the selected provider's registered API stream
+with `maxRetries: 0` so a deterministic auth/quota miss fails fast. Transient
+stream drops such as xAI `terminated` still use Pi's outer turn retry — the
+parent does not `abort_retry` them — and only a settled model-level failure
+hands off to current main. This uses supported extension/RPC surfaces in Node
+and standalone/Bun builds, never rewrites global or project settings, and does
+not alter descendant tool environments. Tool/test failures stay on the same
+model because they are task failures, not model availability failures. A child is
+probed with RPC `get_state` before the first prompt so the 30s command ACK clock
+does not include process boot. Only a zero-activity startup miss can retry — a
+silent fast exit, a `get_state` handshake timeout, or an initial prompt ACK
+timeout before any agent/turn/stream/tool activity. Those transport misses are
+not model-level failures and do not hand the task to the main window. An accepted
+prompt or any activity forbids replay.
+
+Auto thinking starts from the Agent's declared preference (`low` for `explorer`,
+`high` for the other built-ins) and uses Pi's capability map to clamp it to the
+actual model. Non-reasoning models resolve to `off`; `xhigh`/`max` appear in setup
+only when that model explicitly supports them. A selected→main handoff re-clamps
+thinking for the main model.
+
+### Choosing an explorer model
+
+Choose a competent fast code model for `explorer`, not automatically the cheapest
+model. Cheap reconnaissance is useful for mechanical symbol/path discovery, but
+a missed dynamic entrypoint or ownership edge can cost more through downstream
+rework. Direct main-model handoff handles provider/runtime failure; it cannot
+detect a plausible but incomplete answer.
+
+`explorer` therefore returns an index of exact paths, lines, symbols, and explicit
+uncertainty. The main agent, worker, or cleaner must re-read load-bearing files
+before editing or deciding deletion, security, compatibility, persistence, or
+dynamic reachability. Prefer a stronger model or direct specialist for complex
+dynamic loading, concurrency, migrations, and security-sensitive code.
+
+### Resuming retained context
+
+Every run stores its Pi session in a private temp directory. A selected→main
+handoff resumes that same session, so searches, reads, reasoning, and edits remain
+in context. A parked, completed, or failed thread can later be resumed under its
+stable id:
+
+```ts
+subagent_control({ action: "resume", id: 7 });
+subagent_control({ action: "resume", id: 7, objective: "Continue with the repaired credentials." });
+```
+
+Use `fork` when both paths should remain available. `subagent_stop` is the
+explicit destructive operation that retires a retained session; otherwise
+sessions live until the parent Pi session shuts down.
+
+### Configuration migration
+
+Config loading normalizes schema fields and removes invalid or obsolete keys,
+including `agentBackupModels`, global `thinkingLevel`, `maxParallelTasks`, and
+`maxSubagentDepth`. Per-agent thinking preferences remain capability-clamped.
+
+The built-in reconnaissance role is now `explorer`. There is deliberately no
+`explore` alias and no automatic key migration. Existing configurations must change
+that role name in `enabledAgents`, `agentModels`, and `agentThinkingLevels` (or run
+`/subagents-setup`). Configured non-empty names are otherwise preserved, so a stale
+name is not silently rewritten. A pre-existing explicit `enabledAgents` list is
+also still preserved without appending `cleaner`; configs without cleaner receive
+the existing one-time setup notice.
+
+## Agent discovery and overrides
+
+- Built-in agents ship with the package; user agents live in `~/.pi/agent/agents/`;
+  project agents in the nearest `.pi/agents/` directory are loaded only when Pi
+  trusts that project.
+- For duplicate names: project overrides user overrides built-in. Keep the
+  matching filename and `name` field to replace a built-in agent.
+- Optional frontmatter: `model` (default model reference), `thinking` (default
+  thinking strength), `tools` (comma-separated tool allow-list; absent = all
+  tools). Config overrides win at spawn.
+
+## How it stays reliable
+
+- **Direct model recovery** — unavailable selections skip immediately; any
+  selected-model provider/auth/quota/rate-limit failure hands directly to current
+  main with thinking re-clamped to the main model.
+- **Startup-race retries** — a silent zero-activity child exit (concurrent pi
+  startup lock contention) is relaunched with backoff; only clean silent exits
+  qualify, so real work is never duplicated.
+- **Idle watchdog** — a stalled selected-model stream (no output for
+  `idleTimeoutSec`) terminates the child and hands the retained session to current
+  main.
+- **Dispatch failures surface** — partial parallel startup reports every failed
+  item and reason; if none start, the tool throws so Pi records a real tool error.
+  Dispatch crashes likewise produce a failed result instead of a silent hang.
+- **Safe live status** — tool activity is credential-redacted and stripped of terminal control sequences before `subagent_status` can return it.
+- **Leaf children** — no nested delegation, no runaway trees.
+
+## Development
+
+```bash
+npm install
+npm run check
+npm test
+```
+
+The source is modular: `dispatch.ts` (public dispatch contract + auto-fix),
+`thread-lifecycle.ts` (queued generations, resume/fork, and isolation settlement),
+`rpc-run.ts` / `spawn.ts` (persistent child transport + selected→main handoff),
+`worktree.ts` / `session-fork.ts` (filesystem/session branching), `tools.ts`
+(wait/status/control/stop), `widget.ts` (active-only TUI status), `announcements.ts`
+(recovery and feature notices), and `runtime.ts` (session-scoped ownership). No runtime
+dependencies beyond pi peer dependencies.
+
+## License
+
+MIT