pi-crew 0.9.68 → 0.10.2

package/docs/README.md

@@ -30,6 +30,8 @@ stay at the root of this directory; historical/retired docs live in
 | [`COVERAGE-ASSESSMENT-2026-08-01.md`](./COVERAGE-ASSESSMENT-2026-08-01.md) | v0.9.56 coverage assessment + source-reorg history |
 | [`AUDIT-2026-07-30.md`](./AUDIT-2026-07-30.md) | v0.9.56 comprehensive audit (6 parallel streams) |
 | [`REMEDIATION-PLAN-2026-07-30.md`](./REMEDIATION-PLAN-2026-07-30.md) | v0.9.56 remediation plan |
+| [`refactor-plan.md`](./refactor-plan.md) | Maintainability refactor plan (phased, v0.9.68) |
+| [`refactor-plan.review.md`](./refactor-plan.review.md) | Verification sweep 3 (vòng 4) audit trail — 3 large-file baselines CONFIRMED, 6 corrections, Phase 4 reframe |
 
 ## Subdirectories
 

package/docs/actions-reference.md

@@ -2,6 +2,23 @@
 
 The `team` tool is the primary tool that pi-crew registers with Pi. All operations go through `action`.
 
+### `plans` — Inspect and decide a run's plan record (T2/R4)
+
+Reads the versioned `PlanRecord` list at `<stateRoot>/plans/plans.json` (ADR-4). Pre-v2 runs without a record degrade gracefully with the manifest-gate info.
+
+```json
+{ "action": "plans", "runId": "team_..." }
+{ "action": "plans", "runId": "team_...", "config": { "subAction": "get", "rev": 2 } }
+{ "action": "plans", "runId": "team_...", "config": { "subAction": "list" } }
+{ "action": "plans", "runId": "team_...", "config": { "subAction": "diff", "a": 1, "b": 2 } }
+{ "action": "plans", "runId": "team_...", "config": { "subAction": "approve" } }
+```
+
+- `get` (default) renders the current (or pinned `rev`) revision with derived per-item progress (done/failed/running from linked `taskIds`).
+- `list` renders the revision history.
+- `diff` reports item-level changes between two revisions (added / removed / dropped / status changes / retitles).
+- `approve` / `reject` delegate to the api `approve-plan` / `cancel-plan` ops (dual-write: manifest keeps its vocabulary — `cancelled` on deny — the record side uses `rejected`).
+
 ## Quick Reference
 
 | Action | Purpose | When to use |
@@ -10,6 +27,7 @@ The `team` tool is the primary tool that pi-crew registers with Pi. All operatio
 | `run` | Create a run and execute a workflow | Main operation |
 | `parallel` | Fan out independent tasks as concurrent runs | When you need N truly concurrent agents |
 | `plan` | Preview a workflow without running tasks | Dry-run planning |
+| `plans` | Inspect/decide the versioned plan record of a run | Plan revisions, diff, approve/reject |
 | `orchestrate` | Execute from a plan document | Automate a plan |
 | `schedule` | Schedule recurring runs | Periodic automation |
 | `scheduled` | List scheduled jobs | View schedules |
@@ -800,6 +818,19 @@ Returns `status: "error"` when the run finished as `failed`.
 > ⚠️ **Internal / experimental.** Queues a steering message that is delivered to
 > a task's worker session on its next turn. The message is appended to
 > `pendingSteers` and also written to a live steering file for immediate delivery.
+>
+> **Updated (T1/WP-1, 2026-08-17, v0.10.1):** the agent-level tools
+> `steer_subagent` / `crew_agent_steer` are now **implemented** — they resolve the
+> subagent record via the unified ownership map (`src/state/stores/ownership-map.ts`,
+> task ⇄ subagent ⇄ pid ⇄ artifacts) to the run's steering file and append a
+> `{type:"steer", message, ts}` line that the worker polls every 500ms and picks
+> up at its next turn boundary. Terminal-task refusal + size-cap parity with
+> `team action=steer`; the steering file is truncated at each worker spawn so
+> lines are scoped to one worker incarnation. The record's `taskId` is linked at
+> dispatch time (onRunStarted), so mid-run steering of a live one-shot agent
+> works. See ADR `docs/decisions/2026-08-14-agent-steer-tools-planned-not-implemented.md`
+> (status: CLOSED). `team action=steer` (runId + taskId + message) remains the
+> run-level path.
 
 ```json
 {

package/docs/commands-reference.md

@@ -8,21 +8,22 @@ Slash commands are manual actions triggered from the Pi chat. Autonomous tool us
 |---------|-------|
 | `/teams` | List teams, agents, workflows, and recent runs |
 | `/team-run [options] <goal>` | Run a team workflow |
-| `/team-orchestrate <planPath>` | Execute from a plan document |
-| `/team-schedule [options]` | Schedule a recurring run |
-| `/team-scheduled` | List scheduled jobs |
 | `/team-cancel <runId>` | Cancel a run |
 | `/team-status <runId>` | View status |
 | `/team-summary <runId>` | View or write a summary |
 | `/team-resume <runId>` | Resume a stopped run |
-| `/team-search <query>` | BM25-ranked discovery |
-| `/team-graph <runId>` | Load, save, or list run graphs |
+| `/team-retry <runId>` | Retry a failed run |
+| `/team-result <runId>` | Show a task result artifact |
+| `/team-respond <taskId> <message>` | Respond to a waiting worker task |
+| `/team-follow-up [options]` | Manage follow-up notifications |
 | `/team-events <runId>` | View the event log |
 | `/team-artifacts <runId>` | View artifacts |
+| `/team-transcript <runId>` | Open the transcript viewer |
 | `/team-worktrees <runId>` | View worktree metadata |
-| `/team-cleanup <runId>` | Remove worktrees |
+| `/team-cleanup <runId>` | Remove worktrees (interactive menu: `/team-cleanup-menu`) |
 | `/team-forget <runId>` | Fully delete a run |
 | `/team-prune` | Delete multiple old runs |
+| `/team-invalidate [scope]` | Invalidate caches |
 | `/team-export <runId>` | Export a run bundle |
 | `/team-import <path>` | Import a run bundle |
 | `/team-imports` | List imported bundles |
@@ -37,9 +38,16 @@ Slash commands are manual actions triggered from the Pi chat. Autonomous tool us
 | `/team-validate` | Validate resources |
 | `/team-help` | Help text |
 | `/team-doctor` | Diagnose the environment |
+| `/team-mascot` | Toggle the mascot overlay |
 | **`/team-goal`** | **v0.9.0** Start autonomous goal loop (sub-actions: `start/status/pause/resume/stop/step/clear`) |
 | **`/workflows`** | **v0.9.0** List static + dynamic workflows (`.dwf.ts`) |
 
+**Removed (v0.10.1 docs hygiene):** `/team-orchestrate`, `/team-schedule`,
+`/team-scheduled`, `/team-search`, `/team-graph` — phantom entries; none of
+these slash commands is registered. Their functionality is reachable via the
+`team` tool actions (`action: "orchestrate" | "schedule" | "scheduled" |
+"search" | "graph"`) instead.
+
 ---
 
 ## `/team-run` — Details
@@ -296,6 +304,9 @@ Creates directories:
 | `u` | View summary |
 | `i` | API read-manifest |
 | `q` or `Esc` | Close |
+| `1`–`7` | Pane: agents / progress / mailbox / output / health / metrics / **plan** (v0.10.1, `PI_CREW_PLAN_UI=1`) |
+| `A` / `n` | Plan approval approve/deny (progress + plan panes, while pending) |
+| `X` | Plan revision diff (plan pane) |
 
 ---
 

package/docs/resource-formats.md

@@ -107,6 +107,8 @@ Format:
 ---
 name: default
 description: Explore, plan, execute, verify
+maxConcurrency: 3
+specStrict: false
 ---
 
 ## explore
@@ -122,6 +124,15 @@ output: plan.md
 Create a plan for: {goal}
 ```
 
+Frontmatter fields:
+
+| Field | Meaning |
+| --- | --- |
+| `maxConcurrency` | Max parallel tasks for the run |
+| `topology` | Explicit topology classification |
+| `coalesceMicroTasks` | Coalesce micro-tasks sharing role+cwd into one worker prompt |
+| `specStrict` | T4/R6 (ADR-6): strict spec mode — coverage AND machine-check; requires a `verifier` role step (reject-start) |
+
 Step fields:
 
 | Field | Meaning |
@@ -136,5 +147,7 @@ Step fields:
 | `progress` | `true`/`false` |
 | `worktree` | `true`/`false` metadata |
 | `verify` | `true`/`false` verification marker |
+| `specRefs` | T4/R6 (ADR-6): workspace SpecRecord ids this step is held to (frozen into the packet at dispatch) |
+| `specStrict` | Per-step strict override (`true`/`1`) |
 
 Each step starts with `## step-id` followed by recognized step metadata such as `role:` before the blank line. Level-2 headings inside task bodies are preserved unless they look like a step section with recognized metadata; use `###` or lower for maximum compatibility.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-crew",
-  "version": "0.9.68",
+  "version": "0.10.2",
   "description": "Pi extension for coordinated AI teams, workflows, worktrees, and async task orchestration",
   "author": "baphuongna",
   "license": "MIT",
@@ -30,6 +30,7 @@
   ],
   "files": [
     "src/",
+    "!src/benchmark",
     "dist/index.mjs",
     "index.ts",
     "install.mjs",
@@ -67,13 +68,14 @@
   ],
   "scripts": {
     "check": "npm run ci",
-    "ci": "npm run check:lockfile-sync && npm run typecheck && npm run lint && npm run format:check && npm run check:conflict-markers && npm run check:decision-drift && npm run check:event-types && npm run check:lazy-imports && npm run check:bundle-staleness && npm run build:bundle && npm run check:bundle-size && npm run test:bundle && npm test && npm pack --dry-run",
+    "ci": "npm run check:lockfile-sync && npm run typecheck && npm run lint && npm run format:check && npm run check:conflict-markers && npm run check:decision-drift && npm run check:env-vars && npm run check:event-types && npm run check:lazy-imports && npm run check:bundle-staleness && npm run build:bundle && npm run check:bundle-size && npm run test:bundle && npm test && npm pack --dry-run",
     "check:lockfile-sync": "node scripts/check-lockfile-sync.mjs",
     "check:lazy-imports": "node scripts/check-lazy-imports.mjs",
     "check:bundle-staleness": "node scripts/check-bundle-staleness.mjs",
     "check:bundle-size": "node scripts/check-bundle-size.mjs",
     "check:conflict-markers": "node scripts/check-conflict-markers.mjs",
     "check:decision-drift": "node scripts/check-decision-drift.mjs",
+    "check:env-vars": "node scripts/check-env-vars.mjs",
     "check:event-types": "node scripts/check-event-types-registry.mjs",
     "typecheck": "tsc --noEmit && node --experimental-strip-types -e \"await import('./index.ts'); console.log('strip-types import ok')\"",
     "lint": "biome check --linter-enabled=true --formatter-enabled=false --max-diagnostics=50 .",