@bonesofspring/ai-rules 0.2.10 → 0.2.12

package/CHANGELOG.md

@@ -2,6 +2,46 @@
 
 ## Unreleased
 
+### Agent-task metrics
+
+- Added local schema-v1 `metrics.json` ledgers for agent-team pipelines: append-only task, attempt, retry, skip, completion, and human-gate lifecycle events. Allowlisted identifiers/timestamps/outcomes only (no prompts, secrets, file paths/contents, or remote telemetry). Shared-core policy: `presets/_shared/core/agent-team/agent-task-metrics.md` (requestable via orchestrator; no new always-on rule).
+- Instrumented all Cursor/Claude `chain-team-phases.sh` forks with lock-safe temp-and-rename ledger/status writes, `status.json` `activeAttempts` / `lastCompletedAttempt` bridge, and conservative parallel attribution (payload `attemptId`, sole active attempt, or `.metrics-receipts.json` fallback—never guess from `currentAgent`). Repeated work-package retries now target the nearest preceding developer and return to the requesting verifier/reviewer; stale duplicate stops are ignored.
+- Human-gate resume is a hook action: `{"status":"completed","action":"end_human_gate","activateNext":true}` on the installed `hooks/chain-team-phases.sh` (`/task-continue`); optional wrapper `scripts/close-agent-task-human-gate.mjs`.
+- Added `aggregate-agent-task-metrics.mjs` (role/count/duration, gate wait, attribution confidence, and separate missing-ledger, malformed-JSON, duplicate-ID, unterminated, unmatched, and duplicate-terminal completeness categories), Next fixtures under `presets/cursor/next/team/fixtures/`, and a 14-fork lifecycle harness covering retry, concurrent parallel completion, skip, gate, and invalid-ledger scenarios. Technical-retro commands include a Metrics evidence section.
+
+### PHP Laravel preset (`php-laravel`)
+
+- New installable preset **`php-laravel`** (Cursor + Claude twins): Laravel backend/API projects — flat `app/` layout, Eloquent, Form Requests, Policies, Actions/Services, Service Providers, Jobs/Queues/Events; thin controllers with clean-architecture leanings (repositories only when justified).
+- Always-on trio `laravel-app-core`, `architecture-boundaries`, `code-quality-and-refactoring` (session-start body budget ≤140).
+- Rule catalog (34 stems): Laravel domain rules (`http-layer`, `form-requests`, `eloquent-persistence`, `actions-services`, `service-providers`, `policies-authorization`, `jobs-queues`, `events-listeners`, …); no `api-grpc` / `messaging-adapters` stubs (async covered by Jobs/Events); `integration-test-*` agents (no Playwright / XCUITest / a11y).
+- Gates: `php-tooling` + `post-change-test` (prefer `php artisan test` / `composer test`; Pint/PHPStan/Psalm only if configured in repo/CI).
+- Structure scripts include `php-laravel` (`STACKS` + `PHP_FORBIDDEN_*`); meta shared-core + embed adapters list php-laravel as 7th installable stack; chain forks pairwise vs all stacks including `php-hexagonal`.
+- **Coexistence:** distinct install id from **`php-hexagonal`** (framework-agnostic ports/adapters PHP); comparison table in stack READMEs — neither replaces the other.
+- Install: `npx @bonesofspring/ai-rules init cursor|claude --preset php-laravel`.
+- Dogfood: `dogfood/php-laravel` (full `laravel/framework` via create-project; Todo API; committed `composer.lock`; P1–P3 prompts).
+- Review follow-ups (non-blocking): secondary agents (`task-analyst`, `solution-architect`, test planners) may retain hexagonal vocabulary — Laravel pass scheduled; run `sync-dogfood-cursor.sh` before monorepo merge; `check:rule-eval` remains next-only.
+
+### PHP hexagonal preset (`php-hexagonal`)
+
+- New installable preset **`php-hexagonal`** (Cursor + Claude twins): PHP 8.x + Composer + PSR-4 hexagonal backends; framework-agnostic HTTP edge (Slim / Symfony HttpFoundation / Laravel HTTP as documented variants only).
+- Always-on trio `php-app-core`, `architecture-boundaries`, `code-quality-and-refactoring` (session-start body budget ≤140).
+- Rule catalog (34 stems) with thin stub for `api-grpc`; full `messaging-adapters` (java parity); `integration-test-*` agents (no Playwright / XCUITest / a11y).
+- Gates: `php-tooling` + `post-change-test` (Composer / PHPUnit / Pest discovery; PHPStan/Psalm/Pint only if configured in repo/CI).
+- Structure scripts include `php-hexagonal` (`STACKS` + `PHP_FORBIDDEN_*`); meta shared-core + embed adapters list php-hexagonal alongside go/java; chain forks pairwise vs other stacks.
+- Install: `npx @bonesofspring/ai-rules init cursor|claude --preset php-hexagonal`.
+- Dogfood: `dogfood/php-hexagonal` (domain → ports → HTTP/persistence/messaging adapters → composition → PHPUnit unit + integration).
+- Review follow-ups: README titles use install name `php-hexagonal`; dogfood twin-sync paths synced; leakage checker scans `team/fixtures` (`steps[]` FAIL / `skipped[]` anti-pattern OK); composition-root HTTP integration test; AB.md PHP sample note; `check:rule-eval` documented as next-only.
+
+### MCP TypeScript preset (`mcp-ts`)
+
+- New installable preset **`mcp-ts`** (Cursor + Claude twins): TypeScript MCP server authoring with `@modelcontextprotocol/sdk`, stdio-first transport, Context7-only MCP, package-manager–discovered typecheck/test via `mcp-ts-tooling` / `post-change-test`.
+- Full mapped rule catalog (~25 stems); always-on trio `mcp-ts-app-core`, `mcp-server-boundaries`, `code-quality-and-refactoring` (session-start body budget ≤140).
+- Agent-team surface (Phase 1): 6 agents (`task-router`, `task-analyst`, `solution-architect`, `feature-developer`, `build-verifier`, `security-reviewer`), 4 skills, 5 commands, hooks + team fixtures; no Playwright/XCUITest/a11y/`integration-test-*`.
+- Structure scripts (`twin-map`, `token-budget`, `leakage`) include `mcp-ts`; leakage `MCP_TS_FORBIDDEN_*` + pairwise chain forks vs other stacks.
+- Install: `npx @bonesofspring/ai-rules init cursor|claude --preset mcp-ts`.
+- Dogfood: `dogfood/mcp-ts-sample` (scenarios M1–M3 in `reference-features`).
+- Platform asymmetry: `BUGBOT.md` + `hooks.json` Cursor-only; `CLAUDE.md` Claude-only.
+
 ### Go/Java stub thicken + session-start budget raise (`go-java-rules-thicken-dogfood`)
 
 - **Session-start body cap:** **120 → 140** (stretch **≤120**); `alwaysApply` / Claude session-start count stays **exactly 3**. Checker `BODY_CAP=140`; policy core + embed adapters + build-verifier copy updated.

package/README.md

@@ -53,7 +53,7 @@ Where things go on `init`:
 
 Skills of note: `write-adr` (all stacks); `react-performance` (**next only**). Living gotchas: `team/conventions.md`. Opt-in rule samples eval: `yarn check:rule-eval` (workflow `rule-eval.yml`; not part of `check:preset-structure`).
 
-The package ships presets including **`next`** (Next.js stack), **`ios-swift`** (iOS apps on Swift / SwiftUI), **`go`** (Go backends with hexagonal architecture), and **`java`** (Java / Spring Boot hexagonal backends). Other preset names: see `presets/` in the [repo](https://github.com/bonesofspring/ai-rules). Name is a single path segment, no `/` or `..`.
+The package ships presets including **`next`** (Next.js stack), **`ios-swift`** (iOS apps on Swift / SwiftUI), **`go`** (Go backends with hexagonal architecture), **`java`** (Java / Spring Boot hexagonal backends), **`mcp-ts`** (TypeScript MCP servers with `@modelcontextprotocol/sdk`), **`php-hexagonal`** (PHP / Composer hexagonal backends), and **`php-laravel`** (Laravel backend/API with Eloquent, Form Requests, Policies, Actions/Services). Other preset names: see `presets/` in the [repo](https://github.com/bonesofspring/ai-rules). Name is a single path segment, no `/` or `..`.
 
 ```bash
 npx @bonesofspring/ai-rules init cursor --preset ios-swift
@@ -62,6 +62,12 @@ npx @bonesofspring/ai-rules init cursor --preset go
 npx @bonesofspring/ai-rules init claude --preset go
 npx @bonesofspring/ai-rules init cursor --preset java
 npx @bonesofspring/ai-rules init claude --preset java
+npx @bonesofspring/ai-rules init cursor --preset mcp-ts
+npx @bonesofspring/ai-rules init claude --preset mcp-ts
+npx @bonesofspring/ai-rules init cursor --preset php-hexagonal
+npx @bonesofspring/ai-rules init claude --preset php-hexagonal
+npx @bonesofspring/ai-rules init cursor --preset php-laravel
+npx @bonesofspring/ai-rules init claude --preset php-laravel
 ```
 
 Help: `ai-rules --help`.

package/bin/cli.js

@@ -73,10 +73,16 @@ Examples:
   npx @bonesofspring/ai-rules init cursor --preset ios-swift
   npx @bonesofspring/ai-rules init cursor --preset go
   npx @bonesofspring/ai-rules init cursor --preset java
+  npx @bonesofspring/ai-rules init cursor --preset mcp-ts
+  npx @bonesofspring/ai-rules init cursor --preset php-hexagonal
+  npx @bonesofspring/ai-rules init cursor --preset php-laravel
   npx @bonesofspring/ai-rules init claude
   npx @bonesofspring/ai-rules init claude --preset ios-swift
   npx @bonesofspring/ai-rules init claude --preset go
   npx @bonesofspring/ai-rules init claude --preset java
+  npx @bonesofspring/ai-rules init claude --preset mcp-ts
+  npx @bonesofspring/ai-rules init claude --preset php-hexagonal
+  npx @bonesofspring/ai-rules init claude --preset php-laravel
   npx @bonesofspring/ai-rules clean cursor
 `);
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bonesofspring/ai-rules",
-  "version": "0.2.10",
+  "version": "0.2.12",
   "description": "Presets of Cursor and Claude rules/commands for Revy Ross personal use",
   "license": "MIT",
   "author": "Revy Ross",

package/presets/_shared/README.md

@@ -1,6 +1,6 @@
 # Shared preset core (`presets/_shared`)
 
-Stack-agnostic **prose and checklists** for all installable presets: `next`, `ios-swift`, `go`, and `java` (`cursor/<stack>`, `claude/<stack>`).
+Stack-agnostic **prose and checklists** for all installable presets: `next`, `ios-swift`, `go`, `java`, `mcp-ts`, and `php-hexagonal` (`cursor/<stack>`, `claude/<stack>`).
 
 **Not an installable preset.** CLI `init` copies only under `presets/cursor/<name>/` or `presets/claude/<name>/` — this sibling is never shipped to consumer apps.
 

package/presets/_shared/core/agent-team/agent-task-metrics.md

@@ -0,0 +1,30 @@
+# Agent-task metrics
+
+Metrics are local task artifacts for evaluating pipeline effectiveness. They are not remote telemetry and must never contain prompts, secrets, file paths or contents, user identifiers, model transcripts, or arbitrary error text.
+
+## Ledger contract
+
+Each pipeline run owns `<team>/tasks/<slug>/metrics.json`:
+
+```json
+{
+  "schemaVersion": 1,
+  "task": { "runId": "uuid", "slug": "example", "preset": "ai-rules", "stack": "next", "platform": "cursor", "intent": "feature", "profile": "standard" },
+  "events": []
+}
+```
+
+`events` is append-only. A task is one `runId`; a subtask is one invocation of `pipeline.steps[i]`. Every invocation has an `attemptId`, including each member of a parallel step. Events inherit the immutable task dimensions rather than duplicating them.
+
+Allowed event types are `task_started`, `attempt_started`, `attempt_finished`, `retry_scheduled`, `attempt_skipped`, `human_gate_started`, `human_gate_ended`, and `task_completed`. Terminal outcomes are `completed`, `changes_requested`, `validation_failed`, `skipped`, `cancelled`, or `unknown`.
+
+## Collection policy
+
+1. The router creates the ledger and task dimensions, then records `task_started` and an `attempt_started` immediately before every invocation.
+2. The completion hook uses the host clock for `attempt_finished`, `task_completed`, and gate edges. It computes `durationMs` from the stored start timestamp; agents must not calculate durations or rewrite the ledger.
+3. `status.json.activeAttempts` is a transient bridge keyed by `attemptId`; `lastCompletedAttempt` is diagnostic only. The ledger is the history authority.
+4. A retry ends its prior attempt, emits `retry_scheduled`, then creates a new attempt number. In repeated work-package pipelines, the fix targets the nearest preceding developer step and then returns to the exact verifier/reviewer step that requested the fix. A skipped step emits `attempt_skipped`. Human-gate wait is recorded with separate start/end events and never added to attempt duration.
+5. For parallel completion, the hook accepts a payload `attemptId` only when it directly names an active attempt. A supplied stale or unknown `attemptId` is ignored and must not fall back to another active attempt or advance the pipeline. Other runtime IDs have no verified start-time mapping and are not stored as `runtimeIdentity`. If no direct `attemptId` exists, attribute only when exactly one active attempt exists. Otherwise accept an explicit task-local completion receipt and mark it `source: "agent"` with `attributionConfidence: "fallback"`; never guess from `currentAgent`.
+6. Writers use a per-task lock and temporary-file rename for both ledger and status updates so concurrent completions do not lose events or state. Aggregation reports missing ledgers, malformed JSON, duplicate attempt IDs, unterminated attempts, unmatched terminals, and duplicate terminals as separate incomplete-data categories; none count as a successful task.
+
+Only allowlisted identifiers and timestamps belong in events. Use `source` (`router`, `hook`, `agent`) and `attributionConfidence` (`high`, `fallback`, `unknown`) to expose incomplete collection rather than invent precision.

package/presets/_shared/core/agent-team/agent-team-orchestrator.md

@@ -54,6 +54,12 @@ flowchart TD
 
 **Source of truth for order:** `pipeline.json` → `steps[]`. Never hardcode analyst → dev → review → QA when `pipeline.json` exists.
 
+## Metrics lifecycle
+
+For local metrics policy and the `metrics.json` ledger, see **agent-task-metrics**. The router records `task_started` and one `attempt_started` before each invocation; the hook records host-clock terminal and human-gate events. Keep active attempt identities in `status.json.activeAttempts` so parallel completions can be bound safely.
+
+Do not calculate durations in agents. For parallel steps, never attribute a stop using `currentAgent` alone: use a stable runtime identity, the sole active attempt, or an explicit receipt marked with fallback confidence.
+
 ## status.json (pipeline mode)
 
 ```json
@@ -64,7 +70,9 @@ flowchart TD
   "currentAgent": "task-analyst",
   "phase": "executing",
   "state": "in_progress",
-  "awaitingHumanGate": false
+  "awaitingHumanGate": false,
+  "activeAttempts": {},
+  "lastCompletedAttempt": null
 }
 ```
 

package/presets/_shared/core/meta/preset-layering.md

@@ -13,7 +13,7 @@
 ## Куда класть новое правило
 
 1. Общие принципы для всех стеков → **core** + адаптеры с `<!-- shared-core: … -->`.
-2. Только один из installable стеков (`next` / `ios-swift` / `go` / `java`) → **stack-specific** (не в `_shared`).
+2. Только один из installable стеков (`next` / `ios-swift` / `go` / `java` / `mcp-ts` / `php-hexagonal` / `php-laravel`) → **stack-specific** (не в `_shared`).
 3. Только Cursor или только Claude packaging → platform artifact / frontmatter, не дублировать семантику.
 
 ## SoT

package/presets/_shared/core/meta/preset-no-cross-stack-leakage.md

@@ -6,7 +6,8 @@
 
 - `next` must not require iOS-only agents/tooling (`xcuitest-*`, `xcodebuild` as required handoff, Keychain-only rules).
 - `ios-swift` must not ship Playwright handoffs or Next lint recipes as required commands.
-- `go` / `java` must not ship Playwright, XCUITest, Next lint recipes, or `xcodebuild`/`SwiftLint` as required handoffs.
+- `go` / `java` / `php-hexagonal` / `php-laravel` / `mcp-ts` must not ship Playwright, XCUITest, Next lint recipes, or `xcodebuild`/`SwiftLint` as required handoffs.
+- `mcp-ts` must not ship Go tool commands (`go test`, `go vet`, `golangci-lint`) or `integration-test-*` agents as required delivery gates.
 - Shared core must not embed stack paths (`app/src/**`, `Features/**`, `playwright`, `xcodebuild`).
 
 ## Forbidden in ios-swift hooks/agents (FAIL)
@@ -24,16 +25,28 @@ Required instead: `xcuitest-test-*` stems; build-verifier mentions `xcodebuild`
 - `xcuitest-test-` as handoff stems
 - Requiring `xcodebuild` / SwiftLint as the default post-change gate for Next app work
 
-## Forbidden in go / java hooks/agents (FAIL)
+## Forbidden in go / java / php-hexagonal / php-laravel hooks/agents (FAIL)
 
-Shared backend-stack bans (both `go` and `java`):
+Shared backend-stack bans (`go`, `java`, `php-hexagonal`, and `php-laravel`):
 
 - `playwright-test-`, `playwright-e2e`, `user-playwright`, `app/__tests__/e2e`
 - Positive yarn recipes `lint:js` / `lint:css`
 - `xcuitest-test-`, `xcodebuild`, `SwiftLint` as required handoffs
 - Agent file `accessibility-reviewer` (no UI surface)
 
-Required instead: `integration-test-*`; build-verifier / `post-change-test` plus stack tooling — `go-tooling` (`go test`, `go vet`) or `java-tooling` (`mvn test` / `./gradlew test`).
+Required instead: `integration-test-*`; build-verifier / `post-change-test` plus stack tooling — `go-tooling` (`go test`, `go vet`), `java-tooling` (`mvn test` / `./gradlew test`), or `php-tooling` (`php artisan test` / `composer test` / `vendor/bin/phpunit`).
+
+## Forbidden in mcp-ts hooks/agents (FAIL)
+
+Same UI/mobile bans as backend stacks, plus:
+
+- `playwright-test-`, `playwright-agents`, `playwright-e2e`, `user-playwright`, `app/__tests__/e2e`
+- Positive yarn recipes `lint:js` / `lint:css` as required gates
+- `xcuitest-test-`, `xcodebuild`, `SwiftLint` as required handoffs
+- Go tool commands as required MCP delivery gates: `go test`, `go vet`, `golangci-lint`
+- Agent files: `playwright-test-*`, `xcuitest-test-*`, `integration-test-*`, `accessibility-reviewer`
+
+Required instead: package-manager–discovered typecheck/test (`mcp-ts-tooling`, `post-change-test`); Phase 1 gates `build-verifier` + `security-reviewer`.
 
 ## Soft / document
 
@@ -42,4 +55,4 @@ Required instead: `integration-test-*`; build-verifier / `post-change-test` plus
 
 ## Chain fork
 
-`hooks/chain-team-phases.sh` must differ **byte-wise** pairwise among `next`, `ios-swift`, `go`, and `java` (per platform).
+`hooks/chain-team-phases.sh` must differ **byte-wise** pairwise among `next`, `ios-swift`, `go`, `java`, `mcp-ts`, `php-hexagonal`, and `php-laravel` (per platform).

package/presets/_shared/core/meta/preset-pr-checklist.md

@@ -6,7 +6,7 @@
 
 - [ ] **SoT:** правки в `packages/…` (не только root `.cursor/`)
 - [ ] **Core:** если менялся `_shared/core/**` — обновлены все `<!-- shared-core: … -->` consumers (или documented intentional fork)
-- [ ] **Twin:** Cursor change → Claude twin **same PR** (**FAIL** if missing on next, ios-swift, go, and java — см. `preset-twin-sync`)
+- [ ] **Twin:** Cursor change → Claude twin **same PR** (**FAIL** if missing on next, ios-swift, go, java, mcp-ts, php-hexagonal, and php-laravel — см. `preset-twin-sync`)
 - [ ] **Mapping:** `claude/<stack>/rules/README.md` mapping table актуален
 - [ ] **Token budget:** alwaysApply / session-start ≤ 3; meta rules not always-on
 - [ ] **Leakage:** нет чужих stack agent stems / handoff strings (`preset-no-cross-stack-leakage`)

package/presets/_shared/core/meta/preset-twin-sync.md

@@ -10,6 +10,9 @@ Cursor `.mdc` — **SoT per stack**. Claude topic `.md` — derived twin.
 | `ios-swift` | `presets/cursor/ios-swift/rules/*.mdc` | `presets/claude/ios-swift/rules/<topic>/` | Keep in `claude/ios-swift/rules/README.md` |
 | `go` | `presets/cursor/go/rules/*.mdc` | `presets/claude/go/rules/<topic>/` | Keep in `claude/go/rules/README.md` |
 | `java` | `presets/cursor/java/rules/*.mdc` | `presets/claude/java/rules/<topic>/` | Keep in `claude/java/rules/README.md` |
+| `mcp-ts` | `presets/cursor/mcp-ts/rules/*.mdc` | `presets/claude/mcp-ts/rules/<topic>/` | Keep in `claude/mcp-ts/rules/README.md` |
+| `php-hexagonal` | `presets/cursor/php-hexagonal/rules/*.mdc` | `presets/claude/php-hexagonal/rules/<topic>/` | Keep in `claude/php-hexagonal/rules/README.md` |
+| `php-laravel` | `presets/cursor/php-laravel/rules/*.mdc` | `presets/claude/php-laravel/rules/<topic>/` | Keep in `claude/php-laravel/rules/README.md` |
 
 ## Same-PR checklist
 
@@ -29,5 +32,8 @@ Cursor `.mdc` — **SoT per stack**. Claude topic `.md` — derived twin.
 - **ios-swift:** missing Claude twin for a changed Cursor rule → **FAIL**.
 - **go:** missing Claude twin for a changed Cursor rule → **FAIL** (same as next / ios-swift; mapping table required).
 - **java:** missing Claude twin for a changed Cursor rule → **FAIL** (same as next / ios-swift / go; mapping table required).
+- **mcp-ts:** missing Claude twin for a changed Cursor rule → **FAIL** (same as next / ios-swift / go / java; mapping table required).
+- **php-hexagonal:** missing Claude twin for a changed Cursor rule → **FAIL** (same as next / ios-swift / go / java / mcp-ts; mapping table required).
+- **php-laravel:** missing Claude twin for a changed Cursor rule → **FAIL** (same as next / ios-swift / go / java / mcp-ts / php-hexagonal; mapping table required).
 
 Dogfood: sync packages → root `.cursor/` after Cursor SoT changes (not the reverse). Use `packages/ai-rules/scripts/sync-dogfood-cursor.sh`.

package/presets/_shared/core/review/technical-retro.md

@@ -4,4 +4,6 @@
 
 **`commands/technical-retro.md`** → `/technical-retro`
 
+When a task has `metrics.json`, use the local aggregation script and state the cohort, sample size, incomplete artifacts, attribution confidence, active attempt duration, and human-gate wait separately. Do not score malformed artifacts or gate wait as agent work.
+
 Используй rule только если command недоступен. Не дублируй содержимое command в ответе — следуй command-файлу и skill `technical-retro`.

package/presets/claude/go/agents/build-verifier.md

@@ -29,7 +29,7 @@ Hard gates: twin-map (stacks include **go**), token-budget (alwaysApply/session-
 
 Document PASS/FAIL in `validation-report.md`.
 
-1. **Root AGENTS template:** each installable preset (`cursor|claude` × `next|ios-swift|go|java`) must ship `REPO_AGENTS.md` when validating preset packaging → **FAIL** if missing.
+1. **Root AGENTS template:** each installable preset (`cursor|claude` × `next|ios-swift|go|java|mcp-ts|php-hexagonal`) must ship `REPO_AGENTS.md` when validating preset packaging → **FAIL** if missing.
 2. **Hooks cookbook:** default `hooks.json` must **not** reference `hooks/examples/` → **FAIL** if it does.
-3. **No react-performance on ios/go/java:** skill directory / required handoff must be absent under ios-swift, go, and java → **FAIL** (leakage). Prefer `check:preset-leakage`.
+3. **No react-performance on ios/go/java:** skill directory / required handoff must be absent under ios-swift, go, java, mcp-ts, php-hexagonal, and php-laravel → **FAIL** (leakage). Prefer `check:preset-leakage`.
 4. **Rule eval (WARN):** `scripts/check-rule-eval.mjs` / `yarn check:rule-eval` should exist for maintainers; smoke optional. Hard fail only when the opt-in eval job runs — not part of default `check:preset-structure`.

package/presets/claude/go/agents/task-router.md

@@ -67,3 +67,7 @@ Write `.cursor/team/tasks/<slug>/pipeline.json` and initial `status.json` via th
 ## Handoff
 
 Intent + steps table + skipped + gates → invoke `steps[0].agent` (unless retro).
+
+## Metrics initialization
+
+Alongside `pipeline.json` and `status.json`, create `metrics.json` with `schemaVersion: 1`, immutable task dimensions, `task_started`, and a generated `runId`. Before the parent invokes each agent, append `attempt_started` and set `status.activeAttempts[attemptId]` with agent, step index, attempt number, and host-clock `startedAt`. For parallel work create one entry per agent; do not use `currentAgent` to identify their completion.

package/presets/claude/go/commands/task-continue.md

@@ -22,22 +22,19 @@ description: Continue an agent-team task after a human gate, pause, or review ch
 ## Алгоритм
 
 1. Запиши `.claude/team/active-task.json` → `{ "slug": "<slug>" }`.
-2. Сбрось gate: `awaitingHumanGate: false`.
-3. Определи **следующий шаг**:
-   - После approval analyst/architect: `pipelineIndex + 1` → следующий agent в `steps`.
-   - После `changes_requested`: снова **feature-developer** (тот же index или найди developer в steps).
-4. Обнови `status.json`:
-   ```json
-   {
-     "pipelineIndex": <n>,
-     "currentAgent": "<steps[n].agent>",
-     "phase": "executing",
-     "state": "in_progress",
-     "awaitingHumanGate": false
-   }
-   ```
-5. Вызови subagent для `currentAgent` с контекстом slug и артефактов.
-6. Если hook `chain-team-phases.sh` включён в проекте и `autoChain: true`, он продолжит цепочку по `pipeline.json`; иначе продолжай следующий шаг вручную.
+2. **Human gate resume** (`awaitingHumanGate` / `awaiting_approval`):
+   1. **Do not** manually increment `pipelineIndex` or invent `currentAgent` first.
+   2. Close the gate + activate the next attempt via the **installed** hook (single authority for ledger + status):
+
+```bash
+printf '%s\n' '{"status":"completed","action":"end_human_gate","activateNext":true}' | bash .claude/hooks/chain-team-phases.sh
+```
+
+   3. Re-read `status.json` (and the hook `followup_message`).
+   4. Invoke the subagent named in `status.currentAgent` / followup — do **not** advance the pipeline a second time.
+3. **Review `changes_requested`** (not a human gate): set `currentAgent` to `feature-developer` (same index / find developer in steps), `state: in_progress`, then invoke developer. Do **not** call `end_human_gate` unless `activeHumanGateEventId` is set.
+4. **Other resume**: restore `pipelineIndex` / `currentAgent` from status and invoke that agent.
+5. After the invoked agent completes, `chain-team-phases.sh` on `subagentStop` continues the chain when `autoChain: true`.
 
 ## Артефакты для handoff
 
@@ -47,3 +44,7 @@ description: Continue an agent-team task after a human gate, pause, or review ch
 | build-verifier | git diff, decomposition.md |
 | code-reviewer | brief.md, validation-report.md, git diff |
 | qa-tester | brief.md, decomposition.md, scope from pipeline step |
+
+## Metrics
+
+On human-gate resume, the hook action above appends `human_gate_ended` under the per-task lock **before** any new `attempt_started`. Agents must not calculate durations.

package/presets/claude/go/commands/task.md

@@ -47,3 +47,7 @@ description: Route a natural-language task through the agent-team pipeline: task
 ## Legacy
 
 `/feature-start` и `/feature-continue` остаются совместимыми; для новых задач предпочитай `/task`.
+
+## Metrics
+
+Initialize `metrics.json` before the first invocation with frozen task dimensions and `task_started`; add one `activeAttempts[attemptId]` entry plus `attempt_started` per scheduled agent. On `/task-continue`, append `human_gate_ended` (referencing the stored gate event) before activating the next attempt. Do not calculate durations in agents.

package/presets/claude/go/commands/technical-retro.md

@@ -56,3 +56,7 @@ description: Facilitate a technical retrospective for a period, incident, releas
 - Не приписывать команде цитаты или факты, которых не было во входе.
 - Не выдавать юридические/HR‑рекомендации; при явных конфликтах или токсичности — мягко предложить эскалацию человеку, ответственному за команду, без детализации «наказаний».
 - Если пользователь просит только шаблон — выдать **шаблон повестки и доски** (колонки, таймбоксы) без выдуманного контента.
+
+### Metrics evidence
+
+When `metrics.json` exists, run the local aggregation script and report the cohort, sample size, incomplete artifacts, attribution confidence, active attempt duration, and human-gate wait separately. Do not treat gate wait or malformed ledgers as agent duration.