peaks-loop 4.0.0-beta.8 → 4.0.1

package/skills/peaks-code/references/startup-sequence.md

@@ -0,0 +1,110 @@
+---
+name: peaks-code-startup-sequence
+description: peaks-code startup sequence Steps 0 / 0.5-0.87 / 1 / 2 / 2.3 / 2.5 / N / N+1 / N+2 — extracted from SKILL.md 2026-07-16 to keep SKILL.md under 25,000-byte cap
+---
+
+# Peaks-Loop Startup sequence (MANDATORY — execute in order)
+
+> **Extracted from SKILL.md on 2026-07-16.** This file is the
+> authoritative reference; SKILL.md points here with a 1-line pointer.
+> Do not edit `SKILL.md` for content changes — edit this file instead.
+
+### Peaks-Loop Slice 011 — workspace consolidate + session checkpoint/resume
+
+`peaks workspace consolidate` is the slice-011 umbrella primitive. `peaks session checkpoint` (context-overflow defense) and `peaks session resume` surface as Step 0.75 + Step N. See `references/cross-date-session-check.md`.
+
+### Peaks-Loop Step 0.75: Resume from checkpoint (BLOCKING on same-day re-invocation)
+
+When a NEW conversation opens on a session whose `lastActivity` is from today AND `.peaks/_runtime/<sessionId>/checkpoints/` has `*.json`, surface via `peaks session info --active --json` + `peaks session resume --from <path> --project <repo>`. Prompt via `AskUserQuestion` (resume / fresh); if "resume", prepend the block. Step 0.75 is no-op on precondition fail. See `references/checkpoint-resume.md`.
+
+### Peaks-Loop Step N: Periodic checkpoint (auto-fire, no user action)
+
+Proactive context-overflow defense. CLI: `peaks session checkpoint [--reason <r>] [...flags]`. Fire `--reason periodic` every 20 tool calls (G1 hard-coded), `--reason artifact-written` per PRD/RD/QA/TXT write. See `references/periodic-checkpoint.md` (otherwise `tests/unit/code/checkpoint-periodic-frequency.test.ts` fails).
+
+### Peaks-Loop Step 0.6: Audit + Goal (NEW)
+
+Invoke peaks-audit after need expressed; store approved goal at `.peaks/_runtime/<sessionId>/audit-goal/<rid>.json`. **All autonomous work requires approved goal.** → `references/audit-goal-gate.md`.
+
+### Peaks-Loop Step N+1: Final Review (NEW)
+
+After autonomous work (RD, QA, security, perf), invoke peaks-final-review for 4-dim evidence. All 4 pass → final delivery; else loop back. See `references/final-review-gate.md`.
+
+### Peaks-Loop Step N+2: Main-session context monitor (D6 + slice 2026-07-02)
+
+After every 4th tool call, probe via `peaks code context-now --project <repo> --json`. Thresholds: 50% / **0.85 pre-compact / 0.95 red-line** (v2.13.0). **In Job mode ≥ 0.85 is MANDATORY auto-compact** (`peaks compact auto --execute`). Pass `--enforce-job-mode` for single-rid. **≥ 0.95** red line: next Bash/Task fires `peaks compact auto`. **Karpathy §4 exception**: compact red line keeps runner alive — zero-intervention wins; LLM MUST NOT ask user to run `/compact`. Honor `--in-flight-batch` (D6.e).
+
+### Peaks-Loop Step 0: Anchor the workflow (MANDATORY FIRST ACTIONS — no bail-out)
+
+Run `peaks workspace init` + `peaks skill presence:set peaks-code` BEFORE any analysis, role handoff, or mode-selection question. Even "分析下这个项目" / "看一下代码" / one-line questions anchor first.
+
+→ see `references/anchoring-and-session-info.md` for the full Step 0 contract + session-conflict resolution.
+
+#### Step 0 bypass — Claude Code [Fact-Forcing Gate] (slice 2.0.1-bug3)
+
+`peaks workspace init` materializes `.claude/settings.local.json` (a PreToolUse hook allow-list for `.peaks/**` paths and `peaks <subcommand>` Bash) so the [Fact-Forcing Gate] does not block Step 0. Pass `--no-claude-hooks` to opt out.
+
+→ see `references/anchoring-and-session-info.md` for the full bypass contract, recovery flow, and anti-bail-out rule.
+
+### Peaks-Loop Step 0.7: Detect unfinished work and offer resume (BLOCKING on first invocation per session)
+
+After Step 0, run the resume-detection probe; surface via `AskUserQuestion` if a slice is in flight. **v3.1.2 resume rule:** if `.peaks/_runtime/<sessionId>/job/<jid>/progress.json` exists, read FIRST and surface `Next: slice #N of M`. **v2.11.0 D7 override:** if user just `/compact`ed, run `peaks code post-compact-detect --project <repo> --json` FIRST; `shouldAutoResume: true` skips AskUserQuestion. → `references/resume-detection.md`.
+
+### Peaks-Loop Step 0.55: 1.x → 2.0 detection (BLOCKING on first invocation per session, when the project is not on a 2.0 layout)
+
+After Step 0.7 returns "fresh", run `peaks upgrade --detect-1x --project <root> --json`. If `isOneX: true`, surface `AskUserQuestion`. → `references/step-0-55-1x-detection.md`.
+
+### Peaks-Loop Step 0.8 — Job 启动 (BLOCKING on LLM judgement — v3.1.1 patch + v3.1.2 mechanical gates)
+
+> **CLI reality check (D-001 sediment, 2026-07-09):** The CLI surface has changed since this section was last verified. The actual command is **`peaks job init --job-id <jid> --slice-list <list> --main-loop-strategy <single|rotating> [--parallelism-hint <serial|llm-decides>] [--exit-policy <strict|best-effort>] [--project <repo>]`**. The legacy `peaks code detect-job --is-job/--suggested-job-id/--confidence` form described below is **no longer present** in 4.0.0-beta.6. If you find `--is-job` rejected, fall back to `peaks job init`.
+
+The CLI is a **recorder + gate** for job-shape (the LLM judges). LLM calls `peaks job init --job-id <jid> --slice-list <list> --main-loop-strategy <single|rotating>` (with `--caller-id <id>` and `--project <repo>` in non-Claude-Code environments), which writes `.peaks/_runtime/<sessionId>/job/<jid>/state.json` and emits a `[job-event] {kind: job-started, jobId, total, strategy}` line. Downstream steps call `peaks job status --job-id <jid>` to read state.
+
+**Hard rule (v3.1.1 red-line #10):** LLM MUST NOT skip the job-init call. If `peaks job status` reports `done: 0` and the slice work has begun, the LLM forgot to initialize — re-call `peaks job init` immediately.
+
+> **D-003 sediment (2026-07-09):** The historical `JOB_SHAPE_NOT_DECIDED` exception thrown by `read-job-shape` is no longer wired in 4.0.0-beta.6. The current CLI uses `peaks job status` which reports a passive `done: 0` instead of throwing. Both behaviors enforce the same intent (no work may begin before init) but the failure mode is a soft warning rather than a hard exception. LLM should treat `done: 0` after Step 0.8 as a recoverable miss, not an unrecoverable error.
+
+**v3.1.2 mechanical gates** (recorder-only was bypassed twice):
+
+PreToolUse hook on `peaks code gate-step-08`; size-fear ban on `peaks code emit-handoff`; forced auto-compact at ≥ 0.85 in Job mode; on-disk slice progress via `peaks job checkpoint` (canonical reader `peaks job progress --job-id <jid>`). Full mechanics (judgement criteria, hook table, backup-regex rationale, hook wiring) at `references/step-0-8-gate.md`.
+
+### Peaks-Loop Step 0.81 — per-slice 收尾
+
+After commit: `peaks job checkpoint --slice-id <rid> --state done --commit-sha $(git rev-parse HEAD)` → `peaks job status` → loop (remaining>0 → Step 1; ==0 → Step 8/9/10/11; blocked → Step 0.85).
+
+### Peaks-Loop Step 0.85 — slice 阻塞处理
+
+Trigger: `repair-status` atCap=true, context-now red-line ≥5min, or `subagent-cleanup` fails twice. Action: `peaks job block --slice-id <rid> --reason "<reason>"` then STOP.
+
+### Peaks-Loop Step 0.86 — main session rotation
+
+Active in rotating mode; fires every `rotateEvery` slices or via `peaks job rotate-now`. Sequence: cycle-summary → checkpoint rotate-marker → `peaks session rotate` → resume via `peaks session resume --job-id <jid>`.
+
+### Peaks-Loop Step 0.87 — sub-agent cleanup gate
+
+After every `peaks sub-agent dispatch --batch-id <id>` inside a Job, BEFORE next slice checkpoint: `peaks job subagent-cleanup --job-id <jid> --batch-id <id> --force`. Non-zero → block.
+
+→ see `references/job-loop.md` for state machine, visibility, rotation cadence, cross-day recovery, 9 red lines.
+
+### Peaks-Loop Step 1: Mode selection
+
+Use `AskUserQuestion` with `Full auto (Recommended)` first when user did not name a profile OR skill presence is stale (run `peaks skill presence:check-stale --project <path> --json` first; `stale: true` ⇒ re-ask). → `references/mode-selection.md`, `references/fast-mode.md`, `references/mode-selection-with-stale-presence.md`.
+
+### Peaks-Loop Step 2: Re-set skill presence with the chosen mode
+
+Re-run `peaks skill presence:set peaks-code --mode <mode-value> --gate startup`. Install statusline on first run (`peaks statusline install`). → `references/skill-presence-and-title.md`.
+
+### Peaks-Loop Step 2.3: Load project memory (durable, LLM-authored memories)
+
+Run `peaks project memories --project <repo> --json` to read decisions / conventions / modules / rules / lessons from `.peaks/memory`. → `references/project-memory-loading.md`.
+
+### Peaks-Loop Step 2.5: Set session title
+
+Extract a short title from the user's first request (8-20 Chinese chars or 4-10 English words). Run `peaks session title <sessionId> "<title>" --json`. Skip if a title is already set. → `references/skill-presence-and-title.md` (same file as Step 2). D-002: sid is positional, NOT `--session-id` flag (rejected with `error: unknown option`).
+
+## Sub-agent session sharing (MANDATORY — one conversation = one sid)
+
+When peaks-code dispatches a sub-agent (peaks-rd, peaks-qa, peaks-ui, peaks-txt, peaks-sc), the prompt MUST include the parent's sid. The sub-agent MUST NOT call `peaks workspace init` (would orphan the parent's binding).
+
+## Boundaries
+
+Peaks-Loop Code may identify scenarios, recommend profiles, coordinate role skills via artifacts, coordinate project memory extraction, request user confirmation at risk/commit boundaries. MUST NOT silently install hooks / create agents / enable MCP / modify Claude settings / create GitHub repos / bypass role-skill artifacts. → `references/boundaries.md`.

package/skills/peaks-code/references/step-0-8-gate.md

@@ -45,7 +45,7 @@ Refuses to emit a final handoff while `remaining > 0` under Job mode. Pass `--fo
 
 ### 3. Forced auto-compact — `--enforce-job-mode`
 
-`peaks code context-now --enforce-job-mode` (auto-enabled when `job-shape.json` says `isJob=true`) returns `action: 'auto-compact-now'` at ≥ 0.85; Code MUST call `peaks session auto-compact --execute` without confirmation.
+`peaks code context-now --enforce-job-mode` (auto-enabled when `job-shape.json` says `isJob=true`) returns `action: 'auto-compact-now'` at ≥ 0.85; Code MUST call `peaks compact auto --execute` without confirmation.
 
 ### 4. On-disk slice progress — `progress.json`
 

package/skills/peaks-code/references/sub-agent-dispatch.md

@@ -85,7 +85,10 @@ peaks sub-agent dispatch <role> --prompt <text> [--request-id <rid>] [--session-
       "artifactBytes": 0,
       "totalBytes": 4321
     },
-    "artifactMetas": []
+    "artifactMetas": [],
+    "orchestratorVisibleHint": "⏳ Spawning sub-agent via Task tool: rd for rid=002-2026-06-07, batch-id=… (ETA ~45s)",
+    "artifactsPublicPaths": [],
+    "expectedCompletionSeconds": 45
   },
   "warnings": [],
   "nextActions": [
@@ -248,6 +251,94 @@ once the poller has read them.
 - AC-25..AC-32: G5 resource lifecycle (RL-1 batch counter, dispatch record schema, reducer dispose, slice archive, cancel dispose)
 - AC-33..AC-37: G6 heartbeat CLI, schema, poller, status line, E2E
 
+## G11.5 — visibility contract (slice 2026-07-28-sub-agent-visibility)
+
+The dispatch CLI is a `dry-run` primitive — it never spawns the Task
+tool itself. The LLM orchestrator (peaks-code / peaks-rd / peaks-qa)
+executes the returned `data.toolCall` via the IDE's `Task` agent. The
+harness does not always render the spawned agent's progress in the
+user's terminal, so the user frequently sees "peaks dispatched" but
+no Task UI feedback. This rule closes that gap with three
+machine-readable + one prose obligation.
+
+**Envelope additions (single dispatch and `--from-dag`):**
+
+- `data.orchestratorVisibleHint: string` — a single-line prose string
+  the orchestrator MUST echo verbatim in the user-facing turn
+  (i.e. in the orchestrator's response, NOT just the tool call).
+  Format: `⏳ Spawning sub-agent via Task tool: <role> for rid=<rid>, batch-id=<batchId> (ETA ~<seconds>s)`.
+- `data.artifactsPublicPaths: readonly string[]` — public artifact
+  paths the sub-agent will own. Empty when `--write-artifact` is not
+  passed; otherwise contains the single user-supplied path. The
+  orchestrator surfaces this list before fan-out so the user can see
+  what files will appear on disk.
+- `data.expectedCompletionSeconds: number` — ETA budget. Single
+  dispatch defaults to 45s; `--from-dag` defaults to 60s. Use as a
+  hint, not a hard deadline (the heartbeat poller at §G6 still owns
+  the 5-min stale threshold).
+
+**Orchestrator obligations (text-locked, no skip):**
+
+1. Before every `Agent` tool call that materializes a `data.toolCall`
+   returned by `peaks sub-agent dispatch`, the orchestrator MUST emit
+   one line of prose to the user containing
+   `⏳ Spawning sub-agent via Task tool: <description>`. `<description>`
+   MUST be ≥ 6 chars and MUST include the rid. (`description` is the
+   short label rendered by the IDE; the prose line is the user-visible
+   counter-part.) The line is rendered in the assistant's reply text,
+   not in the tool result block.
+2. When the CLI returns a fan-out envelope (`dispatchCount > 1`),
+   the orchestrator MUST emit the prose line once for the whole fan-out
+   (not once per leaf) and then issue all `Agent` tool calls in the
+   same turn. Repeating the line N times defeats the point of fan-out.
+3. If the orchestrator cannot satisfy (1) (e.g. an inline tool call
+   failed and the orchestrator wants to retry), it MUST explain the
+   retry in the next user-visible message — silent retry is a
+   visibility regression.
+
+**Why the prose, not just the envelope field**: the
+`orchestratorVisibleHint` field is machine-readable, but a terminal
+user reads prose. The LLM MUST transcribe the hint into the visible
+reply. The two surfaces are not redundant — the field is for
+downstream tooling, the prose is for the human. Same payload,
+different readers.
+
+**Regression guard**: `tests/unit/dispatch/dispatch-fanout-mandatory.test.ts`
+parses the same SKILL.md text and asserts the G11.5 paragraph is
+present; do not delete the heading without also updating the guard.
+
+---
+
+## D21 — LLM-side completion signal (mandatory)
+
+`peaks sub-agent dispatch` writes the record in `queued` state. The
+dispatcher has no other way to learn that the spawned Task tool
+actually returned. **The LLM runner MUST call `peaks sub-agent finalize`
+once per dispatched Task, in the post-completion branch** (after the
+Task tool's `tool_result` has been received). Without it, the record
+stays `queued, no-execution` forever.
+
+```sh
+# After each Task tool returns:
+peaks sub-agent finalize --request-id <rid> --outcome done
+
+# On Task failure:
+peaks sub-agent finalize --request-id <rid> --outcome failed --error "<msg>"
+
+# Crash-recovery sweep (only if the runner was interrupted and many
+# records are stuck in queued):
+peaks sub-agent finalize --all-stale --outcome done
+```
+
+`finalize` calls `markCompleted` (in dispatch-record-writer.ts), which
+transitions the record to the chosen status and unregisters the path
+from `active-dispatches.json`. The next runner does not see a stale
+ghost.
+
+This was added in slice 2026-07-17-D21 after 13 stale records were
+observed in the peaks-loop repo (`active-dispatches.json` would have
+grown unboundedly without this fix).
+
 ## How to apply
 
 When writing a SKILL.md that fans out sub-agents:

package/skills/peaks-code/references/worktree-governance.md

@@ -0,0 +1,123 @@
+# Worktree governance — 3-layer design
+
+> **Source of truth.** This document is the canonical reference for peaks-loop's worktree posture. The SKILL.md `## Worktree governance` chapter carries the heading + 1-line pointers; the full design lives here. Editing peaks-loop source propagates to every consumer project on `npm install`.
+
+## The problem this solves
+
+Sub-agents using raw `git worktree add` to escape governance:
+
+1. The superpowers chain ends in `superpowers:using-git-worktrees` SKILL.md line 96 (`git worktree add "$path" -b "$BRANCH_NAME"`). An LLM that follows `brainstorming` → `writing-plans` → `subagent-driven-development` → `using-git-worktrees` will land on raw `git worktree add` without ever touching peaks-loop's hooks.
+2. Worktrees created outside peaks-loop's governance are invisible to `peaks worktree auth grant|revoke|status` — the L2 hook gate cannot deny or grant them.
+3. Worktree-only artifacts (`.peaks/_runtime/<sessionId>/<role>/requests/<rid>.md` written from inside a worktree) may be lost when the worktree is pruned mid-run. Lesson 1 of [[2026-07-27-rid-016-monorepo-delete-5-subpackages]] documents the failure mode.
+
+## Three-layer design (defence in depth)
+
+| Layer | Mechanism | Strength | Failure mode |
+|---|---|---|---|
+| L1 | Sub-agent system prompt `MUST NOT` on the superpowers chain | Weak — LLM may forget | Prose-only; no enforcement |
+| L2 | PreToolUse hook → `peaks gate enforce` → `evaluateWorktreeAuth` | Medium — sees every tool call but cannot intercept Skill selection | Hook bypassed if user disables `peaks hooks install` |
+| L3 | IDE `permissions.deny: ["UseSkill(superpowers:using-git-worktrees)"]` | Strongest — IDE refuses Skill BEFORE LLM sees it | Requires `peaks hooks install` to have run; bypassed by `--dangerously-skip-permissions` |
+
+The 3 layers stack. L3 is the strongest because it operates BELOW the LLM's tool-call decision: the Skill never appears in the LLM's tool list, so the LLM cannot choose to invoke it.
+
+## L1 — Sub-agent system prompt
+
+The explicit `MUST NOT` block lives in:
+- `peaks-rd/references/rd-sub-agent-dispatch.md`
+- `peaks-qa/...`
+- `peaks-ui/...`
+- `peaks sub-agent dispatch` CLI prompt template
+
+Each prompt template carries the block:
+> MUST NOT follow the superpowers chain (`brainstorming` → `writing-plans` → `subagent-driven-development` → `using-git-worktrees`) for worktree decisions. The chain is reference material, not a workflow. For worktree operations, use `peaks worktree spawn --rid <id> --ttl <duration> --purpose <text>` (or, until that ships, fall back to L2 hook gate + manual `peaks worktree auth grant`).
+
+L1 is the weakest layer because the LLM is the decision-maker and may forget the block under load. L3 exists to catch L1 failures.
+
+## L2 — Hook + lifecycle gate
+
+`peaks hooks install` writes a PreToolUse hook that calls `peaks gate enforce` BEFORE the IDE permission system. For worktree operations:
+
+```bash
+# User authorizes a worktree operation
+peaks worktree auth grant \
+  --rid rid-2026-07-29-worktree-spawn \
+  --reason "RD sub-agent for worktree spawn slice" \
+  --ttl 5m
+# Output: grant token, single-use by default
+```
+
+The token authorizes exactly one `git worktree add` matching the granted scope (path / branch / rid). 5 deny codes:
+- `REQUIRED` — no grant on file, request denied
+- `EXPIRED` — TTL elapsed, re-authorize
+- `REQUEST_MISMATCH` — git args don't match the granted scope
+- `CONSUMED` — token already used (single-use)
+- `FILE_INVALID` — grant file unreadable / corrupted
+
+Implementation: `src/services/hooks/worktree-authorization-gate.ts`. Sediment: `.peaks/memory/2026-07-27-worktree-user-auth-hard-gate.md`.
+
+## L3 — IDE `permissions.deny`
+
+`peaks hooks install` ALSO writes a `permissions.deny` block into the IDE settings.json:
+
+```json
+{
+  "permissions": {
+    "deny": [
+      "UseSkill(superpowers:using-git-worktrees)"
+    ]
+  }
+}
+```
+
+The deny entry is wrapped in the IDE's `UseSkill(...)` envelope (Claude Code's permission-system syntax). Claude Code refuses to invoke any Skill listed there BEFORE the LLM sees the Skill's name in the available-tool list.
+
+Single source of truth: `SUPERPOWERS_DENIED_SKILLS` in `src/services/skills/hooks-settings-service.ts`. To deny a new skill, append its id there and re-run `peaks hooks install`.
+
+Sediment: `.peaks/memory/2026-07-29-worktree-layer3-deny.md`.
+
+## Sub-agent worktree contract (mandatory)
+
+Every sub-agent dispatched by peaks-rd / peaks-qa / peaks-ui / peaks-sc / peaks-txt MUST:
+
+1. **Use `peaks worktree spawn --rid <rid> --ttl <duration> --purpose <text>`** for any worktree operation. Raw `git worktree add` is forbidden.
+2. **Treat the superpowers chain as reference material only.** It MAY inform PRD / RD planning but MUST NOT drive tool selection.
+3. **Write artifacts under `.peaks/_runtime/<sessionId>/<role>/requests/` in the MAIN checkout.** Dispatch with `--project .` (the main workdir), NOT `--project .claude/worktrees/...`. Worktrees are ephemeral; main is durable.
+
+## Operator runbook
+
+- **Manual authorization**: `peaks worktree auth grant|revoke|status` (L2 surface).
+- **Inspect L3 state**: `peaks hooks status` reports `permissionsDenyEntries` (desired) + `permissionsDenyOnDisk` (reality).
+- **Refresh L3 deny block**: `peaks hooks uninstall && peaks hooks install` re-writes the block from `SUPERPOWERS_DENIED_SKILLS`.
+- **Diagnose L3 failures**: the on-disk `permissions.deny` is the source of truth. If a Skill bypasses the deny, the IDE permission system is broken (not peaks-loop's problem).
+
+## Future rid path
+
+| rid | Layer | Goal | Status |
+|---|---|---|---|
+| rid-L1 | L1 | Hard-coded prompt hardening across all dispatch templates | Pending |
+| rid-L2-extended | L2 | `peaks worktree spawn` with lease lifecycle, removing manual `peaks worktree auth grant` loop | Pending |
+| rid-L3-extended | L3 | Append new deny skills to `SUPERPOWERS_DENIED_SKILLS` as needed | Trigger-only |
+
+## Related memories
+
+- [[2026-07-27-worktree-user-auth-hard-gate]] — L2 baseline, shipped earlier.
+- [[2026-07-29-worktree-layer3-deny]] — L3 Minimal Viable rationale.
+- [[2026-07-29-worktree-l2-extended-part1]] — lease foundation + spawn/release CLI.
+- [[2026-07-29-worktree-l2-extended-part2]] — renew/list/gc/status + hook integration + dispatch bridge + lifecycle e2e.
+- [[2026-07-29-worktree-l2-extended-part3]] — auto-release on terminal finalization.
+- [[2026-07-29-worktree-l2-extended-part4]] — observability emitter + v3 schema.
+- [[2026-07-29-worktree-l2-extended-part5]] — leak rate + cross-session aggregation.
+- [[2026-07-29-worktree-l2-extended-part6-9]] — stats summary + v3.1 schema + container contract bridge.
+
+## Lease observability (Parts 4-6)
+
+The L2 ecosystem emits a `lease` category into the existing observability stream (`.peaks/_runtime/<sid>/metrics/slices.jsonl`). Seven `kind` values: `spawn` / `renew` / `release` / `gc` / `autoRelease` / `autoRelease-failed` / `autoRelease-skipped`. Two CLIs read the stream:
+
+- `peaks lease-metrics [--session <sid>] [--project <root>] [--rate] [--all-sessions] [--json]` — per-kind counts + 5-event tail. With `--rate`, adds `estimatedActive` / `estimatedLeaked` / `avgLifetimeMs` / `p99LifetimeMs`. With `--all-sessions`, aggregates across every session under `.peaks/_runtime/`. The two flags are independent; `--rate --all-sessions` is the project-wide leak view.
+- `peaks lease-stats --project <root> --json` — project-wide summary: counts + rate + per-rid (top 20) + per-role + per-isolation breakdown. Always aggregates cross-session. Use this when you want a "status board" rather than the raw event stream.
+
+The `estimated` prefix is intentional: the canonical "alive" set is the on-disk lease files in `.peaks/_runtime/<sid>/worktree-leases/` (queryable via `peaks worktree list`). The aggregation is a metrics-stream view — useful for time-series dashboards, but the FS state is the source of truth for "is this lease still alive right now?".
+
+For dashboards: pipe the JSON envelope into `jq`, paste into a `peaks lease-stats --json` reader, or feed into the `peaks audit metrics` flow. The schema is the same `ObservabilityEvent` v1 used by slice / dispatch / cycle events.
+- [[2026-07-27-rid-016-monorepo-delete-5-subpackages]] — Lesson 1: worktree-only artifact fragility.
+- [[2026-07-24-peaks-code-bridge-002-rootcause]] — peaks-code ↔ superpowers bridge baseline.

package/skills/peaks-code/test-prompts.json

@@ -0,0 +1,12 @@
+[
+  {
+    "id": 1,
+    "prompt": "/peaks-code https://example.feishu.cn/wiki/PRODUCT_DOC_ID 是产品文档。请根据这份产品文档，对 /Users/yuanyuan/Desktop/work/platform-rag-web 这个前端项目进行迭代开发。目前只是前端项目，没有后端 swagger.json，可以先 mock，等后面有真实 swagger.json 数据时再联调替换。允许修改目标仓代码，这是实战测试。",
+    "expected": "peaks-code 选择 full-auto/实战路径，使用本仓 skills 目录下的 peaks-prd/rd/ui/qa/sc/txt，Feishu 走 headed gstack/browse/dist/browse 可见浏览器登录 handoff，先校验 https 和批准的 Feishu/Lark tenant 域名，生成 .peaks/_runtime/<session-id>/prd|rd|qa|txt 脱敏中间产物；RD 完成实现、单测、CR、安全检查、dry-run 后必须进入 QA；前端 RD 自测和 QA 验证都必须使用 headed gstack/browse/dist/browse 有头浏览器端到端验证并确认可见浏览器；QA 完成 API/mock 验证、浏览器 E2E、报告、安全和性能检查。若 QA 报告存在失败、阻塞、缺失或未验证项，必须回到 RD 修复并再次 QA，直到 QA 全部通过或生成 blocked handoff；full-auto 可用 Claude Code goal 维持 RD fixes until QA passes all acceptance items 的控制目标；不得保留登录 URL、cookie、headers、tokens、storage state、browser traces、raw browser/network state 或 PII/SSO/MFA 截图/日志。"
+  },
+  {
+    "id": 2,
+    "prompt": "/peaks-code 对 /Users/yuanyuan/Desktop/work/platform-rag-web 做前端需求迭代，需求来自需要登录的 Feishu 产品文档；如果浏览器登录没打开或文档仍不可读，需要阻塞并告诉我具体要做什么，不要用 unauthenticated fetch，也不要绕过 SSO/MFA。",
+    "expected": "流程必须验证可见浏览器 handoff，不能降级到 fetch；登录或访问失败时生成 blocked PRD/TXT 胶囊，只记录脱敏状态和下一步，不保存登录 URL、cookie、headers、tokens、storage state、浏览器 traces，或包含 PII/SSO/MFA 内容的截图和日志。"
+  }
+]

package/skills/peaks-doctor/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: peaks-doctor
 description: Orchestrate peaks-loop's L3 doctor (peaks audit + peaks doctor + peaks openspec from-doctor) for project health. Use when the user asks for a project health check, doctor report, audit, or wants to convert doctor findings into OpenSpec change records. Coordinates the L2 audit framework + the L3.2 doctor + the L3.3 from-doctor proposal generator. Triggers on `/peaks-doctor`, "peaks doctor", "项目健康", "doctor report", "health check", "check the project", "audit my repo".
-internal: true
+visibility: internal
 ---
 ---
 

package/skills/peaks-ide/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: peaks-ide
 description: Orchestrate peaks-loop's IDE-aware behavior (hooks + statusline + handle) for a user's specific IDE. Detects the current state (which IDE the user is on, what peaks has already installed), plans the install / switch / status / uninstall actions, and invokes the existing peaks CLI primitives. Triggers on `/peaks-ide`, "set up peaks for my IDE", "switch peaks to Trae", "what did peaks install", "uninstall peaks hooks". Sits between the user and `peaks hooks install` / `peaks statusline install` / `peaks hook handle` — those are the CLI primitives; this skill is the user-facing surface.
-internal: true
+visibility: internal
 ---
 ---
 

package/skills/peaks-solo/SKILL.md

@@ -163,6 +163,27 @@ multiSelect: false
 | 用户主动说"沉淀这个" | **立即问** | `user_explicit` |
 | LLM 识别 ≥ 2 个 reuse signals | **附带 4-section brief 问** | `llm_suggested` |
 
+### 5.3 4-level precedence for keyword triage (rid-020b)
+
+When the user's NL description contains a 24h-mode keyword (`24h` / `通宵` / `夜跑` / `通宵跑` / `夜机` / `不计成本` / `不停机`) AND a code-domain keyword, apply the 4-level precedence table to pick the leaf:
+
+| # | Precedence | If matched | Routes to |
+|---|---|---|---|
+| 1 | peaks-* skill name (highest) | user names `/peaks-code` or `/peaks-content` or `/peaks-doctor` explicitly | named leaf, regardless of 24h keyword |
+| 2 | keyword match | `24h` / `通宵` / `夜跑` etc. present | add `code-domain-evidence` column to triage table; default to `/peaks-code` for code work |
+| 3 | code-domain evidence | change-id / src path / `peaks-code` mention in NL | `/peaks-code` (with `--24h` flag passed via `peaks code run --24h`) |
+| 4 | default (lowest) | none of the above | peaks-solo self-planning fallback (§4) |
+
+### 5.4 24h keyword sub-section (special cases)
+
+Three explicit edge cases — peaks-solo still routes to the named leaf, but the LLM MUST add the 24h flag at the right invocation:
+
+| 24h keyword NL snippet | Default leaf | 24h flag injection |
+|---|---|---|
+| `24h 跑 dispatcher` / `24h triage` | /peaks-solo (self) | — (peaks-solo has no 24h flag; record sediment) |
+| `24h 内容` / `24h 公众号` / `24h 博客` | /peaks-content | peaks-solo surfaces `AskUserQuestion` asking the user to add the 24h flag at the leaf dispatch |
+| `24h 体检` / `24h doctor` / `24h health` | /peaks-doctor | peaks-solo surfaces `AskUserQuestion` asking the user to add the 24h flag at the leaf dispatch |
+
 ---
 
 ## 6. Out of scope (HC-11 锁死)

package/skills/peaks-solo/references/triage-decision-table.md

@@ -16,6 +16,7 @@
 | 跑测试 / test / vitest / unit test | → | /peaks-test |
 | 恢复 / resume / 继续 / 接上次 | → | /peaks-resume |
 | 不知道 / 随便 / 帮我 / 你帮我决定 / 我想做点东西 | → | peaks-solo triage (this skill) |
+| 24h / 通宵 / 通宵跑 / 夜跑 / 夜机 / 不计成本 / 不停机 (code-domain-evidence) | → | /peaks-code (with `--24h` via `peaks code run --24h`; see SKILL.md §5.3) |
 | 切片 / slice / 拆解任务 / 大任务拆小 | → | /peaks-slice-decompose |
 | 想法 / brain-storm / 创意 / 探索思路 | → | /peaks-ide |
 | 最终审查 / final review / 验收 / 4-dim evidence | → | /peaks-final-review |
@@ -29,3 +30,8 @@
 3. **(e) "都不对" → 自规划兜底。** 走 SKILL.md §4。
 4. **空表条目:** "研究 / 信息查询 / GitHub trending / top 10" 等一次性信息查询,**当前 skill 池没有专属 leaf**,peaks-solo 直接走自规划(deep-search / WebSearch)。
 5. **本表是 opportunistic 表,**不是 exhaustive 表。S2/S3 后续可在 dogfood 中发现新关键词补行。
+6. **24h keyword row (rid-020b):** when the NL contains a 24h keyword AND code-domain evidence (e.g. `peaks-code` mentioned, a `src/` change-id, or `peaks code run`), add a `code-domain-evidence` column to the AskUserQuestion and pre-select `/peaks-code` with the `--24h` flag surfaced as a follow-up. The 4-level precedence in SKILL.md §5.3 wins on ties.
+7. **Special case rows (rid-020b):**
+   - `24h dispatcher` / `24h triage` → peaks-solo (self); record sediment after.
+   - `24h 内容` / `24h 公众号` / `24h 博客` → /peaks-content; surface AskUserQuestion asking the user to add the 24h flag.
+   - `24h 体检` / `24h doctor` → /peaks-doctor; surface AskUserQuestion asking the user to add the 24h flag.

package/skills/peaks-sop/test-prompts.json

@@ -0,0 +1,17 @@
+[
+  {
+    "id": 1,
+    "prompt": "我想要一个发布流程：先写好 CHANGELOG，再跑通测试，最后才能发布。帮我做成一个 SOP，我不想自己写 JSON。",
+    "expected": "peaks-sop 访谈出 phases（如 draft/review/ship）和门禁，调用 peaks sop init --apply 生成骨架，把门禁写进 sop.json（changelog 用 file-exists、tests 用 command 并提示需要 --allow-commands），跑 peaks sop lint 到 ok:true，用 peaks sop check 验证每个门禁，再 peaks sop register 注册。全程不让用户手写 JSON。"
+  },
+  {
+    "id": 2,
+    "prompt": "我的 SOP lint 报错说门禁绑定了未知阶段，帮我修。",
+    "expected": "peaks-sop 读 lint findings 定位违例门禁，在 sop.json 里把门禁的 phase 改成已声明的阶段（或补声明该阶段），重新 peaks sop lint 直到 ok:true，不直接 register 未通过 lint 的 SOP。"
+  },
+  {
+    "id": 3,
+    "prompt": "这个发布 SOP 真的能在测试没过时把我拦住吗？我想先确认再用。",
+    "expected": "peaks-sop 用 peaks sop check 在好/坏两种状态下验证门禁返回 pass/fail，再用 peaks sop advance --to ship --dry-run 干跑确认门禁未通过时阻断推进且不落盘，向用户展示阻断行为后再交付。"
+  }
+]

package/examples/video-demo/README.md

@@ -1,142 +0,0 @@
-# peaks-loop video demo (v5)
-
-Remotion-based ~27-second brand walk-through for peaks-loop 4.x. Mirrors the multi-domain + slash-command narrative of the top-level README.
-
-## Composition
-
-- Two compositions, one per locale:
-  - `peaks-loop-demo`     — Chinese primary
-  - `peaks-loop-demo-en`  — English primary
-- Same frame budget for both; dimensions 1920×1080 @ 30fps.
-- Scene timeline (frozen in `src/Root.tsx`, v5 narrative):
-
-  | # | Scene | Slug | Frames |
-  | --- | --- | --- | --- |
-  | 1 | `IntroScene` | intro | 80 |
-  | 2 | `RecordingScene` | install | 100 |
-  | 3 | `RecordingScene` | first-slash | 100 |
-  | 4 | `RecordingScene` | domains | 100 |
-  | 5 | `RecordingScene` | sediment | 100 |
-  | 6 | `CreditScene` | credit | 130 |
-  | 7 | `ClosingScene` | closing | 110 |
-
-  Plus 12-frame overlap transitions between every scene.
-
-## Visual style
-
-- Background: `#0f172a` (slate-900) via `bg-brand-bg` + `BrandBackground` (drifting indigo halo + diagonal sweep line + bottom green glow).
-- Accent: `#6366f1` (indigo-500) · Stable: `#22c55e` (green-500) · Warn: `#f59e0b` (amber-500).
-- Typography: `font-mono` for kickers, captions, IDE/terminal HUD; `font-sans` for headlines.
-- Three caption styles (`CaptionOverlay`):
-  - **subtitle** — black semi-transparent bar at the bottom.
-  - **callout** — indigo gradient emphasis box, slides in from the left.
-  - **annotation** — amber speech-bubble, drops in from the top-right.
-
-## Scene components
-
-```
-src/scenes/
-├── IntroScene.tsx       — hero title + tagline (kicker / brand / tagline)
-├── RecordingScene.tsx   — the workhorse: 4 instances, each plays an HUD + captions + cursor
-├── HudWindow.tsx        — stylised terminal/IDE panel that types out lines one at a time
-├── CaptionOverlay.tsx   — 3-style caption layer (subtitle / callout / annotation)
-├── CursorHighlight.tsx  — macOS-style cursor with smooth waypoint interpolation + click ripple
-├── BrandBackground.tsx  — shared brand background (halo + sweep + bottom glow)
-├── CreditScene.tsx      — tribute + recommended stack
-└── ClosingScene.tsx     — install chip + 5,439 tests stats
-```
-
-## Copy registry
-
-All on-screen text lives in `src/copy.ts`. Both locales are strict: no hardcoded
-CN/EN inline in scene components. The `recordings[]` array holds 4 scene slugs
-(install / first-slash / domains / sediment), each with its own captions,
-cursor choreography, and HUD payload.
-
-## v5 recording pipeline (Playwright + Remotion)
-
-The `RecordingScene` slots are designed to host **real recorded footage**
-when you have a Chromium-equipped machine. The HUD window is the stand-in
-for that footage until the recording pass runs.
-
-### 1. Install the recording side
-
-```bash
-cd examples/video-demo
-pnpm add -D playwright
-pnpm exec playwright install chromium
-```
-
-### 2. Author a scenario JSON
-
-Write `recordings/<slug>.json` per scene using the
-`web-demo-video_skills` schema (`goto` / `click` / `fill` / `wait` / `screenshot`).
-For our 4 slugs, the natural targets are:
-
-| slug | URL / action |
-| --- | --- |
-| `install` | terminal typing `npm i -g peaks-loop` |
-| `first-slash` | Claude Code chat with `/peaks-code ...` |
-| `domains` | same chat, scrolling through the 5 `/peaks-*` slash commands |
-| `sediment` | same chat, completing a run + asking to sediment |
-
-### 3. Record per scene
-
-```bash
-pnpm exec tsx scripts/record-scenes.ts
-# outputs recordings/<slug>.webm  +  recordings/<slug>.cursor.json
-```
-
-### 4. Swap the HUD for the recorded clip
-
-In `src/scenes/RecordingScene.tsx`, replace the `<HudWindow>` block with a
-`<Video src={...}>` from the matching `recordings/<slug>.webm`. Drop the
-`HudWindow` import when no scene needs it any more.
-
-### 5. Render
-
-```bash
-pnpm install
-npx remotion render peaks-loop-demo out/peaks-loop-demo.mp4
-npx remotion render peaks-loop-demo-en out/peaks-loop-demo-en.mp4
-
-mkdir -p preview
-cp out/peaks-loop-demo.mp4    preview/peaks-loop-demo.mp4
-cp out/peaks-loop-demo-en.mp4 preview/peaks-loop-demo-en.mp4
-```
-
-Fallback (no Chromium available — renders the HUD stand-in only):
-
-```bash
-npx remotion still peaks-loop-demo out/still.png --frame=400
-```
-
-## Preview during dev
-
-```bash
-npx remotion studio   # http://localhost:3000
-```
-
-## Why this layout
-
-The 4 recording slugs follow the **black-box mental model** the README now
-uses: install → first slash command → tour the 5 domains → sediment. Each
-scene is **self-contained** (own HUD payload, captions, cursor ticks),
-so you can rearrange them, drop one, or A/B test a new one without
-touching the others. The 12-frame overlap transitions cross-fade between
-scenes so the result feels like one continuous take rather than a slide
-deck.
-
-## Adding a new scene
-
-1. Add a new entry to `recordings` in `src/copy.ts` (both `zh` and `en`):
-   - `slug` (kebab-case, used for the corner chip + recording filename)
-   - `captions` (array of `{ text, style }` triples)
-   - `cursor` (array of waypoints; empty = no cursor overlay)
-   - `hud` (terminal / ide, with the lines you want typed out)
-2. Add an extra frame budget to `src/Root.tsx` if you want a longer hold.
-3. Typecheck: `pnpm exec tsc --noEmit`.
-4. Re-render.
-
-If you have a recording, drop a `<Video>` into `RecordingScene.tsx` and
-point it at the new `recordings/<slug>.webm`.