peaks-loop 4.0.0-beta.3 → 4.0.0-beta.30

package/skills/peaks-code/SKILL.md

@@ -18,7 +18,7 @@ The boundary is closed under this slice:
 
 ## Single-scope-axis naming convention (2.7.1)
 
-The `.peaks/` workspace has a **single scope axis** (session-id) plus a nested **sub-agent axis** under `.peaks/_sub_agents/<sessionId>/...`. Use `<sessionId>` (NEVER bare `<sid>`). Reviewable artifacts live at `.peaks/_runtime/<sessionId>/<role>/...` (gitignored); the change-id is an optional filename slug and does NOT route filesystem writes. CLI: session-id → `peaks session *`; sub-agent → `peaks sub-agent *`. OpenSpec's `openspec/changes/<change-id>/` vocabulary (L4) is preserved. Test `skills-skill-md-naming.test.ts` enforces (a) zero bare `<sid>`, (b) axis labels, (c) this callout.
+The `.peaks/` workspace has a **single scope axis** (session-id) plus a nested **sub-agent axis** under `.peaks/_sub_agents/<sessionId>/...`. Use `<sessionId>` (NEVER bare `<sid>`). Reviewable artifacts live at `.peaks/_runtime/<sessionId>/<role>/...` (gitignored); the change-id is an optional filename slug and does NOT route filesystem writes. CLI: session-id → `peaks session *`; sub-agent → `peaks sub-agent *`. Test `skills-skill-md-naming.test.ts` enforces (a) zero bare `<sid>`, (b) axis labels, (c) this callout.
 
 ## Karpathy guidance (Slice 1/6 — karpathy prompt-injection-lift)
 
@@ -46,118 +46,64 @@ This skill is the **primary surface**. The `peaks <cmd>` CLI is **auxiliary**
 
 ## Peaks-Loop Startup sequence (MANDATORY — execute in order)
 
-### Peaks-Loop Slice 011 — workspace consolidate + session checkpoint/resume
+Full content extracted to **`references/startup-sequence.md`** (Steps 0 / 0.5-0.87 / 1 / 2 / 2.3 / 2.5 / N / N+1 / N+2 + sub-agent sharing + boundaries). Read that file in full at session start; the sequence is MANDATORY.
 
-`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 0.8 — Job-shape detection (BLOCKING on LLM judgement)
 
-### Peaks-Loop Step 0.75: Resume from checkpoint (BLOCKING on same-day re-invocation)
+> **BLOCKING on LLM judgement.** Before Step 1 mode selection, Code MUST surface a Job-shape decision via `peaks code detect-job` and persist it to `.peaks/_runtime/<sessionId>/job-shape.json`. The CLI is a recorder + gate — the LLM makes the judgement; downstream steps call `read-job-shape` and refuse if missing. See `references/step-0-8-gate.md` for the full contract (LLM criteria, action sequence, hard rule red-line #10). Failure mode: skipping `peaks code detect-job` blocks the workflow at the next `read-job-shape` call (`JOB_SHAPE_NOT_DECIDED`).
 
-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`.
+**v3.1.2 mechanical gates** (the recorder-only design was bypassed twice under load; the gate must be un-bypassable):
 
-### Peaks-Loop Step N: Periodic checkpoint (auto-fire, no user action)
+1. **PreToolUse hook — `peaks code gate-step-08`.** Installed by `peaks workspace init` on the `Bash` matcher; checks `job-shape.json` presence + fail-closed backup regex. If `job-shape.json` exists AND `progress.json` exists, surfaces `Next: slice #N of M (<currentSlice>)` so the LLM cannot wake up cold.
+2. **Size-fear ban — `peaks code emit-handoff`.** Refuses to emit a final handoff while `remaining > 0` under Job mode. Pass `--force-under-job` only with explicit user approval.
+3. **On-disk slice progress — `peaks job progress`.** `peaks job checkpoint --state done` writes `progress.json`. `peaks job progress --job-id <jid> [--allow-missing]` is the canonical reader; `peaks code gate-step-08` also surfaces it on every Bash call.
+4. **Forced auto-compact — `--enforce-job-mode`.** `peaks code context-now --enforce-job-mode` returns `action: 'auto-compact-now'` at ≥ 0.85. **Job mode at ≥ 0.85 is MANDATORY auto-compact** — Code MUST call `peaks session auto-compact --execute` without confirmation.
 
-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).
+**Step 0.7 resume rule (read-FIRST):** on resume, `peaks code gate-step-08` reads `progress.json` first and surfaces `Next: slice #N of M (<currentSlice>)` so the orchestrator picks up at the right slice without re-reading the artifact tree.
 
-### Peaks-Loop Step 0.5: OpenSpec first-run opt-in (conditional)
+### Peaks-Loop Step N+2: Auto-compact at the warning line (v2.13.0 zero-pause contract)
 
-Run when `openspec/` is absent and `.peaks/.peaks-openspec-opt-in.json` is missing. Asks once and persists the decision.
+> **Zero-pause contract.** When context usage crosses the pre-compact threshold, peaks-loop **automatically fires `peaks session auto-compact --execute` — the LLM does NOT prompt the user to run `/compact` manually**. This is the single biggest UX regression vector in the system: a stale prose that says "ask the user to compact" will silently send the LLM into "wait for human" mode and stall the entire workflow. The v2.13.0 contract makes auto-compact a system responsibility, not a user action.
 
-→ see `references/openspec-workflow.md` for the full opt-in flow + lifecycle.
+**Thresholds (v2.13.0, replacing legacy 50/75/90):**
 
-### Peaks-Loop Step 0.6: Audit + Goal (NEW)
+| ratio zone | zone name | action |
+|---|---|---|
+| `< 0.85` | normal | skip — LLM keeps working, no action |
+| `0.85 ≤ ratio < 0.95` | **pre-compact zone** | `peaks session auto-compact --execute` fires **automatically** (deferred only if an in-flight sub-agent batch is still running; fires the moment the batch lands). The LLM does not prompt the user. |
+| `ratio ≥ 0.95` | **red-line (Karpathy §4 compact-red-line exception)** | synchronous gate — `peaks session auto-compact --execute` is invoked immediately; `peaks code context-now` returns `action: 'red-line'` and refuses to advance until ratio drops below 0.85. **Karpathy §4 treats this as an automatic exception** — the LLM cannot opt out. |
 
-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`.
+**Probe primitive (single source of truth):** `peaks code context-now --json`. Do NOT use `peaks context check --prompt-size` (deprecated, pre-v2.13.0 surface; will silently under-report ratio). The CLI returns `{ ratio, action: 'ok' | 'soft-warn' | 'auto-compact-now' | 'red-line' }` — Code reads `action` and dispatches `peaks session auto-compact --execute` on `auto-compact-now` or `red-line` without any user confirmation.
 
-### Peaks-Loop Step N+1: Final Review (NEW)
+**Enforcement layers (defense in depth — no single layer is the gate):**
 
-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`.
+1. `src/services/code/auto-compact-orchestrator.ts` — `evaluateAutoCompactDecision` default-returns `shouldCompact: true` for both `pre-compact` and `red-line` zones. The only deferral is `inFlightBatch.hasInFlightBatch` (D6.e); no LLM / human approval branch.
+2. `--enforce-job-mode` (v3.1.2) — Job mode elevates ≥0.85 to MANDATORY regardless of in-flight batch (see Step 0.8 §4 above).
+3. `peaks code gate-step-08` (PreToolUse hook) — surfaces `auto-compact-now` on every Bash call when ratio is in the zone, so the LLM cannot wake up cold and forget.
+4. The Karpathy §4 exception: `peaks session auto-compact --execute` is fired *by the orchestrator*, not by the user running `/compact`. If you find yourself about to write prose that says "ask the user to compact" or "prompt the user to run `/compact`", STOP — that is the regression.
 
-### Peaks-Loop Step N+2: Main-session context monitor (D6 + slice 2026-07-02)
+**Anti-pattern (regression marker — DO NOT introduce):** any of these strings in skills/* or comments signals the zero-pause contract has been broken:
 
-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** (not advisory): Code MUST call `peaks session auto-compact --execute --project <repo>` autonomously. Pass `--enforce-job-mode` to force the MANDATORY semantics in single-rid mode too (`job-shape.json` auto-enables it). **≥ 0.95** red line: next Bash/Task call fires `peaks session auto-compact-hook` (PreToolUse), in-band spawning `claude --compact`. **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).
+- "ask the user to compact"
+- "prompt the user to run `/compact`"
+- "the user should run `peaks session auto-compact` manually"
+- "the user is responsible for context management"
+- legacy pre-v2.13.0 thresholds (mid / seventy-five / ninety percent tiers; current is 0.85 / 0.95)
 
-### Peaks-Loop Step 0: Anchor the workflow (MANDATORY FIRST ACTIONS — no bail-out)
+If the prose audit (`peaks audit red-lines`) flags any of the above, the slice is **blocked** until the prose is rewritten.
 
-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.
+## CLI Drift Index (sediment 2026-07-09)
 
-→ see `references/anchoring-and-session-info.md` for the full Step 0 contract + session-conflict resolution.
+> **Reading guide:** Verified against peaks-loop 4.0.0-beta.6. Each drift below is annotated inline at the relevant step with a `> CLI reality check`. On `error: unknown option ...`, **read the inline reality check first** before guessing.
 
-#### Step 0 bypass — Claude Code [Fact-Forcing Gate] (slice 2.0.1-bug3)
+| Drift ID | Step | Symptom | Fix | Inline location |
+|---|---|---|---|---|
+| **D-001** | 0.8 | `peaks code detect-job --is-job ...` rejected with `error: unknown option '--is-job'` | Use `peaks job init --job-id <jid> --slice-list <list> --main-loop-strategy <single\|rotating>` | §Step 0.8 first paragraph |
+| **D-002** | 2.5 | `peaks session title --session-id <sid> ...` rejected with `error: unknown option '--session-id'` (this is the bare `<sid>` anti-pattern) | sid is positional: `peaks session title <sessionId> "<title>" --json` | §Step 2.5 |
+| **D-003** | 0.8 | `JOB_SHAPE_NOT_DECIDED` exception expected but never thrown | Current behavior is `peaks job status` reports `done: 0` passively — treat as recoverable miss, not hard error | §Step 0.8 third paragraph |
+| **D-010** | 11c | `peaks memory extract` returns `extractedCount: 0` despite `<!-- peaks-memory:start -->` existing | Block requires YAML frontmatter (`title:` + `kind:` + `---`) + closing `<!-- peaks-memory:end -->`. Bare `peaks-memory:start` is parsed silently but produces no writes | §Step 11c + 11d |
 
-`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 it FIRST and surface `Next: slice #N of M (<currentSlice>)`. `peaks job progress --job-id <jid> [--allow-missing]` is the canonical reader; the same JSON is surfaced by `peaks code gate-step-08` on every Bash call.
-
-**v2.11.0 D7 override:** if user just `/compact`ed, run `peaks code post-compact-detect --project <repo> --json` FIRST; `shouldAutoResume: true` skips AskUserQuestion (D7.b). Log to `.peaks/_runtime/<sessionId>/txt/auto-decisions.md`.
-
-→ see `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`. Persist to `.peaks/preferences.json`. → `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)
-
-The CLI is a **recorder + gate** for job-shape (the LLM judges). LLM calls `peaks code detect-job --is-job <bool> --rationale <text> --suggested-job-id <jid> --suggested-strategy <single|rotating> --confidence <high|medium|low>`, which writes `.peaks/_runtime/<sessionId>/job-shape.json`. Downstream steps call `read-job-shape` and refuse if missing.
-
-**Hard rule (v3.1.1 red-line #10):** LLM MUST NOT skip `peaks code detect-job`. If `read-job-shape` throws `JOB_SHAPE_NOT_DECIDED`, record a decision before proceeding.
-
-**v3.1.2 mechanical gates** (recorder-only was bypassed twice):
-
-- **PreToolUse hook** — `peaks workspace init` installs the `Bash` matcher `peaks code gate-step-08 --project .`. Exits 0 when `job-shape.json` exists (also emits `Next: slice #N+1 of M` from `progress.json`); exits 2 BLOCKED when missing AND prompt matches a fail-closed backup regex; otherwise 0 (allow).
-- **Size-fear ban** — `peaks code emit-handoff` refuses a final handoff while `remaining > 0` under Job mode.
-- **Forced auto-compact** — `peaks code context-now --enforce-job-mode` (auto-enabled on `isJob=true`) returns `auto-compact-now` at ≥ 0.85; **In Job mode ≥ 0.85 is MANDATORY auto-compact** (not advisory).
-- **On-disk slice progress** — `peaks job checkpoint --state done` writes `progress.json`; `peaks job progress --job-id <jid> [--allow-missing]` is the canonical reader.
-
-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 with TXT handoff.
-
-### Peaks-Loop Step 0.86 — main session rotation
-
-Active in rotating mode; fires every `rotateEvery` slices or on-demand via `peaks job rotate-now`. Sequence: cycle-summary → checkpoint rotate-marker → `peaks session rotate` → next turn resumes 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` with the active sid. Skip if a title is already set. → `references/skill-presence-and-title.md` (same file as Step 2).
-
-## 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 through artifacts, coordinate project memory extraction, request user confirmation at risk/commit boundaries. Peaks-Loop Code must NOT silently install hooks / create agents / enable MCP / modify Claude settings / create GitHub repos / bypass role-skill artifacts. → `references/boundaries.md`.
+> **Sediment lesson (master record):** `.peaks/memory/peaks-code-runbook-4-0-0-beta-6-skill-md-cli-d-001-d-002-d-003-d-010.md`
 
 ## Peaks-Loop GStack integration
 
@@ -169,7 +115,9 @@ Step 0 creates the workspace; all intermediate artifacts land under `.peaks/_run
 
 ## Peaks-Loop Pre-RD project scan checklist (MANDATORY)
 
-Before handing off to `peaks-rd`, scan the project and record findings to `.peaks/_runtime/<sessionId>/rd/project-scan.md`. → `references/project-scan-checklist.md`.
+Before handing off to `peaks-rd`, scan the project and record findings to `.peaks/project-scan/project-scan.md` (project-level, git-tracked; sibling of `.peaks/PROJECT.md`). Slice 2026-07-15-project-scan-bootstrap (G1 + G2): `peaks workspace init` and `peaks project context` both bootstrap this file automatically; only re-run the manual flow when the scan output is genuinely out of date. → `references/project-scan-checklist.md`.
+
+> **Slice 2026-07-15-project-scan-bootstrap G4b / AC9:** after `peaks workspace init`, the consumer project's `.peaks/project-scan/` directory carries 5 files: `project-scan.md` (generated by the scan logic) + `business-knowledge.md` + `security-template.md` + `perf-template.md` + `audit-output-schema.md` (the 4 audit/business templates bundled at `src/services/workspace/templates/project-scan/*.md`, copied on init; idempotent — existing files preserved unless `--force-project-scan-templates`). RD / UI / TXT / QA read from this tree; nothing else.
 
 ## Peaks-Loop Frontend-only development mode
 
@@ -177,15 +125,13 @@ When the project has no live backend (no swagger.json, no API server), Code must
 
 ## Peaks-Loop Request type classification + Workflow order + Transition verification gates
 
-The contract for the 6-type classification table, the 11-step workflow order, and the 7 transition verification gates (A through G with their `ls` / `grep` shell snippets) lives in `references/workflow-gates-and-types.md`. The peaks-code narrative references those gate numbers (Gate A through Gate G) — keep both files in lockstep when adding or renaming a gate.
+The 6-type table + 11-step order + 7 transition gates (A-G) live in `references/workflow-gates-and-types.md`. peaks-code narrative references Gate A-G — keep both files in lockstep.
 
 ## Peaks-Loop Default sub-agent fan-out (≥ 2 leaves/topological level → `--from-dag`)
 
 > **Slice 5:** when the slice DAG has ≥ 2 leaves at one topological level, dispatch via `peaks sub-agent dispatch --from-dag <dag-file>` (wall-time ≈ max, not sum).
 
-Write DAG → `.peaks/_runtime/<sessionId>/sc/slice-dag.json`, run `peaks sub-agent dispatch --from-dag <dag-file> --batch-id <id>` once; orchestrator emits N parallel `buildToolCall` (`dispatchCount === N`).
-
-> **2026-06-28 校准:** 主路径 = 唯一蜂群;config/docs/chore 跳过不打断。详见 `.peaks/memory/peaks-loop-24h-ai-programmer-positioning.md`。
+Write DAG → `.peaks/_runtime/<sessionId>/sc/slice-dag.json`, run `peaks sub-agent dispatch --from-dag <dag-file> --batch-id <id>` once; orchestrator emits N parallel `buildToolCall` (`dispatchCount === N`). 主路径 = 唯一蜂群;config/docs/chore 跳过不打断。详见 `.peaks/memory/peaks-loop-24h-ai-programmer-positioning.md`。
 
 ### Hard constraint: fan-out is mandatory (slice 2026-06-24-audit-5th-p2)
 
@@ -229,23 +175,7 @@ After final validation, refresh project-local standards via `peaks standards ini
 
 ## Peaks-Loop Step 11: Memory sediment (BLOCKING on workflow complete)
 
-> **Hard rule.** Code MUST NOT declare a workflow complete until Step 11 has produced ≥ 1 file in `.peaks/memory/` (the durable, LLM-authored memory store) OR the user has explicitly approved a no-sediment outcome via AskUserQuestion. Applies to **all modes** including `assisted` and `strict` — `assisted` previously skipped Step 10 because there was no `[CONFIRM]` gate; Step 11 fixes that.
-
-### Substeps (BLOCKING)
-
-**11a — Gate A (txt/ inventory):** `find .peaks/_runtime/<sessionId>/txt/ -type f -name '*.md' | head`. If **0 files** → STOP. Dispatch `peaks-txt` first to write `handoff.md`, then return to 11c.
-
-**11b — Gate B (memory block embed scan):** `grep -c 'peaks-memory:start' .peaks/_runtime/<sessionId>/txt/handoff.md || true`. If **0 AND this session surfaced a stable project fact** (decision / convention / approved refactor / hard rule), STOP and tell peaks-txt to embed at least one `<!-- peaks-memory:start -->` block first.
-
-**11c — Canonical extract** (the only CLI that writes `.peaks/memory/`): `peaks memory extract --project <repo> --artifact .peaks/_runtime/<sessionId>/txt/handoff.md --apply --json`. `--apply` is REQUIRED (without it the command only previews — no files land).
-
-**11d — Gate C (zero-write outcome):** If `extractedCount === 0` after 11c, fire AskUserQuestion:
-
-> "本次 code 未沉淀任何 `.peaks/memory` 文件。可选: (a) 回去在 handoff.md 嵌入至少 1 个 `peaks-memory:start` block 后重试; (b) 显式接受 no-sediment 并记录为 lesson; (c) 取消完成。"
-
-Default option = (a). Code MUST NOT silently accept (b) without user pick. Full bash + flow at `references/runbook.md` §Step 11.
-
-**Why this step exists:** audit 2026-07-03 confirmed 2 consecutive sessions produced zero `.peaks/memory/` files despite completing RD + QA + handoff artifacts; `assisted` mode silently skipped runbook Step 10 (no STOP condition). **Why `peaks memory extract` (not `peaks project memories:extract`):** the artifact-scoped extract is canonical; the batch-scoped sibling is for non-handoff flows. Always use `peaks memory extract --apply`.
+> **Hard rule.** Code MUST NOT declare a workflow complete until Step 11 has produced ≥ 1 file in `.peaks/memory/` OR the user has explicitly approved a no-sediment outcome via AskUserQuestion. Canonical CLI: `peaks memory extract --project <repo> --artifact .peaks/_runtime/<sessionId>/txt/handoff.md --apply --json` (the artifact-scoped extract; the batch-scoped sibling `peaks project memories:extract` is for non-handoff flows only). Substeps 11a/11b/11c/11d (Gate A/B/C), D-010 fix root cause check → `references/step-11-memory-sediment.md` + `references/runbook.md` §Step 11.
 
 ## Peaks-Loop External references and lifecycle
 
@@ -259,9 +189,9 @@ Default option = (a). Code MUST NOT silently accept (b) without user pick. Full
 
 Main LLM reducer sees metadata-only view (~200 chars/sub-agent); on-demand `Read` for full content. Threshold table: 50% soft warn, 75% `CONTEXT_NEAR_LIMIT`, 80% hard reject (CLI + hook double-guard). → `references/context-governance.md`.
 
-## Sub-agent cross-batch signal — G8.4 share / shared-read / await (slice 2026-06-23-audit-3rd)
+## Sub-agent cross-batch signal — G8.4 share / shared-read / await
 
-Three CLI primitives let sibling sub-agents coordinate: `peaks sub-agent share / shared-read / await` (last-write-wins, ≤ 1KB warn / ≥ 64KB reject). Channel file gitignored under `.peaks/_sub_agents/<sessionId>/shared/`.
+Three CLI primitives: `peaks sub-agent share / shared-read / await` (last-write-wins, ≤ 1KB warn / ≥ 64KB reject). Channel gitignored under `.peaks/_sub_agents/<sessionId>/shared/`.
 
 ## References
 
@@ -288,7 +218,6 @@ Index of every `references/` file. Read on demand.
 | `references/local-artifact-workspace.md` | Workspace tree + root-prohibition. |
 | `references/micro-cycle.md` | RD micro-cycle + repair loop. |
 | `references/mode-selection.md` | Step 1 mode + `--mode`. |
-| `references/openspec-workflow.md` | Step 0.5 OpenSpec. |
 | `references/playwright-mcp-multi-terminal.md` | Multi-terminal Playwright MCP. |
 | `references/project-memory-loading.md` | Step 2.3 memories. |
 | `references/project-scan-checklist.md` | Pre-RD scan + template. |

package/skills/peaks-code/references/external-references.md

@@ -10,8 +10,6 @@ Do not run upstream installer flows, mutate agent settings, or commit `.codegrap
 
 **External skills**: All external skill references (`mattpocock/skills`, `awesome-design-md`, `taste-skill`, `shadcn/ui`, `Chrome DevTools MCP`, `Figma Context MCP`, `Context7`, etc.) follow the three-stage pattern: capability discovery via `peaks capabilities` before naming, references only (no execute/install/persist), Peaks-Loop CLI for all side effects. Do not execute upstream installers, do not install upstream resources, do not persist sensitive examples — Peaks-Loop gates remain authoritative. External skills inform, they do not approve.
 
-**OpenSpec lifecycle**: `render → validate → show → to-rd → validate → archive`. Code's default runbook handles the exit gate (validate → archive after QA pass). Entry-gate validation (to-rd before slicing) is available when `openspec/` exists pre-workflow; Code delegates it to `peaks-rd` during implementation.
-
 **MCP lifecycle**: `list → plan → apply --yes → call → rollback`. `apply` backs up settings and refuses non-peaks entries unless `--claim` is passed.
 
-Detailed rules: `references/external-skill-invocation.md`, `references/openspec-mcp-workflow.md`, `references/workflow.md`, `references/existing-system-extraction.md`. For an informational mapping of peaks artefact paths to the A2A (Agent2Agent) protocol's Task / Artifact / Part / Message / AgentCard vocabulary (no A2A implementation, just a shared naming layer), see `references/a2a-artifact-mapping.md`.
+Detailed rules: `references/external-skill-invocation.md`, `references/workflow.md`, `references/existing-system-extraction.md`. For an informational mapping of peaks artefact paths to the A2A (Agent2Agent) protocol's Task / Artifact / Part / Message / AgentCard vocabulary (no A2A implementation, just a shared naming layer), see `references/a2a-artifact-mapping.md`.

package/skills/peaks-code/references/external-skill-invocation.md

@@ -40,7 +40,7 @@ The skill body must not silently:
 - commit or sync intermediate artifacts;
 - create remote repositories.
 
-All of these must route through the Peaks CLI under the appropriate command (`peaks artifacts …`, `peaks memory …`, `peaks openspec …`, `peaks standards …`, `peaks codegraph …`, `peaks capabilities …`), with dry-run preview where supported and `--yes` / `--apply` where a real write is required. The `peaks mcp …` command tree was retired in slice #016; MCP install / dispatch is the LLM runtime's job, not the CLI's.
+All of these must route through the Peaks CLI under the appropriate command (`peaks artifacts …`, `peaks memory …`, `peaks standards …`, `peaks codegraph …`, `peaks capabilities …`), with dry-run preview where supported and `--yes` / `--apply` where a real write is required. The `peaks mcp …` command tree was retired in slice #016; MCP install / dispatch is the LLM runtime's job, not the CLI's. Note: peaks-code's 11-step workflow does NOT route through any external artifact lifecycle (OpenSpec or otherwise) — all reads/writes are confined to the LLM-authored artifact workspace under `.peaks/_runtime/<sessionId>/<role>/`. External CLI surfaces remain available for users who call them directly, but Code never calls them on the user's behalf as of the 2026-07-08 RR decoupling slice.
 
 ## Allowed in-process references
 

package/skills/peaks-code/references/project-scan-checklist.md

@@ -2,7 +2,9 @@
 
 > Extracted from `skills/peaks-code/SKILL.md` on 2026-06-09 (slice 019 — slim skill files to references) to keep SKILL.md under the 800-line cap from `common/coding-style.md`. The content below is the verbatim Pre-RD project scan checklist that was previously inline; nothing was paraphrased, just relocated.
 
-Before handing off to `peaks-rd`, scan the project and record findings to `.peaks/_runtime/<sessionId>/rd/project-scan.md`. RD and UI roles read this before starting work. **project-scan.md is a session-scoped singleton** — check if it already exists before regenerating (e.g. via `ls .peaks/_runtime/<sessionId>/rd/project-scan.md`). If it exists and is complete (has `## Archetype` and `## Project mode` sections), reuse it. Only regenerate if missing or incomplete.
+Before handing off to `peaks-rd`, scan the project and record findings to `.peaks/project-scan/project-scan.md` (project-level, git-tracked; sibling of `.peaks/PROJECT.md`). RD and UI roles read this before starting work. **project-scan.md is a project-scoped singleton** — check if it already exists before regenerating (e.g. via `ls .peaks/project-scan/project-scan.md`). If it exists and is complete (has `## Archetype` and `## Project mode` sections), reuse it. Only regenerate if missing or incomplete.
+
+> **Slice 2026-07-15-project-scan-bootstrap (G1 + G2 + G4b):** `peaks workspace init` and `peaks project context` now both bootstrap `.peaks/project-scan/project-scan.md` automatically. On a brand-new 0-1 project (no `package.json` or no source files) the file is seeded with `archetype: unknown` + `(empty)` placeholder rows. On an existing project it is filled from `peaks scan archetype` + `peaks scan libraries` output. Idempotent — re-running `workspace init` does NOT overwrite the existing file unless `--force` is passed.
 
 ### 0. Project archetype detection (MANDATORY — run FIRST, deterministic CLI)
 

package/skills/peaks-code/references/runbook.md

@@ -26,8 +26,8 @@ peaks skill runbook peaks-code --json
 peaks workspace init --project <repo> --json
 peaks workspace reconcile --project <repo> --json
 peaks scan archetype --project <repo> --json
-# → copy archetype, frontendOnly, signals into .peaks/_runtime/<session-id>/rd/project-scan.md (Peaks-Loop Gate A)
-# → copy libraries[] into .peaks/_runtime/<session-id>/rd/project-scan.md under `## Library versions`
+# → copy archetype, frontendOnly, signals into .peaks/project-scan/project-scan.md (Peaks-Loop Gate A; project-level, git-tracked — slice 2026-07-15-project-scan-bootstrap G1)
+# → copy libraries[] into .peaks/project-scan/project-scan.md under `## Library versions`
 peaks scan libraries --project <repo> --json
 # → if archetype != greenfield AND archetype != unknown:
 peaks scan existing-system --project <repo> --json
@@ -149,9 +149,7 @@ peaks sc retention --slice-id <rid> --prd <prd> --rd <rd> --qa <qa> --json
 peaks sc validate --slice-id <rid> --json
 peaks sc boundary --slice-id <rid> --artifact <artifact> --code <file> --json
 
-# 9. Peaks-Loop OpenSpec archive (exit gate; only after QA pass, when openspec/ exists)
-peaks openspec validate <cid> --project <repo> --json
-peaks openspec archive <cid> --project <repo> --apply --json
+# 9. Peaks-Loop Workspace reconcile (slate cleanup)
 peaks workspace reconcile --project <repo> --apply --older-than 7
 
 # 10. Peaks-Loop TXT handoff — invoke peaks-txt which embeds memory markers (BLOCKING)

package/skills/peaks-code/references/standards-preflight.md

@@ -7,7 +7,7 @@
 
 Use `standards init` for first-time creation and `standards update` for existing `CLAUDE.md` append/review behavior. In `full-auto` and `swarm` profiles, `--apply` runs automatically after `--dry-run` succeeds — these files live inside the target project, are required for downstream skill preflight, and producing them is part of finishing the workflow (Peaks-Loop Gate G enforces this). `assisted` and `strict` profiles pause for explicit user confirmation between dry-run and apply.
 
-**CRITICAL — Standards must reflect the project scan.** When generating or updating `CLAUDE.md`, the content must reference concrete findings from `.peaks/_runtime/<sessionId>/rd/project-scan.md`: the detected component library (e.g. "This project uses antd 5.x"), CSS solution (e.g. "Uses Less via Umi"), build tool, state management, and routing. Never emit a generic template that says "read .claude/rules/..." without naming the actual project stack. If the project-scan has not been run yet, run it before standards init/update.
+**CRITICAL — Standards must reflect the project scan.** When generating or updating `CLAUDE.md`, the content must reference concrete findings from `.peaks/project-scan/project-scan.md`: the detected component library (e.g. "This project uses antd 5.x"), CSS solution (e.g. "Uses Less via Umi"), build tool, state management, and routing. Never emit a generic template that says "read .claude/rules/..." without naming the actual project stack. If the project-scan has not been run yet, run it before standards init/update.
 
 **Legacy projects additionally** — when archetype ∈ {legacy-frontend, legacy-fullstack, frontend-monorepo}, the `CLAUDE.md` Conventions section MUST extract concrete naming, directory, service-layer, and hooks conventions from `.peaks/_runtime/<sessionId>/system/existing-system.md` and record them as hard constraints for new code. It must also list the `## Legacy constraints` from `project-scan.md` (class components, moment, enzyme, etc.) and instruct that new code in the same module preserves those patterns unless PRD explicitly authorizes modernization. A `CLAUDE.md` for a legacy project that contains only generic rule pointers without naming the actual conventions is a blocking violation — regenerate it.
 

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 session auto-compact --execute`). Pass `--enforce-job-mode` for single-rid. **≥ 0.95** red line: next Bash/Task fires `peaks session auto-compact-hook`. **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-0-55-1x-detection.md

@@ -73,10 +73,11 @@ The decision is persisted via `peaks preferences set --project <root> --key auto
 
 ## How this integrates with the rest of the workflow
 
-- Step 0.5 (OpenSpec opt-in) — runs first if `openspec/` is missing.
 - Step 0 (anchor) — runs always.
 - Step 0.7 (resume detection) — runs after Step 0.
 - **Step 0.55 (1.x detection) — runs after Step 0.7, only when the project is not on a 2.0 layout.**
 - Step 1 (mode selection) — runs after Step 0.55.
 
 The 1.x detection is intentionally placed AFTER Step 0.7 because the user might already have a 1.x-converted in-flight slice; the resume flow takes precedence over the upgrade prompt.
+
+> **Historical note (prior to 2026-07-08 RR slice):** an OpenSpec first-run opt-in step (Step 0.5) used to sit between Step 0 and Step 0.7. peaks-code is now decoupled from OpenSpec and does not surface that step; this reference still preserves the historical ordering for context.

package/skills/peaks-code/references/step-11-memory-sediment.md

@@ -0,0 +1,56 @@
+# Peaks-Loop Step 11 — Memory sediment (substep details)
+
+Companion to `SKILL.md` §"Peaks-Loop Step 11". This file holds the substep-by-substep bash + flow + sediment history; the SKILL.md only carries the hard-rule one-liner and the link.
+
+## Blocking rule
+
+Code MUST NOT declare a workflow complete until Step 11 has produced ≥ 1 file in `.peaks/memory/` OR the user has explicitly approved a no-sediment outcome via AskUserQuestion. Applies to **all modes** including `assisted` and `strict`.
+
+## Substeps
+
+### 11a — Gate A (txt/ inventory)
+
+```bash
+find .peaks/_runtime/<sessionId>/txt/ -type f -name '*.md' | head
+```
+
+If **0 files** → STOP. Dispatch `peaks-txt` first to write `handoff.md`, then return to 11c.
+
+### 11b — Gate B (memory block embed scan)
+
+```bash
+grep -c 'peaks-memory:start' .peaks/_runtime/<sessionId>/txt/handoff.md || true
+```
+
+If **0 AND this session surfaced a stable project fact** (decision / convention / approved refactor / hard rule), STOP and tell peaks-txt to embed at least one `<!-- peaks-memory:start -->` block first.
+
+### 11c — Canonical extract
+
+```bash
+peaks memory extract --project <repo> --artifact .peaks/_runtime/<sessionId>/txt/handoff.md --apply --json
+```
+
+`--apply` is REQUIRED (without it the command only previews — no files land).
+
+> **CLI reality check (D-010 sediment, 2026-07-09):** The `<!-- peaks-memory:start -->` block must be followed immediately by a **YAML frontmatter** (`title: ...`, `kind: lesson | decision | convention`) and a `---` separator. Each block must close with `<!-- peaks-memory:end -->`. A bare `peaks-memory:start` without the YAML fields is parsed but produces no `plannedWrites` — the CLI silently returns `extractedCount: 0`.
+
+### 11d — Gate C (zero-write outcome)
+
+If `extractedCount === 0` after 11c, fire AskUserQuestion:
+
+> "本次 code 未沉淀任何 `.peaks/memory` 文件。可选: (a) 回去在 handoff.md 嵌入至少 1 个 `peaks-memory:start` block 后重试; (b) 显式接受 no-sediment 并记录为 lesson; (c) 取消完成。"
+
+> **D-010 fix root cause check:** When firing 11d, first inspect whether the block has the YAML frontmatter (`title:` + `kind:` + `---`). If the `<!-- peaks-memory:start -->` exists but no `title:` line follows, fix the block format and re-run 11c — don't ask the user yet. Default option = (a). Code MUST NOT silently accept (b) without user pick.
+
+## Why Step 11 exists
+
+Audit 2026-07-03 confirmed 2 consecutive sessions produced zero `.peaks/memory/` files despite completing RD + QA + handoff artifacts; `assisted` mode silently skipped runbook Step 10 (no STOP condition).
+
+## Why `peaks memory extract` (not `peaks project memories:extract`)
+
+The artifact-scoped extract is canonical; the batch-scoped sibling is for non-handoff flows. Always use `peaks memory extract --apply`.
+
+## Related
+
+- `references/runbook.md` §Step 11 — full bash + flow.
+- `references/project-memory-loading.md` — Step 2.3 memory loading (read-side).

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

@@ -248,6 +248,36 @@ 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
 
+## 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/swarm-dispatch-contract.md

@@ -184,7 +184,7 @@ A skill cannot itself trigger sub-agents — the Skill tool runs in the main loo
    - `feature` / `refactor` / `bugfix` → RD(planning) and QA(test-cases) are always in the swarm
    - `config` / `docs` / `chore` → no swarm. RD/QA artefacts are not required by Gates B/C/D for these types. Skip the Swarm phase entirely and proceed to step 4 (RD implementation) with only the PRD in hand.
 3. **Frontend touch** — does the request affect user-visible behavior? This is decided by:
-   - Reading `.peaks/_runtime/<sessionId>/rd/project-scan.md` `## Project mode` for `frontendOnly` (project-shape signal)
+   - Reading `.peaks/project-scan/project-scan.md` `## Project mode` for `frontendOnly` (project-shape signal)
    - **AND** scanning the PRD body for frontend keywords: 页面 / 组件 / 表单 / 弹窗 / 表格 / 样式 / 布局 / 交互 / UI / UX / page / component / form / modal / table / styling / layout / interaction
    - UI joins the swarm when (a) is `true` OR (b) matches. Both signals required `false` to skip UI.
 

package/skills/peaks-code/references/workflow-gates-and-types.md

@@ -29,7 +29,7 @@ When Peaks-Loop Code coordinates development in a code repository, keep this ord
 
 0. **Peaks-Loop Snapshot** — `peaks doctor` + `peaks project dashboard` to capture baseline state before anything else;
 0.5. **Peaks-Loop Workspace initialization** — `.peaks/_runtime/<session-id>/` created, directory structure verified;
-0.6. **Peaks-Loop Project scan** — archetype, component library, CSS framework, build tool, state management, routing, data fetching, legacy signals detected and recorded to `.peaks/_runtime/<session-id>/rd/project-scan.md`;
+0.6. **Peaks-Loop Project scan** — archetype, component library, CSS framework, build tool, state management, routing, data fetching, legacy signals detected and recorded to `.peaks/project-scan/project-scan.md`;
 0.7. **Peaks-Loop Existing-system extraction** (MANDATORY when archetype ∈ {legacy-frontend, legacy-fullstack, frontend-monorepo}; SKIP for greenfield) — extract visual tokens and code conventions from the live codebase to `.peaks/_runtime/<session-id>/system/existing-system.md`. The path lives under `system/` (not `ui/`) because the file also records non-UI conventions (service-layer signatures, hooks, naming) that backend-only or legacy-fullstack work consumes. See `references/existing-system-extraction.md`. UI design-draft and RD implementation MUST treat the extracted tokens and conventions as hard constraints;
 1. **Peaks-Loop Standards preflight** — `peaks standards init/update --dry-run`, must reference concrete project-scan findings (never emit generic templates);
 2. **Peaks-Loop PRD phase** — capture request as canonical artifact, extract scope and acceptance criteria:
@@ -44,7 +44,7 @@ When Peaks-Loop Code coordinates development in a code repository, keep this ord
 6. **Peaks-Loop QA validation** (auto-proceed from RD in full-auto) — execute test cases + API checks + Playwright MCP headed browser E2E for frontend + security/perf checks + test report;
 7. **Peaks-Loop RD↔QA repair loop** — if QA verdict is `return-to-rd`, loop back to step 4 (RD implementation) and re-run through QA; max 3 repair cycles, then emit blocked TXT regardless;
 8. **Peaks-Loop SC phase** — change-control evidence: impact, retention, validate, boundary;
-9. **Peaks-Loop OpenSpec archive** — exit gate: validate → archive only after QA verdict=pass (when `openspec/` exists);
+9. **Peaks-Loop Workspace reconcile** — slate cleanup;
 10. **Peaks-Loop TXT handoff capsule** — mode, validated decisions, artifact paths, standards deltas, open questions, next action;
 11. **Peaks-Loop Final snapshot** — `peaks project dashboard` + `peaks skill doctor` to confirm the workflow closed cleanly.
 
@@ -54,8 +54,8 @@ You cannot declare a phase complete from memory. Each gate below is a `ls` comma
 
 **Peaks-Loop Gate A — After workspace init + project scan:**
 ```bash
-ls .peaks/_runtime/<sessionId>/rd/project-scan.md
-# Expected output: .peaks/_runtime/<sessionId>/rd/project-scan.md
+ls .peaks/project-scan/project-scan.md
+# Expected output: .peaks/project-scan/project-scan.md
 # "No such file" → STOP, run project scan first
 # File present but missing `## Archetype` or `## Project mode` sections → INCOMPLETE, rerun scan
 # File present and complete → reuse (project-scan is a session-scoped singleton)