Skip to content

pm-dispatch(execution): an idle lane seat does four pre-triage chores on its own cards, never grades, and rings the triage mailbox instead of waiting - #28935

Merged
objectstack-fleet[bot] merged 3 commits into
mainfrom
claude/project-thread-b1yw53
Nov 2, 2026
Merged

pm-dispatch(execution): an idle lane seat does four pre-triage chores on its own cards, never grades, and rings the triage mailbox instead of waiting#28935
objectstack-fleet[bot] merged 3 commits into
mainfrom
claude/project-thread-b1yw53

Conversation

@objectstack-fleet

@objectstack-fleet objectstack-fleet Bot commented Nov 2, 2026

Copy link
Copy Markdown
Contributor

Requested by Irene · project thread

Fixes #28899
Clause-②: no
Metadata-protocol: none

维护者速读

  • 改了什么:车道席没有卡可认领、或在等分诊时,改做四件分诊前杂务(交付收尾、旧卡复读、给新卡写一行建议级别、补 finding 复现),永不自己定级;等分诊必须先敲分诊信箱,报告里引那条评论。分诊定级时读这行建议级别。
  • 为什么改:维护者常听到车道说「在等分诊」;这四件不需要定级权,车道闲着就能做,分诊接到的卡也已带好读数。
  • 风险与回滚:只改两份规则文本,无代码、无门禁、无新标签、上限未动;为守行数上限删了 7 行在别处已有原文的复述(出处见下表)。回滚 = revert 本 PR。
  • 先看:① references/execution-duties.md 〈候选与批次〉末尾新增 6 行;② references/triage-duties.md 新增 1 行;③ 下方 7 行删除的出处表。
  • 两分钟验证:pnpm check:pm-skill-ratchet && pnpm check:pm-skill-id-lint 全绿,wc -l 两文件仍为 188 / 102。
  • 需要你做的:无,席位落地(Tier S .claude/**,席位达档复核 PASS 后入队)。

What changes

Text only, two files under .claude/skills/pm-dispatch/references/, inside the claim's file surface.

The shared block — six lines at the end of execution-duties.md 〈候选与批次〉. SKILL.md's phase index names this file as read by every execution seat on every fire, so every lane seat reads the block without a per-lane copy:

  • Scope: a lane seat with no claimable card, or one waiting on triage, works its own lane's cards only; it never grades (the lane self-triage exception stands as written).
  • ① Delivery hygiene: a delivered card gets a closure record; a release or shift end never removes a pm:* label without setting the next state in the same write; a card whose Blocked-by: targets have all closed is put to triage for unblocking.
  • ② Stale-card re-read: the oldest open cards are re-read against origin/main (premise still true, already fixed) with the evidence posted; a fixed card goes to triage or the maintainer to close, never closed by the lane.
  • ③ Pre-read: on a new card, a reproduction reading, the dedupe query with its hit count, and one Suggested-grade: pN · REASON line; no priority:* label is written.
  • ④ Finding repro: the lane's finding cards get the missing reach: or reproduction, or a stated reason they miss the three classes.
  • Never waiting on triage in silence: ring the matching mailbox key (project-forms.md); a report that says "waiting on triage" cites the mailbox comment id it rang, and one that cannot has not asked.

The triage pointer — one line in triage-duties.md beside the self-triage review line: the grading pass reads lanes' Suggested-grade: lines as input, not a grade, and confirms or changes them, giving a reason when it does not adopt one. The 「车道席自分诊」 exception lines are byte-identical.

Paying the ratchet — no ceiling moved. Both files stay at their ceilings (execution-duties.md 188/188, triage-duties.md 102/102), every added line is at most 120 bytes, and no line was re-wrapped to buy room. Seven lines left, each restating a rule that still lives elsewhere:

Deleted line Where the rule still lives
execution-duties 〈候选与批次〉「无分诊审计评论即非候选;例外:sub-issue 承父卡首触,main-red 锚卡见 landing-operations.md」 SKILL.md 〈域车道〉 (read every fire): 「domain:*、队列与级别只由分诊落笔;两例外:sub-issue 承父卡域、main-red 锚卡,免审计评论」 and 「无分诊审计评论 ⛔ 不得认领」; main-red handling in landing-operations.md 约定③
execution-duties 〈候选与批次〉「解锁那一刻 PM 自己的判断最不可信:裁决收窄或关掉了那张卡是假设不是前提」 same section: 「派发前做前提过时检查,裁决同罪」, 「引用读到的现状,⛔ 不引用裁决写作时的描述」 and the 「阻塞解除后延后单重定价」 pair
execution-duties 〈派发〉「条款②判据即 triage 机械边界测试与 references/lanes/spec.md 席内分派判据,⛔ 不另抄」 the test itself: execution-duties 〈认领〉「判据:本卡放宽接受集或扩大公开面吗」 and triage-duties 「机械边界测试(条款②判据)」; its boundaries stay in the lines below it
execution-duties 〈派发〉「无条件条款只住角色文件,冲突时它胜、错了修那里,⛔ 不靠派发词临时覆盖」 .claude/agents/os-dev.md 〈标准条款住在这里〉「派发词与本文件冲突时以本文件为准:无条件条款住这里,错了也在这里改」, and execution-duties 「标准非协商条款 ⛔ 不抄进派发词」
execution-duties 〈收集〉「停摆永不自愈:携带任务中途状态的完成通知本身就是停摆信号」 the next two lines: 「报告末句是意图不是结果 = 停摆非完成:当刻 SendMessage 续派」 and 「停摆 ⇒ 立刻 SendMessage …⛔ 不等任何静默阈值」
execution-duties 〈复核〉「报告在草稿 PR 时点到达,CI 收敛读数只属于复核侧:gate in_progress 是诚实读数」 os-dev.md DoD 「报告在本地验证走完时交付,CI 收敛等待归 PM 不归你」 and 「门禁状态如实记录,in_progress 是诚实值」; review-checklist.md 「收敛期转红走补丁轮续派原 dev,⛔ 不作 REWORK 的理由」
triage-duties 「轮次报告设呈裁清单专节(卡、推荐、所引权威),⛔ 无代裁清单、无追认表」 seat-lifecycle.md 〈轮次报告与节奏〉 (every seat, every fire) 「报告加升级项、呈裁清单…」 and SKILL.md 红线 「⛔ 无代裁通道」; the column hint (card, recommendation, authority) is the one detail not restated elsewhere

No issue number in the skill text (check:pm-skill-id-lint green). No package publishes from this diff, so skip-changeset.

Tests

No code changes; the gates are the tests. Derived with node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (merge base on origin/main, 2 paths, +7/−7) and run one by one with the exit code captured before any pipe, on head ec896289:

19 derived commands, all exit 0; --ran reconciliation 19/19, 0 NOT-MEASURED 
node scripts/check-closing-keyword-parity.mjs :: exit 0
node scripts/check-closing-keyword-parity.mjs --self-test :: exit 0
node scripts/check-comment-mask-corpus.mjs :: exit 0
node scripts/pm/check-harness-current.mjs --self-test :: exit 0
pnpm --filter @objectstack/lint run check:doc-formula-expressions :: exit 0
pnpm check:agent-test-spelling :: exit 0
pnpm check:cross-package-test-inputs :: exit 0
pnpm check:doc-authoring :: exit 0
pnpm check:driver-memory-census :: exit 0
pnpm check:gitlink-declared :: exit 0
pnpm check:nul-bytes :: exit 0
pnpm check:pm-governed-merges :: exit 0
pnpm check:pm-half-states :: exit 0
pnpm check:pm-label-write :: exit 0
pnpm check:pm-skill-id-lint :: exit 0
pnpm check:pm-skill-ratchet :: exit 0
pnpm check:refd-timer-probe :: exit 0
pnpm check:skill-frame-sync :: exit 0
pnpm check:watch-hint-literal :: exit 0

dispatch-gates --ran: 19 derived families accounted for, 19 run, 0 NOT-MEASURED (a derived zero: all 19 recorded an exit code and none is 3). Also run, outside the derived list: pnpm check:pm-settings-deny-roster (its roster sits under .claude), pnpm check:pm-governed-prose, node scripts/check-published-list-mirrors.mjs, each exit 0.

Acceptance notes

  • Premise, half held. No lane file lists idle-lane chores today (holds). But execution-duties.md has no empty-queue step: the per-card empty-queue table and the standby list live in seat-lifecycle.md 〈停止条件与待命〉, outside the claim's file surface. The block sits at the end of 〈候选与批次〉, which that table's blocker vocabulary already points at (「候选定义反面(execution-duties.md)」). A one-line pointer in seat-lifecycle.md's standby list would be the direct link; not made here (surface). Carrier: the seat.
  • No per-lane pointer lines (the dispatch's optional route). Every lane seat already reads execution-duties.md on every fire, so six pointers would add lines with no reader the block lacks; lanes/ui.md, lanes/server.md and lanes/hotcrm.md sit at headroom 0, so each would cost another deletion. It also leaves the lane files to the sibling card that adds a pointer line there. The card's acceptance allows the shared section in place of per-lane copies.
  • Unblock signal. The card text says Notify: unblocked to triage, but in the closed mailbox vocabulary (scripts/pm/record-recognisers.mjs) Notify: unblocked runs from the seat that queued the card to the lane, for p0/p1 only. A lane asking triage to unblock uses Request: triage (project-forms.md: 「问卡去留、路由或定级」). The chore says 「请解锁」 and the bell line routes it through the existing keys; no key was added.
  • lanes/skills.md 常设承诺 「首触定级每轮」 reads as the skills lane grading at first touch, next to two lines that hand finding grading to triage. Against the ruling that grading stays with triage it may be stale. Outside this card's wording, not touched. Carrier: the seat.

Verification

Head ec896289 (main merged once before opening). Line budget after the edit: execution-duties.md 188/188, triage-duties.md 102/102; longest added line 120 bytes. The os-dev-report comment on the card carries the same readings.

Generated by Claude Code

claude added 2 commits November 2, 2026 04:24
@claude
pm-dispatch(execution): an idle lane seat does four pre-triage chores…
915e5bc 
… on its own cards, never grades, and rings the triage mailbox instead of waiting

One shared block at the end of execution-duties.md 〈候选与批次〉, read by
every execution seat every fire: delivery hygiene, stale-card re-reads, a
`Suggested-grade:` pre-read that writes no priority label, and finding
repro; a "waiting on triage" report cites the mailbox comment it rang.
triage-duties.md gains one line: the grading pass reads `Suggested-grade:`
lines as input. The lane self-triage exception is unchanged.

Paid by deletion, no ceiling moved: six restated lines leave
execution-duties.md and one leaves triage-duties.md (sources in the PR body).

Claude-Session: https://claude.ai/code/session_01J8VVAoanyVrJxtZ2gGqdZw
Co-authored-by: Claude <noreply@anthropic.com>
@claude
Merge origin/main into claude/project-thread-b1yw53
ec89628 
Claude-Session: https://claude.ai/code/session_01J8VVAoanyVrJxtZ2gGqdZw
Co-authored-by: Claude <noreply@anthropic.com>
@objectstack-fleet objectstack-fleet Bot added the skip-changeset PR has no user-facing published change; bypasses the changeset gate label Nov 2, 2026
@github-actions github-actions Bot added size/s documentation Improvements or additions to documentation labels Nov 2, 2026
@objectstack-fleet

objectstack-fleet Bot commented Nov 2, 2026

Copy link
Copy Markdown
Contributor Author

domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J

Contract-review-claim: session_01J8VVAnMUWA9rufXzFaEh4J · head ec89628 · 2026-11-02T04:41Z

The seat review runs at CONTRACT_REVIEW_TIER in an isolated subagent (the seat session is not at that tier).

@objectstack-fleet objectstack-fleet Bot mentioned this pull request Nov 2, 2026
Closed
@objectstack-fleet

objectstack-fleet Bot commented Nov 2, 2026

Copy link
Copy Markdown
Contributor Author

domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: ec896289c99d6e3e7b215500c93ea041acb3a732
Local-runs: none
Scope: .claude/skills/pm-dispatch/references/execution-duties.md, .claude/skills/pm-dispatch/references/triage-duties.md + .claude/skills/pm-dispatch/SKILL.md (state model, phase index, domain-lane rules, 升级与决策), references/seat-lifecycle.md (轮次报告, 停止条件与待命), references/dispatch-runbook.md (巡查常设项), references/project-forms.md (信箱 PR), references/landing-operations.md (main-red), references/review-checklist.md, references/decision-analysis.md, references/coordinator.md, references/contract-review.md, references/lanes/{hotcrm,skills}.md, .claude/agents/os-dev.md, scripts/pm/record-recognisers.mjs (mailbox table), scripts/pm/check-skill-line-ratchet.mjs (ceilings)

Reviewer: an isolated subagent at CONTRACT_REVIEW_TIER, adopted by the seat; inputs were the card and its comments, the PR body, file list and net diff, and the head's check-runs.

① Derived judgments

Net diff against the merged-in main: two files, +7/−7. execution-duties.md gains six lines at the end of 候选与批次 and loses six restatements; triage-duties.md gains one pointer line and loses the 呈裁清单 line. Ceilings held (188/188, 102/102); longest added line 120 bytes; no issue numbers; the self-triage exception lines are byte-identical to main.

  • Every execution seat (domain:* and repo:hotcrm): idle or waiting on triage ⇒ works its own lane's cards, does the four chores, never grades. Reader path holds (the phase index sends every execution seat to execution-duties.md every fire). Right.
  • Chore ① delivery hygiene — closure record; no pm:* stripped without the next state in the same write: right. Third clause 「Blocked-by: 全关请解锁」: wrong against the tree (see ④).
  • Chore ② stale re-read — evidence posted, a fixed card goes to triage or the maintainer, the lane never closes on merits: right.
  • Chore ③ pre-read — repro, dedupe count, one Suggested-grade: line, no priority:* label: right.
  • Chore ④ finding repro — right.
  • Never silent — ring the matching key; a report saying 等分诊 cites the rung comment id: right; the three lane→triage keys are exactly the mailbox table's, no key added.
  • Triage — reads Suggested-grade: lines as input in its grading pass, gives a reason when not adopting: right; grading and routing stay triage's.

② Semver level

Nothing published; skip-changeset correct, Check Changeset success. Clause-②: no, Metadata-protocol: none correct.

③ Boundary flags

PR author objectstack-fleet[bot]. os-dev-report 6569223549 open_questions: [].

  1. No per-lane pointers — accepted on the "(or the shared section)" arm. (The headroom reason holds for ui/server/hotcrm only; devx/skills/spec each have one spare line.)
  2. Empty-queue step lives in seat-lifecycle.md 〈停止条件与待命〉, outside the surface; no pointer added — see ④ (should-fix).
  3. Chore ① unblock via Request: triage instead of the card's Notify: unblocked — the key reading is right (Notify: unblocked runs from the re-queuing seat to the lane), but the tree answers whether the lane should ask at all the other way — see ④.
  4. No MCP assign/review — correct.
  5. Model-free trailers — per AGENTS.md.

Seven deleted lines, each checked for a surviving home with the same reader: 无分诊审计评论即非候选 (SKILL.md domain-lane rules, 待命 blocker vocabulary) — survives; 解锁那一刻判断最不可信 (前提过时检查, 延后单重定价, same section) — survives; 条款②判据不另抄 (criterion in the same file and lanes/spec.md) — survives; 无条件条款只住角色文件 (os-dev.md role-file-wins + the 派发词 no-copy rule) — survives, PM-facing "fix it there" now implicit (should-fix); 停摆永不自愈 (报告末句是意图不是结果) — survives; CI 收敛读数属复核侧 (review-checklist, os-dev.md) — survives; 呈裁清单专节 (seat-lifecycle 轮次报告, SKILL.md 无代裁通道) — survives, the column hint has no verbatim home (decision-analysis carries the item shape; acceptable).

④ Design judgment

  • Acceptance 1 (four chores + never-grade line, lane files or shared section): met.
  • Acceptance 2 (empty-queue step points at the chores; "waiting on triage" names the doorbell): the card's premise is off — execution-duties.md has no empty-queue step; that table and 待命五项 live in seat-lifecycle.md. Substance met (line 57 triggers on 无卡可认领或等分诊; line 62 names the doorbell). Should-fix, carrier the seat: one pointer from 待命五项.
  • Acceptance 3 (one triage pointer; self-triage unchanged): met.
  • Acceptance 4 (no new gate or label): met.
  • MUST-FIX — chore ①'s 「Blocked-by: 全关请解锁」 sends the lane to triage for an unblock the lane already owns. dispatch-runbook.md 〈巡查常设项〉「车道巡查同跑:hold/closed 解锁重验回队」, seat-lifecycle.md 待命五项 (pm:blocked/pm:on-hold 解锁扫描) and SKILL.md's Blocked-by: reverse-index duties (上游关单放回解锁卡) all make the lane the re-queuer of its own blocked cards. Read with line 62 the clause becomes a Request: triage round trip — the very 「在等分诊」 the ruling 「因为我经常听到车道说在等分诊」 set out to remove. Reword to a self-act (e.g. 「Blocked-by: 全关即重验回队」, shorter than the line it replaces) and leave Notify: unblocked where the mailbox table puts it.
  • Rule style: one executable sentence per line, no narrative, date or tracker number. Holds.

Check-runs at record time: all success or skipped except Lint · PM dispatch-gates (in progress).

Implemented-by: claude/project-thread-b1yw53
Reviewed-by: session_01J8VVAnMUWA9rufXzFaEh4J

VERDICT: FAIL

Generated by Claude Code

@claude
pm-dispatch(execution): an all-closed Blocked-by is the lane's own re…
c53bdb0 
…-verify-and-requeue, not a triage round trip

Chore 1's last clause sent the lane to triage for an unblock the lane already
owns (the lane patrol's hold/closed re-verify, the standby unblock scan, the
Blocked-by reverse-index duty). It now says the lane re-verifies and
re-queues the card itself; Notify: unblocked stays where the mailbox table
puts it. "释放/收班" becomes SKILL.md's own word 离手 so the line keeps
within 120 bytes (118).

Claude-Session: https://claude.ai/code/session_01J8VVAoanyVrJxtZ2gGqdZw
Co-authored-by: Claude <noreply@anthropic.com>
@objectstack-fleet objectstack-fleet Bot mentioned this pull request Nov 2, 2026
Closed
@objectstack-fleet

objectstack-fleet Bot commented Nov 2, 2026

Copy link
Copy Markdown
Contributor Author

domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: c53bdb054bb966d60dfb21f119ee914462ed44a7
Local-runs: none
Scope: .claude/skills/pm-dispatch/references/execution-duties.md, .claude/skills/pm-dispatch/references/triage-duties.md + .claude/skills/pm-dispatch/SKILL.md (106, 133-140, 146), references/seat-lifecycle.md (38, 85-92), references/seat-post-protocol.md (35), references/state-machine.md (41), references/dispatch-runbook.md (47, 66), references/project-forms.md (34-49), scripts/pm/record-recognisers.mjs (mailbox table 1262-1268), scripts/pm/check-skill-line-ratchet.mjs (1277, 1312, 1351)

Delta record over the FAIL record on ec896289 (comment 6569334797). The new head is one commit on ec896289, no force-push, parent verified; the delta touches one line, execution-duties.md:58. Everything the FAIL record judged and did not name as must-fix stands unchanged: the net diff against the merged main is still two files, +7/-7, with the same seven deletions, the same six-line block and the same triage pointer line. This record re-judges the must-fix, the wording change made to pay for it, and the ratchet readings; ①–④ below cover the delta and restate only what the delta touches.

① Derived judgments

Delta: 「① 交付卫生:交付卡补收尾记录;释放/收班摘 pm:* 同笔设下一态;Blocked-by: 全关请解锁。」 becomes 「① 交付卫生:交付卡补收尾记录;离手摘 pm:* 同笔设下一态;Blocked-by: 全关即重验回队。」 (118 bytes).

  • Lane seat, chore ① third clause: a card whose Blocked-by: targets have all closed is re-verified and re-queued by the lane itself. RIGHT — this is the duty the tree already assigns to the lane in the same words: dispatch-runbook.md:47 「车道巡查同跑:hold/closed 解锁重验回队」, seat-lifecycle.md:91 (待命五项 includes the pm:blocked/pm:on-hold 解锁扫描 for the seat whose queue is empty), SKILL.md:135-136 (one Blocked-by: read feeds three duties, the first 上游关单放回解锁卡). 重验 carries the cross-repo criterion at SKILL.md:138-140 (consumer-installable, not merely merged) because it is the same re-verify the runbook names. No triage round trip remains in the block; the never-silent line (62) now governs only the cases that are triage's (grade, route, review, a card's fate).
  • Mailbox table: untouched on this head (no diff under scripts/pm/ or project-forms.md). Notify: unblocked stays where record-recognisers.mjs 1268 and project-forms.md:44 put it — from the seat that moved the card to pm:queue, to the lane, p0/p1 only. The lane that re-queues its own card under chore ① is that seat for its own cards and owes the notify only when a sibling seat of the same lane is the reader; nothing in the delta changes that. RIGHT.
  • Lane seat, chore ① second clause: 离手 replaces 释放/收班. 离手 is SKILL.md's own term (SKILL.md:106 「离手恒走释放 —— 三因 = 改路由、前提证伪、弃飞无接管,去向 = 新标签态或车道」) and state-machine.md:41 defines the 释放者 as 「让卡离手的席位」. Shift-end is covered: seat-lifecycle.md:38 「收班 = 状态 flush + 在飞默认交回队列」 and seat-post-protocol.md:35 「收班默认交回」 make the sign-off handback a release, so every card that leaves the seat at 收班 is 离手; a stay-behind card does not leave the seat and strips no pm:*, and a card closed at sign-off sets its state by the closure itself (SKILL.md:133). The term is therefore at least as wide as 释放/收班 for the rule's risk shape (the H13 rows: delivered, released, no next state); it does not narrow the rule. RIGHT, no finding.
  • Every other obligation from the FAIL record's ① (chores ②–④, the never-silent line, the triage pointer, the untouched self-triage exception) is unchanged on this head.

② Semver level

Unchanged: nothing published, two instruction files; skip-changeset, Clause-②: no, Metadata-protocol: none remain correct; labels documentation, size/s, skip-changeset on the PR.

Check-runs on c53bdb05: success — Lint · PM dispatch-gates (pm-skill-ratchet and pm-skill-id-lint, the two gates this diff can move), Lint · repo & tooling gates, Lint · docs & code guards, Lint · ESLint, Lint · CI release & engine gates, Lint · scoped ×3, Lint & Repo Gates, Type Check · source gates, Type Check · consumer gates, Test Core, Test UI, Dogfood Regression Gate, Governed Surface Queue Guard, PR-body guards, Check PR Size, Auto Label, filter. Still in progress at record time: Type Check · debt ledger, Type Check · workspace. Check Changeset shows only a skipped instance on this head so far (it passed on ec896289 under the same label). Skipped: matrix, opt-in and path-filtered jobs; none red.

③ Boundary flags

PR user.login = objectstack-fleet[bot]; open_questions: []; one delta commit, model-free trailer pair.

  1. Optional role-file half-line restore not taken — ANSWERED, accepted: the delta freed no line (188/188), and the FAIL record filed it as should-fix only.
  2. seat-lifecycle.md untouched — ANSWERED, consistent with the claimed surface; the standby-list pointer stays a should-fix carried by the seat (seat-lifecycle.md at 96/96).
  3. Dev flags the PR body as stale (head sha ec896289 and the Unblock-signal acceptance note, which still describes the Request: triage route) and does not PATCH it — correct per os-dev; the body now misdescribes the tree on one point, so the seat's ACCEPT or body correction must name the new route (should-fix, on the body, not the tree).
  4. Ratchet readings, by reading the head's blobs (no local run): execution-duties.md 188 lines against ceiling 188 (check-skill-line-ratchet.mjs 1351), triage-duties.md 102 against 102 (1312), seat-lifecycle.md 96 against 96 (1277); longest line in each of the three files 120 bytes; the changed line 58 is 118 bytes; no ceiling row moved. CI's Lint · PM dispatch-gates success on this head is the gate verdict.
  5. Carried from the FAIL record, unchanged and non-blocking: the implicit PM-side 「错了修角色文件」 imperative, the lost 呈裁清单 column hint, lanes/skills.md:21 「首触定级每轮」 as a separate finding, the overstated headroom reason in the round-0 report.

④ Design judgment

  • The must-fix is closed: execution-duties.md:58 now makes the lane the self-acting re-verifier and re-queuer of its own all-closed Blocked-by: cards, in the runbook's own words, and leaves Notify: unblocked to the mailbox table. The ruling's rationale 「因为我经常听到车道说在等分诊」 is served rather than contradicted: the one unblock case that used to be a triage wait is now a lane act.
  • The paid wording (离手) is the repo's established term for a card leaving a seat for any cause, release and sign-off handback included; the rule is not narrowed and reads tighter.
  • Acceptance bullets 1, 3 and 4 remain met as on ec896289; bullet 2 remains met in substance (the block self-triggers on the empty-queue condition in the file every execution seat reads every fire), with the direct pointer from seat-lifecycle.md's standby list still owed outside the surface.
  • Byte and line budgets hold at the cap with no re-wrap and no ceiling move; no gate, label or mailbox key added; no issue number in rule text.
  • Must-fix: none.

Implemented-by: claude/project-thread-b1yw53
Reviewed-by: session_01J8VVAnMUWA9rufXzFaEh4J

VERDICT: PASS

Generated by Claude Code

@objectstack-fleet

objectstack-fleet Bot commented Nov 2, 2026

Copy link
Copy Markdown
Contributor Author

domain:skills#1 · huangyiirene · session_01J8VVAnMUWA9rufXzFaEh4J

维护者速读

  • 改了什么: execution-duties.md「候选与批次」末加六行:车道席无卡可认领或在等分诊时,做本车道卡上的四项杂务(① 交付卫生 ② 陈卡复读 ③ 预读并只给一行 Suggested-grade: ④ finding 复现),永不定级、不贴 priority:*;要分诊出手时按信箱表敲门,报告里写「等分诊」须引敲门评论 id。triage-duties.md 加一行:分诊定级时把 Suggested-grade: 当输入、不采纳要给理由。两文件各删等量重述,行数不变。
  • 为什么改: 落实你的裁决「因为我经常听到车道说在等分诊」——车道空闲时有事做,真要分诊时敲门而不是干等。
  • 风险与回滚: 纯内部指令文本,两文件 +7/−7,上限未动(188/188、102/102);不加门禁、标签或信箱键。回滚即 revert。
  • 先看: execution-duties.md 第 57–62 行;triage-duties.md 新增的指针行。
  • 两分钟验证: node scripts/pm/check-skill-line-ratchet.mjs 绿;grep -n "Suggested-grade" .claude/skills/pm-dispatch/references/*.md 命中这两处。
  • 需要你做的: 无。Tier S,本席落地。
  • 席位意见: ACCEPT。返工一轮改掉了唯一必改项:Blocked-by: 全关的卡由车道自己重验回队,不再去等分诊。

席位核对 — ACCEPT · domain:skills#1 · head c53bdb054bb966d60dfb21f119ee914462ed44a7

  1. 形态: draft → main,正文首段 Fixes #28899,无他卡关键词;作者 objectstack-fleet[bot](中继)。
  2. 前提实测: main 上两文件无空闲车道杂务块、无 Suggested-grade: 输入行(复核记录 6569334797 读树确认,delta 记录 6569472160 沿用)。
  3. 门禁 check-runs(钉 head c53bdb05): success 22 / skipped 18 / 无红无在跑;Lint & Repo Gates、Type Check · workspace、Lint · PM dispatch-gates success;Check Changeset 本 head 为 skipped(skip-changeset,上一 head ec896289 为 success)。check-governed-merges --pr 28935:Tier S(.claude/** ×2),+7/−7。
  4. 接受的偏差: 返工轮把 「释放/收班」 换成 SKILL.md 自己的 「离手」 以守 120 字节;delta 复核判其不收窄(收班交回即释放)。可选的角色文件半句未恢复(无空行可付)。PR 正文已过时,以此为准: head 应为 c53bdb05;正文 「Unblock signal」 一条所述 Request: triage 路线已不存在,现为 「Blocked-by: 全关即重验回队」,车道自做,Notify: unblocked 仍按信箱表。
  5. 范围外去向: seat-lifecycle.md 待命五项缺指向本块的指针(96/96,本席另行承运);lanes/skills.md 「首触定级每轮」 歧义作为 finding 另记;两条已有家的重述(角色文件 「错了修那里」、呈裁清单列提示)不另立卡。Skills-docs: 不适用(Metadata-protocol: none)。
  6. 落地条款: Tier S,现 head 同形达档 PASS 在案(6569472160)+ 全绿 ⇒ 本席经中继 pr_ready + automerge_enable。本会话分类器已拦过同类落地(docs(pm-dispatch): seat sessions run at effort high, checked at seat-taking #28926,记于 pm-dispatch(seats): the triage seat and every lane PM seat run at effort high; a dispatched os-dev keeps its per-card effort (floor xhigh for protocol-semantics or docs/adr/** cards) #28901),本席不重试,等维护者。

Generated by Claude Code

@objectstack-fleet objectstack-fleet Bot mentioned this pull request Nov 2, 2026
Merged
@objectstack-fleet
objectstack-fleet Bot marked this pull request as ready for review November 2, 2026 05:18
@objectstack-fleet
objectstack-fleet Bot enabled auto-merge November 2, 2026 05:18
@objectstack-fleet
objectstack-fleet Bot added this pull request to the merge queue Nov 2, 2026
Merged via the queue into main with commit 00d0668 Nov 2, 2026
41 checks passed
@objectstack-fleet
objectstack-fleet Bot deleted the claude/project-thread-b1yw53 branch November 2, 2026 05:33
@objectstack-fleet objectstack-fleet Bot mentioned this pull request Nov 2, 2026
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

documentation Improvements or additions to documentation size/s skip-changeset PR has no user-facing published change; bypasses the changeset gate

Projects

None yet

Milestone

No milestone

1 participant

@claude