[Decision] A card triaged to one lane carries an item that lands in another (#23513: an engine card with a skills/** item) — who splits it, and when

Ruled: 6184818831 · letter D · 2026-10-15T05:02Z

Filed by domain:engine seat 2 (#20966) · os-steve · session_01ADzJtzYTLUfgrRZHxkagkX, on the maintainer's word in this seat's session, 2026-10-15, verbatim: 「这种情况该怎么处理立决策卡」. The maintainer's follow-up in the same discussion, verbatim: 「如果你自己把 skills 相关的开发拆分出来立卡呢」 (option D below).

One-line question

A card triaged to one lane carries an item whose fix lands in another lane. The live case: #23513, domain:engine, item 1 edits skills/**. Who splits that item out, and when, so the off-lane work is neither done by the wrong lane nor left waiting?

Background

grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513 is a carried follow-up of C7b S3, triaged domain:engine p2. Its item 1 restates a rule in skills/objectstack-data/rules/security.md, which is governed and Tier H. Items 2 and 3 are engine texts.
Triage's grade (6172504723) and its unlock (6173812728) say "Item 1 is a governed skills/** PR on its own", but leave the whole card domain:engine.
When dispatching, this seat first planned to carry item 1 as well. The maintainer asked why it is not the skills lane.

Governing text

SKILL.md 〈域车道〉:
- "锚定规则:每个包恰属一个域;domain:* = 修复落地包的域"
- "domain:* 只由分诊席产出,例外仅在飞卡衍生 sub-issue 承父卡域"
- "跨域例外路径:拆不动的跨域单 PR 由分诊席指定一个车道 PM 认领"
The domain table: domain:skills = "governed 面全量", and skills/** is a governed surface.
SKILL.md 状态模型 pm:retriage: "向分诊提问(改判、跨域 PR 指定车道、改路由、拆卡…)… 与现行 pm:* 并存;带本标签的 pm:queue 卡跳过派发"
references/execution-duties.md:
- "部分落地(PR 带 Refs #N (item k))即释放… 余项需换道/拆分加 pm:retriage"
- "自在飞派发升级出的卡首行一句:本卡承哪一半、父卡留哪一半"
Precedents:
- [finding] Cards whose fix LANDS on a governed surface but whose SYMPTOM is engineering are being routed to domain:devx — four in one round, one already caused a cross-lane dispatch #13593: the triage ruling that a card whose fix lands on a governed surface is domain:skills.
- decision: which model serves the clause-② contract review — fable on every lane (ruling 5771798588), or fable only for the spec / skills lanes and governed surfaces with opus elsewhere? #23352 / pm-dispatch SKILL.md 〈路径分叉〉: the under-tier route for a skills-face hunk is one route — an isolated review at the face's tier (#23352 ruling B′), not "hand it to the skills seat" beside it #23445 (ruling B′): a skills-face hunk can be reviewed by an isolated review at the face's tier.
Protocol statement: options B and C use today's rules. Option D needs one rule line: a split-off off-lane item does NOT inherit the parent's domain. Today a derived sub-issue inherits it, which would label item 1 domain:engine again.

Premises (each with a re-check)

P1. grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513 is unclaimed pm:queue, domain:engine, item 1 in skills/**. Re-check: curl -sS https://api.github.com/repos/objectstack-ai/objectstack/issues/23513 | jq '.labels[].name,.assignees'.
P2. No rule names a mixed-lane queued card. Re-check: git grep -n "拆卡\|跨域" origin/main -- .claude/skills/pm-dispatch.

Options

	What the seats do	What the customer sees	Workload
A	The executing seat carries the whole card: one dev, a separate PR per surface; the skills PR is reviewed at the face's tier (B′) and approved by a human (Tier H)	Fastest; governed skill text written and judged by a non-skills seat	0 rule lines; the same PRs either way
B	Split first: the seat sets `pm:retriage`; triage splits; the seat dispatches only after	Clean, but every lane item waits a triage round	0 rule lines; one triage round per case
C	Claim and ask in parallel: the seat claims its lane's items, adds `pm:retriage`, and triage splits the off-lane item while the dev works	No waiting; for a while the card holds a claim and a retriage request at once	0 rule lines (allowed today)
D	The seat splits it itself: it files a new card for the off-lane item (no domain label; first line "carries item 1 of #N; #N keeps items 2, 3"); triage stamps `domain:skills` at first touch; the original card keeps only its lane's items and closes with `Fixes`	No waiting, one card per lane at once, nothing forgotten	One `.claude/skills/pm-dispatch/references/execution-duties.md` line (Tier S, skills lane); no precedent PR

Business meaning, one line each:

A: one team does the work of two, fast but outside its role.
B: stop and ask the dispatcher to re-route before starting.
C: start your part and ask the dispatcher to re-route the rest.
D: start your part and hand the rest to the right team yourself, with the dispatcher only confirming the team.

Recommendation

D, fallback C. Confidence gap: no measurement of how often triage files mixed-lane cards (#13593 counted four in one round).

The end state in two years: every card belongs to exactly one lane, and the lane that owns a surface writes and reviews its text. That is the shape GitHub CODEOWNERS and monorepo ownership tools take.
D reaches that state at the moment the mix is found, by the seat that holds the context, with no waiting. A leaves governed text with a non-owner. B costs a round each time. C reaches the same end state but later, and through a mixed-label interval.
Self-check: reading ① alone gives D; ②③④ do not flip it.

os-decision-facets

① 项目长远合理性:D 让「一卡一车道」在发现混卡的那一刻成立，归属与审查都回到本车道，不扩大任何特例；A 让非本车道写受管文本，是特例增生。
② 实际业务拉动:今天就撞上(grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513 的第 1 项);[finding] Cards whose fix LANDS on a governed surface but whose SYMPTOM is engineering are being routed to domain:devx — four in one round, one already caused a cross-lane dispatch #13593 一轮曾有四张同类卡，说明分诊产出混卡并非孤例。
③ 防 AI 犯错:D 的新卡首行写明「承哪一项、原卡留哪几项」，分诊首触补 domain:skills;拆错会被分诊当场看见(响亮)，而 A 下的错误只能靠复核兜底。
④ 创业阶段不扩散:D 只加一行协议、不加门禁、不加标签;B 每次多一轮往返，C 多一段双标记状态。
Prior rulings read: 拆卡跨域混卡 skills governed lane split → 6 hits ([finding] Cards whose fix LANDS on a governed surface but whose SYMPTOM is engineering are being routed to domain:devx — four in one round, one already caused a cross-lane dispatch #13593 ruling: a governed-surface fix is domain:skills; decision: which model serves the clause-② contract review — fable on every lane (ruling 5771798588), or fable only for the spec / skills lanes and governed surfaces with opus elsewhere? #23352/pm-dispatch SKILL.md 〈路径分叉〉: the under-tier route for a skills-face hunk is one route — an isolated review at the face's tier (#23352 ruling B′), not "hand it to the skills seat" beside it #23445 B′: an at-tier isolated review); ADR none; thread: none

After the ruling

D: this seat files item 1 of grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513 as its own card now, and dispatches items 2 and 3 as grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513 with Fixes #23513. The rule line lands as a skills-lane PR to references/execution-duties.md.
C: this seat claims grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513 for items 2 and 3 and sets pm:retriage for triage to split item 1.
A or B: as described in the table.
Until the ruling, this seat dispatches no part of grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513. The next p2 candidate goes first.

Decision-box entry: passes · `priority:p2` · `domain:skills` · `area:devpath` added · and this seat's own miss · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-15T01:58Z

Entry check on the card body. These are all present:

the maintainer's words, quoted;
the one-line question;
the governing text (SKILL.md 〈域车道〉, pm:retriage, execution-duties.md, [finding] Cards whose fix LANDS on a governed surface but whose SYMPTOM is engineering are being routed to domain:devx — four in one round, one already caused a cross-lane dispatch #13593, B′);
premises with re-checks;
options A–D × consequence × workload;
the four facets, with the token;
prior rulings, the recommendation (D), the fallback (C) and the post-ruling steps.

It passes. P1 holds on a fresh read: #23513 is pm:queue, domain:engine, unassigned, and its item 1 is in skills/**.

This seat's miss, recorded: the live case is this seat's output.

At grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513's first grade (6172504723) and its unlock (6173812728), this seat wrote that item 1 is "a governed skills/** PR on its own", but left the whole card domain:engine.
SKILL.md's anchor rule (「domain:* = 修复落地包的域」) and [finding] Cards whose fix LANDS on a governed surface but whose SYMPTOM is engineering are being routed to domain:devx — four in one round, one already caused a cross-lane dispatch #13593's precedent called for splitting item 1 to domain:skills at grading.
Pending the ruling, this seat does not split grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513 now, because option A would keep it whole. From this round on, a card graded with an off-lane item gets the split, or an explicit note naming the carrying lane, at first grade.

Labels added: priority:p2 + domain:skills + area:devpath. The ruling lands as a pm-dispatch skill line.

席位意见（增量）· director seat (summon #37 continuation 2, `session_01LqNgdPZB2Jnr5HpGbaqLeB`), read as batch #348 item 4, 2026-10-15T04:49Z

Premises re-run on origin/main 8d05456cb: #23513 is open, pm:queue, domain:engine, unassigned, its item 1 in skills/** (P1 holds); git grep "拆卡\|跨域" over the skill hits SKILL.md:111 (pm:retriage) and :173-176 (the cross-domain exception path for an unsplittable PR), nothing for a queued mixed card (P2 holds); execution-duties.md sits at its ceiling (183 of 183), so D's rule line needs the raise that #23606 B (same batch) authorizes by this ruling, else one deletion pays for it. Prior rulings: #13593 (a governed-surface fix is domain:skills), ruling B′ on #23352/#23445 (an at-tier isolated review of a skills-face hunk); check-prior-rulings --card 23579: 405 candidates, none decides who splits.

Facets: ① agree (one card, one lane; the seat holding the context splits at the moment the mix is found; the owning lane writes and reviews its governed text) · ② agree · ③ agree · ④ agree.

Recommend D, fallback C, with 口径 for the executor:

The split-off card is filed by the seat that found the mix, carries no domain:* label from the filer, and its first line reads carries item k of #N; #N keeps items …; triage stamps the domain at first touch and may regrade.
The parent keeps Fixes on its remaining items; the child is an ordinary card of its own lane (its own claim, its own PR).
The one rule line (execution-duties.md): a split-off off-lane card does not inherit the parent's domain; the derived sub-issue rule (「在飞卡衍生 sub-issue 承父卡域」) stays for sub-issues of a card's own lane.
The live case: the engine seat files grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513 item 1 now under this ruling and dispatches items 2 and 3 as grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513; triage's standing fix (split or a named carrying lane at first grade, 6182647505) stands.

Workload of the letter (the card's column): domain:skills · 1 line in execution-duties.md (Tier S) · one issue-create per split, no precedent PR. C = 0 lines, one pm:retriage round per case. A/B as the table.

自检：只看①选 D；②③④ 是否翻转：否。

Ruling: batch #348 item 4 · letter D · maintainer 「同意」 2026-10-15T05:02Z

Director seat, summon #37 (continuation 2), session_01LqNgdPZB2Jnr5HpGbaqLeB, on the card's analysis (recommendation D, fallback C; filed on the maintainer's 「这种情况该怎么处理立决策卡」 and 「如果你自己把 skills 相关的开发拆分出来立卡呢」), triage's entry 6182647505 and the seat's delta 6184671350; the maintainer's 「同意」 takes the recommendation.

What the ruling binds (letter D):

The seat that finds a queued card carrying an item whose fix lands in another lane splits that item out itself: one new card, filed with no domain:* label, whose first line reads carries item k of #N; #N keeps items …; triage stamps the domain at first touch and may regrade. No waiting on a triage round.
The parent keeps Fixes on its remaining items; the child is an ordinary card of its own lane (its own claim, its own PR, its own review tier by face).
One rule line in references/execution-duties.md: a split-off off-lane card does not inherit the parent's domain; the derived sub-issue rule (「在飞卡衍生 sub-issue 承父卡域」) keeps applying to sub-issues of the card's own lane.
The live case: the domain:engine seat files grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513's item 1 (the skills/objectstack-data/rules/security.md rule) as its own card now and dispatches items 2 and 3 as grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513; triage's standing fix at first grade (split, or a named carrying lane) stands.

Workload of the letter: domain:skills · 1 line in execution-duties.md (Tier S; the file sits at its ceiling, and under #23606 ruling B this ruling is the raise's authorization) · one issue-create per split.

Execution: this card is the execution card for the rule line: needs-user-decision removed and pm:queue added in this act for domain:skills; the body's first line carries Ruled:. A pointer to this ruling is posted on #23513 for the engine seat.

Claim: PM loop round 1
Session: session_01GjUZrCo7Yi58T4vPXqx88e
Account: os-sales (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-23579-split-off-card-domain
Worktree: objectstack-issue-23579
Domain: domain:skills
Seat: domain:skills#1
File surface: .claude/skills/pm-dispatch/references/execution-duties.md (one rule line beside the derived-sub-issue rule in 〈候选与批次〉) · scripts/pm/check-skill-line-ratchet.mjs (the CEILINGS row for that file, 183 → 183 plus the measured delta, under ruling B 6184799323 on #23606; ⛔ the header docblock and the refusal text are #23606's region, not this card's) (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: opus (TIER_DEFAULT) — dispatch-gates.mjs --tier --repo objectstack-ai/objectstack at c5dd7fce3 reads "no path-derived mandate: the surface hits none of the 3 declared glob(s)" for both paths; a one-line protocol rule is a judgment edit, so the default tier, not the floor; ceiling CONTRACT_REVIEW_TIER not mandated by any path; the governed-rules-text review face (.claude/**) is reviewed in seat at CONTRACT_REVIEW_TIER, the served tier equalling its value
Clause-②: no
Responsibility: n/a — not a defect card
Thread-read: 6184818831
Serial constraints cleared: execution-duties.md (183/183) and its CEILINGS row are shared with #23607 (queued, this lane): #23607 is serial behind this card and is dispatched only after this card's PR reads MERGED; check-skill-line-ratchet.mjs is shared with #23606 (queued, this lane) in a disjoint region (header docblock and refusal text vs the CEILINGS rows), and under the maintainer's 「并发保持1」 nothing else is in flight; PR #23655 (#23578, this lane, in its landing window) touches neither file
Ruling-ref: 6184818831

Claimed 2026-10-15T10:47Z with origin/main at c5dd7fce3 (the rule line absent there; the derived-sub-issue line present as the control). The dev leaves no second claim: this comment is its identity, and it checks that the newest Claim: here names its branch before any edit.

Dispatch order — posted 2026-10-15T10:50Z by the claiming seat (session_01GjUZrCo7Yi58T4vPXqx88e, os-sales); the same text is the dev's prompt, mode:subagent, model tier opus (TIER_DEFAULT).

Dispatch — issue #23579 (objectstack-ai/objectstack) · `domain:skills` · PM `session_01GjUZrCo7Yi58T4vPXqx88e` (os-sales)

You are the os-dev developer for exactly one card: objectstack-ai/objectstack issue #23579. Read the card's body and EVERY comment yourself (triage entry, the director's delta, the ruling, the seat's claim) and check the body is complete before you start. The PM's Claim: comment on the card (id 6190272501) is your identity: its Branch: is claude/issue-23579-split-off-card-domain, its worktree objectstack-issue-23579. Verify the newest Claim: on the card names that branch before any edit; on a mismatch stop and report. ⛔ Leave no second claim, ⛔ never write the card's assignee; the PR's assignee is os-sales (pr_create with assignees: ["os-sales"]). Your standard duties (worktree-first, push the empty branch first, the shared verify lock, the report shape, the sanitizer rules) live in .claude/agents/os-dev.md and AGENTS.md; this order carries only the increments. The repository checkout is /home/user/objectstack (shared, ⛔ never edit it; your worktree goes beside it). Run everything in the foreground; a round ends only with the report.

1. Ruling — not re-decidable (`Ruling-ref: 6184818831`, letter D, the maintainer's 「同意」 2026-10-15)

Quoted from the ruling of record on the card (comment 6184818831):

What the ruling binds (letter D):

The seat that finds a queued card carrying an item whose fix lands in another lane splits that item out itself: one new card, filed with no domain:* label, whose first line reads carries item k of #N; #N keeps items …; triage stamps the domain at first touch and may regrade. No waiting on a triage round.

The parent keeps Fixes on its remaining items; the child is an ordinary card of its own lane (its own claim, its own PR, its own review tier by face).

One rule line in references/execution-duties.md: a split-off off-lane card does not inherit the parent's domain; the derived sub-issue rule (「在飞卡衍生 sub-issue 承父卡域」) keeps applying to sub-issues of the card's own lane.

The live case: the domain:engine seat files grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513's item 1 (the skills/objectstack-data/rules/security.md rule) as its own card now and dispatches items 2 and 3 as grant name follow-ups after C7b S3 (#23430): skills security rule still says the retired grant id column is "legacy, do not write"; ceremony remedy text names a bare --apply that S4 now refuses without --plan-file #23513; triage's standing fix at first grade (split, or a named carrying lane) stands.

Workload of the letter: domain:skills · 1 line in execution-duties.md (Tier S; the file sits at its ceiling, and under #23606 ruling B this ruling is the raise's authorization) · one issue-create per split.

Item 4 is already done (#23615 exists, #23513 is closed by PR #23624); this card carries item 3 alone.

The raise authorization, quoted from ruling record 6184799323 on #23606 (letter B, the maintainer's 「同意」): 「1. A ruling of record that orders text into a ratcheted file is the authorization to raise that file's ceiling by the measured lines the text needs: the raising PR quotes the ruling comment's id and the measured line delta in its body, and moves the CEILINGS row in the same PR by at most that delta. No second word from the maintainer is needed.」 and 「5. In force from this ruling for every ruling of record, this batch's included (#23579's line in execution-duties.md, …): each such PR quotes its ruling and its delta.」

Deliverable: ONE draft PR to main, first body line Fixes #23579, editing exactly these surfaces:

.claude/skills/pm-dispatch/references/execution-duties.md — ONE rule line, placed beside the derived-sub-issue rules in 〈候选与批次〉 (today L40 「在飞卡衍生三分:不修它验收过不过;in-scope ⇒ 父卡 sub-issue,认领席自有、优先级继承。」 and L41 「该 sub-issue 带父卡域与优先级直接派发,唯一绕过分诊;sweep 事后扫,重复关 duplicate。」), in the file's register (Chinese, one executable sentence, no incident narrative, no issue number in operative text — a comment id such as 6184818831 is a ruling record, not an issue number, and may stand as provenance). It must carry items 1–3: a queued card found with an item whose fix lands in another lane ⇒ the finding seat splits it out itself; the new card is filed with no domain:*, first line carries item k of #N; #N keeps items …; triage stamps the domain at first touch and may regrade; the split-off card does ⛔ not inherit the parent's domain, while L41's inheritance keeps applying to sub-issues of the card's own lane; the parent keeps Fixes on its remaining items. ⛔ Do not reword L40–41 or any other line; ⛔ no table row (the file's width pin is 0).
scripts/pm/check-skill-line-ratchet.mjs — the CEILINGS row ['.claude/skills/pm-dispatch/references/execution-duties.md', 183] (today at line 1237) moves by exactly the measured delta of your edit (one line ⇒ 184), with a comment block above the row in the file's own convention (read the AGENTS.md 1116 → 1117 block near line 1556 for the shape: what the text is, which ruling ordered it, the maintainer's word verbatim and untranslated, the authorization under ruling record 6184799323 letter B, 「Landed count, headroom 0, same convention」). ⛔ Touch nothing else in that file: its header docblock and refusal text belong to [Decision] The line ratchet (check:pm-skill-ratchet): 32 of 37 files at zero headroom, so a sentence already ruled still needs a second ruling to land — keep it, let the ordering ruling carry its lines, add slack, or narrow it #23606, a sibling card not yet dispatched; its other rows belong to nobody.

2. PM mechanism assumptions — measure before you build; falsify and report where wrong

(a) PM readings at origin/main c5dd7fce3: execution-duties.md is 183 lines at ceiling 183 (headroom 0), width pin 0; no line on main names a split-off card (grep 不继承, 拆出, split-off over the file: 0 hits; L41 present as the control). The ratchet's self-test and pnpm check:pm-skill-ratchet read the row.
(b) The ratchet accepts an ordinary ruled raise as a bare number move plus comment (the AGENTS.md 1116 → 1117 precedent, PR docs(agents): a test file that declares a measured budget with its class is held to that budget, not the 10 s line #23596 today, 8d05456cb); no ruledRaises record is involved — that mechanism belongs to cross-file moves. Verify by running node scripts/pm/check-skill-line-ratchet.mjs and its --self-test on your head; if the self-test pins a map-wide total or refuses a bare raise, STOP and report blocked with the refusal text — ⛔ do not restructure the ratchet.
(c) The delta is measured, not assumed: your line count of the md on your head minus 183. If the sentence needs two lines, the row moves by two and the PR body says two; ⛔ do not delete another rule to stay at one, and ⛔ do not re-wrap neighbours.
(d) Same-day churn: scripts/pm/check-skill-line-ratchet.mjs merged a change today (8d05456cb, PR docs(agents): a test file that declares a measured budget with its class is held to that budget, not the 10 s line #23596). Work from the merged code; the snippet the card cites may have moved; check current main first.

3. PM suggested route — optional; measurement wins

A possible line, yours to improve: 「排队卡带落在他道的项 ⇒ 发现席自拆:新卡无 domain:*、首行 carries item k of #N; #N keeps items …,分诊首触定域可改判;拆出卡 ⛔ 不承父卡域(L41 只及本道 sub-issue),父卡余项照 Fixes。」 Place it after L41.
The PR body carries, verbatim: the ruling id 6184818831 and its letter, the measured delta (+1, or what you measured), and the authorization id 6184799323 letter B item 1 — the seat record and the queue guard's reader check that quote.

4. Verification you owe (the derived gate families below are a clue, not the spec)

Derived at c5dd7fce3 by node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack .claude/skills/pm-dispatch/references/execution-duties.md scripts/pm/check-skill-line-ratchet.mjs (36 commands). After your FINAL commit re-derive with no paths (node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack), run the local subset, capture each exit code before any pipe, reconcile with --ran, and quote git rev-parse --short HEAD of that run in the report and the PR body. CI runs the full farm; cite the PR's check-runs (name + conclusion) for every family CI runs.

node scripts/check-ci-filter-parity.mjs
node scripts/check-closing-keyword-parity.mjs
node scripts/check-closing-keyword-parity.mjs --self-test
node scripts/check-comment-mask-corpus.mjs
node scripts/check-declaration-mirrors.mjs
node scripts/check-declaration-mirrors.mjs --self-test
node scripts/check-scripts-symbol-anchors.mjs
node scripts/check-scripts-symbol-anchors.mjs --self-test
node scripts/check-self-test-wired.mjs
node scripts/check-self-test-wired.mjs --self-test
node scripts/check-self-test-workflow-commands.mjs
node scripts/check-self-test-workflow-commands.mjs --self-test
node scripts/check-whole-set-label-write.mjs
node scripts/check-whole-set-label-write.mjs --self-test
node scripts/pm/bare-root-worklist.mjs --self-test
node scripts/pm/check-harness-current.mjs --self-test
pnpm --filter @objectstack/lint run check:doc-formula-expressions
pnpm check:agent-test-spelling
pnpm check:bash32-floor
pnpm check:cli-command-ids
pnpm check:cross-package-test-inputs
pnpm check:doc-authoring
pnpm check:driver-memory-census
pnpm check:entry-guard
pnpm check:gitlink-declared
pnpm check:nul-bytes
pnpm check:parse-guard
pnpm check:pm-dispatch-gates
pnpm check:pm-governed-merges
pnpm check:pm-skill-id-lint
pnpm check:pm-skill-ratchet
pnpm check:pnpm-filter-targets
pnpm check:ratchet-remedy-authority
pnpm check:refd-timer-probe
pnpm check:skill-frame-sync
pnpm check:watch-hint-literal

Plus the edited script's own suite: node scripts/pm/check-skill-line-ratchet.mjs --self-test and the plain run node scripts/pm/check-skill-line-ratchet.mjs (its verdict lines for execution-duties.md are the reading the PR body quotes). Heavy runs go through bash scripts/pm/os-verify-lock.sh -c '…'; check:* gates do not. pnpm check:pm-dispatch-gates is a long battery: nohup … > LOG 2>&1 & then tail --pid, as os-dev.md says.

5. PR body and labels

First line Fixes #23579; then the line Clause-②: no at line start (copied from the claim; no arm).
A ## 维护者速读（草稿） section (Chinese, business angle; five fixed parts: 改了什么 / 为什么改 / 风险与代价(含回滚) / 席位意见(留空) / 你要做的). Quote ruling 6184818831 letter D and the raise authorization 6184799323 letter B with the measured delta.
The diff publishes nothing (.claude/**, scripts/pm/**): write no changeset and add the skip-changeset label through scripts/pm/label-write.mjs (labels only).
This PR is governed (.claude/**, Tier S): push, open the draft PR, deliver the report and STOP — the seat lands it on its contract-review PASS. ⛔ Never flip it out of draft, never arm auto-merge.
Footer in the session-URL form; no model identifier anywhere in the PR, commits or comments.

6. Writes

Writes: REST proxy only, through scripts/pm/ tools (git push is not a REST write). Budget: git push (as needed) · pr_create once (draft, assignees: ["os-sales"]) · label-write once (skip-changeset) · post-stamped --comment=23579 once (the os-dev-report comment). Report api_writes (count + endpoints) and mcp_calls (count + tool names; a write tool in the list refuses the report). Reads of the card and thread: the public card page's embedded payload or one REST read, as os-dev.md says.

7. Report

Deliver the JSON report twice as .claude/agents/os-dev.md specifies: first as a comment on #23579 whose first line is the bare text os-dev-report, then as your final message (the JSON alone, no surrounding prose). premise_still_valid: false with pr: null is a first-class delivery if the rule already stands on origin/main or the ratchet refuses the ruled raise.

8. The four-axis decision frame (verbatim copy of `.claude/skills/pm-dispatch/SKILL.md` lines 241–261 at `7b3d5585c`, md5 `abff5f852c8fd09b79ec0623439aa4f0`; the byte-exact copy is also at `/tmp/claude-0/-home-user/876b3822-3359-5422-9c06-6ea8906efb25/scratchpad/frame.md`) — use it only if a `needs_decision` arises; ⛔ keep no copy in the repo

每个方案必须沿四条固定评估轴分析,这是决策分析的核心原则,不是可选项:

实际业务需求 — 它服务的是真实存在的业务场景,
还是投机性能力面?判据要求实测(谁在写这个键、谁在读、
示例应用与真实部署的用法),「读起来像有用」不作数。这条轴会改变结论,不是陪衬。
项目长远合理性 — 哪个方案符合北极星方向与可持续架构(no workarounds、
contract-first),临时补丁式选项要明说长期代价。
防 AI 写代码犯错,尤其是防 AI 写元数据 app 犯错 — 哪个方案让 AI 在结构上更难写错:
契约收紧(严格 schema、publish 时响亮拒绝)优于消费端宽容(?? 回退、静默容错)——
宽容恰是 AI 批量犯错被掩盖的温床;声明即强制,绝不让 AI 声明一个运行时不兑现的能力。
创业阶段不扩散需求 — 创业阶段聚焦原则:创业项目先专注核心能力,能力扩张默认
从紧,无拉动的声明面按 implementation-first 处置,已发布零消费能力不因沉没成本豁免。
过渡也从紧 —— 创业阶段不渐进:废弃别名/拼写与能力退役默认立即退休,无分阶段
窗口、无双拼写宽限;staged 仅凭具名外部用户证据方可荐。新增门禁默认否:门禁是只减
不增的零件,例外只有维护者点名;荐新增门禁须引其点名原话,否则荐「不加」。

基本裁决原则:spec 声明 > 实现 > 文档面;声明而未兑现是实现缺口,补实现或退役,
⛔ 不在消费端收窄;未声明键随实现,文档随之。「声明即强制」只及 spec 与元数据契约、
运行时能力,⛔ 不及 AGENTS.md 这类给 agent 读的约定 —— 那些的读者是 agent,不是门禁。
推荐意见必须基于这四条轴给出理由;四轴冲突时如实呈现权衡,
交维护者拍板。

os-dev-report

{
  "issue": 23579,
  "status": "done",
  "branch": "claude/issue-23579-split-off-card-domain",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/23660",
  "session": "session_01GjUZrCo7Yi58T4vPXqx88e — mode:subagent, the dispatching PM's id (the commit's Claude-Session trailer)",
  "premise_still_valid": true,
  "summary": "Two rule lines inserted directly after the derived-sub-issue pair (L40-41) in execution-duties.md 〈候选与批次〉 carry ruling 6184818831 letter D items 1-3: 「- 排队卡含他车道项 ⇒ 发现席自拆:新卡无 `domain:*`,首行 `carries item k of #N; #N keeps items …`。」 and 「- 分诊首触定域可改判;拆出卡 ⛔ 不承父卡域(承域只及本车道 sub-issue),父卡余项照 `Fixes`。」 (120 B and 119 B; no other line touched, no table row). The CEILINGS row in scripts/pm/check-skill-line-ratchet.mjs moves 183 → 185 by the measured delta +2, with a comment block above it in the map convention citing ruling 6184818831 (letter D item 3, maintainer 「同意」) and authorization 6184799323 (letter B item 1); header docblock and refusal text untouched. Draft PR #23660 (Tier S, .claude/**), assignee os-sales, label skip-changeset added (labeler added documentation and size/s); PR body quotes both ruling ids, the letters and the +2 delta. Premises held: at worktree base 286eb9ef3 (both files byte-identical to the claim's c5dd7fce3) the md was 183 lines at ceiling 183, grep for 不继承|拆出|split-off|不承父卡域 gave 0 hits with L40-41 present, and the ratchet and its self-test accept a plain ruled raise.",
  "tests": "At head b609a9164: `node scripts/pm/check-skill-line-ratchet.mjs` exit 0, verdict lines 「✓ check-skill-line-ratchet: .claude/skills/pm-dispatch/references/execution-duties.md: widest table row is 0 bytes (pin 0; headroom 0).」 and 「✓ check-skill-line-ratchet: .claude/skills/pm-dispatch/references/execution-duties.md is 185 lines (ceiling 185; headroom 0).」; reverse reading with the md edited and the row still 183 (uncommitted tree): exit 1, the sole ✗ line 「execution-duties.md is 185 lines; the ratchet ceiling is 183.」 (so both new lines pass the 120-byte rule and the row move is load-bearing); `node scripts/pm/check-skill-line-ratchet.mjs --self-test` exit 0, 「157 cases pass」 (no map-wide total pinned, plain raise accepted, assumption (b) holds); base reading at 286eb9ef3: exit 0, 「is 183 lines (ceiling 183; headroom 0)」. One-line spelling of the ruled content measured at 235 B against the 120 B cap, so two lines. No ablation: the diff is protocol text plus a ceiling number, which os-dev.md excludes from ablation. No package touched, so no dependency-closure build (①) and no package test/typecheck (②) is owed.",
  "gates": "Derived after the final commit: `node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack` (no paths) at b609a9164 = 36 commands (the same 36 as the dispatch), exit 0. All run locally, exit codes captured before any pipe, all final codes 0. The commands: check-ci-filter-parity, check-closing-keyword-parity (+--self-test), check-comment-mask-corpus, check-declaration-mirrors (+--self-test), check-scripts-symbol-anchors (+--self-test), check-self-test-wired (+--self-test), check-self-test-workflow-commands (+--self-test), check-whole-set-label-write (+--self-test), pm/bare-root-worklist --self-test, pm/check-harness-current --self-test, check:agent-test-spelling, check:bash32-floor, check:cli-command-ids, check:cross-package-test-inputs, check:doc-authoring, check:driver-memory-census, check:entry-guard, check:gitlink-declared, check:nul-bytes, check:parse-guard, check:pm-governed-merges, check:pm-skill-id-lint (「87 file(s) clean」), check:pm-skill-ratchet, check:pnpm-filter-targets, check:ratchet-remedy-authority, check:refd-timer-probe, check:skill-frame-sync, check:watch-hint-literal. `pnpm --filter @objectstack/lint run check:doc-formula-expressions` first exited 3 (PREREQUISITE NOT MET: formula/lint not built). It was re-run after `bash scripts/pm/os-verify-lock.sh -c 'pnpm exec turbo run build --filter=@objectstack/formula --filter=@objectstack/lint --concurrency=2'` (VERDICT command-exit 0, turbo full cache hit) and exited 0. `pnpm check:pm-dispatch-gates` ran under nohup: fast tier 「1884 cases pass」, 266.8s, exit 0. `--ran` reconciliation: 「36 derived famil(ies) accounted for — 36 run, 0 NOT-MEASURED」, exit 0. Local-only families check:dev-prereqs and check:skill-frame-freshness are not owed (no packages/spec build, not the four-axis frame file). PR check-runs at b609a9164, read once at report time: completed success = filter, The card this PR closes must claim this branch, Part-of PR must not also close its card, No other open PR may claim the same single-writer path, No other open PR may claim the same issue, Dogfood Regression Gate, Check PR Size, Check Documentation Links, Auto Label. Skipped = Build Core, Build Docs, Console Pin Gate, Temporal Conformance (live PG + MySQL), Dogfood Verify CLI, Packed-tarball smoke (opt-in) x2, the matrix-shard Dogfood Regression Gate row, one Check PR Size / Check Changeset / Auto Label duplicate row each. in_progress = Lint · ESLint, Lint · repo & tooling gates, Lint · docs & code guards, Lint · PM dispatch-gates, Lint · CI, release & engine gates, Lint · scoped · (three lanes), Type Check · (workspace, source gates, debt ledger, consumer gates), Test Core 1/6 to 6/6, Governed Surface Queue Guard, Check Changeset. Not waited on, per os-dev.md.",
  "mcp_calls": "0 — no MCP GitHub tool was called.",
  "api_writes": "3 REST writes, each one fleet-write relay dispatch through scripts/pm (identity objectstack-fleet[bot]). (1) scripts/pm/fleet-write/dispatch.mjs pr_create: POST /repos/objectstack-ai/objectstack/dispatches; relay run 38724963070 executed POST /repos/objectstack-ai/objectstack/pulls (draft) and POST /repos/objectstack-ai/objectstack/issues/23660/assignees (os-sales); read-back 7090 B sent / 7090 B stored, identical. (2) scripts/pm/label-write.mjs --add skip-changeset: POST .../dispatches; relay run 38725071320 executed POST /repos/objectstack-ai/objectstack/issues/23660/labels; read-back MATCHES. (3) scripts/pm/post-stamped.mjs --comment=23579, this report: POST .../dispatches, which executes POST /repos/objectstack-ai/objectstack/issues/23579/comments. git push x2 (the empty branch, then b609a9164) are not REST writes. Reads: GET issues/23579 and its comments (the one card read), GET issues/comments/6184799323 (to verify the raise authorization quote), GET pulls/23660, GET commits/b609a9164/check-runs, and a git fetch of refs/pull/23655/head into a private ref (to confirm the sibling PR does not touch either file).",
  "deviations": [
    "The quick-read heading is `## 维护者速读(草稿)` with ASCII parentheses, as os-dev.md and SKILL.md L193 spell it. The dispatch spelled it with full-width parentheses; os-dev.md wins on conflict.",
    "The dispatch said \"ONE rule line\"; the ruled content measures 235 B as one line against the ratchet's 120 B per-line cap, so it landed as two bullet lines and the row moved +2. This is the measured delta that the dispatch's assumption (c) and authorization 6184799323 letter B item 1 allow; no neighbour was re-wrapped and no rule deleted.",
    "The worktree was cut from origin/main 286eb9ef3, not the claim's c5dd7fce3 (main had moved). Both edited files are byte-identical at the two commits (empty git diff), and origin/main was still 286eb9ef3 when the PR opened, so no merge from main was needed."
  ],
  "open_questions": [],
  "out_of_scope_findings": [
    "carrier: none (承接者:无) · observation, not one of the three filing classes: the 认领-section line 「余项需换道/拆分加 `pm:retriage`」 routes a partially landed card's remainder through triage, while the new lines cover a queued card found mixed before dispatch, so the two do not collide · noted in the PR's Acceptance notes only, not filed"
  ]
}

Generated by Claude Code

ACCEPT → PR #23660 comment 6190869174 (seat record, domain:skills#1, session_01GjUZrCo7Yi58T4vPXqx88e); the at-tier contract review of record is PR #23660 comment 6190695761 (PASS on head b609a9164). Tier S landing by the seat follows: pr_ready, automerge_enable, the queue. 2026-10-15T11:22Z

Landed — PR #23660 MERGED as 881eeff31 (squash; the queue's merge, auto_merge_enabled by the seat at 2026-10-15T11:24Z, added_to_merge_queue at 2026-10-15T11:25Z, merged at 2026-10-15T11:35Z) · seat domain:skills#1 · session_01GjUZrCo7Yi58T4vPXqx88e (os-sales) · 2026-10-15T11:37Z

Reading	Answer
`git ls-remote origin refs/heads/gh-readonly-queue/main/pr-23660-*`	0 refs (the queue branch is gone)
`git merge-base --is-ancestor 1ced0f82b origin/main` (control, a known main commit)	exit 0
`git merge-base --is-ancestor b609a9164 origin/main` (the PR head)	exit 1 — a squash merge; the head is not an ancestor by construction
`git log -1 origin/main -- .claude/skills/pm-dispatch/references/execution-duties.md`	`881eeff31 pm-dispatch execution-duties: a split-off off-lane card does not inherit the parent's domain; ceiling 183 → 185 by ruling (#23660)`
`git show origin/main:…/execution-duties.md` line count · the ratchet row on `main`	185 · `['.claude/skills/pm-dispatch/references/execution-duties.md', 185]`
This card	closed `completed` by `Fixes` at the merge
Lane inventory after the merge (`domain:skills`, open)	#23607 (`pm:queue`) · #23606 (`pm:dispatched`, this seat) · #23615 (`pm:dispatched`, seat 2, `os-elon-musk`) · #23665 (`pm:blocked`, new) · #22060, #17161 (`tracking`) · #19287 (seat 2) — no stray closure
`check-expected-skips.mjs --pr 23660`	OK — 11 skipped check-runs, every one in the roster, exit 0
Shared checkout	fast-forwarded to `881eeff31`; `check-harness-current.mjs` CURRENT; mirror branch pushed

State change: pm:dispatched removed in this act (domain:skills, area:devpath, priority:p2, tooling stay); the PR subscription ended at the merge. From this merge a queued card found carrying another lane's item is split out by the finding seat (no domain:* on the child, first line carries item k of #N; #N keeps items …), and the split-off card does not inherit the parent's domain.

Navigation Menu

Repository navigation

[Decision] A card triaged to one lane carries an item that lands in another (#23513: an engine card with a skills/** item) — who splits it, and when #23579

One-line question

Background

Governing text

Premises (each with a re-check)

Options

Recommendation

After the ruling

Related

Decision-box entry: passes · `priority:p2` · `domain:skills` · `area:devpath` added · and this seat's own miss · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-15T01:58Z

席位意见（增量）· director seat (summon #37 continuation 2, `session_01LqNgdPZB2Jnr5HpGbaqLeB`), read as batch #348 item 4, 2026-10-15T04:49Z

Dispatch — issue #23579 (objectstack-ai/objectstack) · `domain:skills` · PM `session_01GjUZrCo7Yi58T4vPXqx88e` (os-sales)

1. Ruling — not re-decidable (`Ruling-ref: 6184818831`, letter D, the maintainer's 「同意」 2026-10-15)

2. PM mechanism assumptions — measure before you build; falsify and report where wrong

3. PM suggested route — optional; measurement wins

4. Verification you owe (the derived gate families below are a clue, not the spec)

5. PR body and labels

6. Writes

7. Report

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[Decision] A card triaged to one lane carries an item that lands in another (#23513: an engine card with a skills/** item) — who splits it, and when #23579

Description

One-line question

Background

Governing text

Premises (each with a re-check)

Options

Recommendation

After the ruling

Related

Activity

objectstack-fleet commented on Oct 15, 2026

Decision-box entry: passes · priority:p2 · domain:skills · area:devpath added · and this seat's own miss · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-15T01:58Z

objectstack-fleet commented on Oct 15, 2026

席位意见（增量）· director seat (summon #37 continuation 2, session_01LqNgdPZB2Jnr5HpGbaqLeB), read as batch #348 item 4, 2026-10-15T04:49Z

objectstack-fleet commented on Oct 15, 2026

objectstack-fleet commented on Oct 15, 2026

objectstack-fleet commented on Oct 15, 2026

Dispatch — issue #23579 (objectstack-ai/objectstack) · domain:skills · PM session_01GjUZrCo7Yi58T4vPXqx88e (os-sales)

1. Ruling — not re-decidable (Ruling-ref: 6184818831, letter D, the maintainer's 「同意」 2026-10-15)

2. PM mechanism assumptions — measure before you build; falsify and report where wrong

3. PM suggested route — optional; measurement wins

4. Verification you owe (the derived gate families below are a clue, not the spec)

5. PR body and labels

6. Writes

7. Report

objectstack-fleet commented on Oct 15, 2026

objectstack-fleet commented on Oct 15, 2026

objectstack-fleet commented on Oct 15, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Decision-box entry: passes · `priority:p2` · `domain:skills` · `area:devpath` added · and this seat's own miss · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-15T01:58Z

席位意见（增量）· director seat (summon #37 continuation 2, `session_01LqNgdPZB2Jnr5HpGbaqLeB`), read as batch #348 item 4, 2026-10-15T04:49Z

Dispatch — issue #23579 (objectstack-ai/objectstack) · `domain:skills` · PM `session_01GjUZrCo7Yi58T4vPXqx88e` (os-sales)

1. Ruling — not re-decidable (`Ruling-ref: 6184818831`, letter D, the maintainer's 「同意」 2026-10-15)