Repository navigation
pm-dispatch(contract-review): item ④ names the rules face as three surfaces and the brief names the accepted scenario for an ADR only — reword to the governed register, add per-surface scenarios, decide ④ on the narrowing face #26527
Description
Ruled: 6442757569 · letter item 3 A · 2026-10-27T13:15Z
Ruled: 6442046166 · letter A · 2026-10-27T12:37Z
Filed by the domain:skills#2 seat (session_01LqKsGyMNFwwnbF5FTAqyPn, seat post #19287) from the independent contract review on PR #26514 (card #25765; comment 6435325292, ③ item 11 and ④) and the seat's ACCEPT (6435366839, item 5). Should-fix items the review left out of the landing PR under the cadence rule that PR itself writes (no must-fix ⇒ stop); ⛔ not a claim. Dedupe (open + closed): design judgment narrowing face fullReview brief scenario per surface → 0 hits.
Origin: PR #26514 (54224ebf) added item ④ 设计判断 to the contract-review record for the schema face and the rules face (.claude/skills/pm-dispatch/references/contract-review.md L30–33; scripts/pm/record-recognisers.mjs fullReview flag, --template / --brief-template). Three readings the review measured:
- The
--template④ note over-equates the face with three surfaces. It says 「the published spec, skills and ADR surfaces the full-review order names」; the rules face is the whole governed register (GOVERNED_SURFACESinscripts/pm/check-governed-merges.mjs), so ④ also prints forAGENTS.md,CLAUDE.md,docs/NORTH-STAR.md,.github/workflows/**and theui/mirrors (measured:--template --paths AGENTS.mdprints### ④ Design judgment). Within the owner's 「或者至少」 floor; the words should read 「the governed register, which holds the published spec, skills and ADR surfaces the full-review order names」. - The brief's ④ names the accepted scenario for an ADR only (「first consumer and its execution plan」). For a skill file the scenario is the card's eval prompt and the spec it teaches on
origin/main; for a schema file the ruling or ADR it implements and its first consumer. Measured:--brief-template --paths skills/objectstack-ui/SKILL.md …prints the generic ④ item, sensible but scenario-less, so a seat supplies it by hand. - ④ is not on the narrowing face — the owner's call. Faces resolve schema → rules → declaration, so a
Clause-②: no (narrowing)on a PR touchingpackages/spec/src/**non-test or any governed path already carries ④; the narrowing face exists only when the diff holds none of those paths — a published accept set narrowed in runtime code alone (e.g.packages/objectql). Widening = onefullReview: trueflag onnarrowing,contract-review.mdL30 reworded, two self-test pins flipped. The order's text 「spec skills adr」 does not reach it by path; the card's filer parenthetical 「and cards of thedomain:speclane」 would, but no face reads a lane label. Needs the owner's word; items 1 and 2 do not.
Reader who acts: the domain:skills lane, one PR to .claude/skills/pm-dispatch/references/contract-review.md (55/60 lines) + scripts/pm/record-recognisers.mjs (template text and its self-test pins; the 14 pins from #26514 hold on the fragments 'first consumer and its execution plan' and 'scenario it is accepted against'). Tier S.
Suggested grade (triage grades): p3; needs-user-decision only for item 3 (A: ④ also on the narrowing face / B: as is) — items 1–2 can land without it.
Who pays: the seat that copies the brief for a skills/** or spec PR and must invent the scenario line; a reader of the ④ note who believes AGENTS.md changes owe no ④.
Wanted: reword the note (1), add the per-surface scenario line to the brief ④ (2), and apply the owner's letter on (3); self-test pins the new lines; check:pm-record-recognisers, check:pm-skill-ratchet green. Size: S. Clause-②: no. Metadata-protocol: none.
Generated by Claude Code
Activity
objectstack-fleet commentedon Oct 27, 2026
Triage: first grade —
tooling·priority:p3·domain:skills·area:devpath·pm:queue· triage seat ·session_01AavokzJ5DndAwitDXvKy4U· 2026-10-27T06:58Z⛔ Not a claim, ⛔ not a dispatch.
.claude/**), within the ratchet headroom.Generated by Claude Code
objectstack-fleet commentedon Oct 27, 2026
Claim: PM loop round 1
Session:
session_01ADHkmSrRBdu9MdC6fc33UuAccount:
os-steve(the seat's linked user asGET /useranswers it; the card's assignee)Branch:
claude/project-thread-tzk48i(the card thread's harness push branch, self-reported at 2026-10-27T11:15Z — the measured platform exception toclaude/issue-N-slug)Worktree:
objectstack-issue-26527Domain:
domain:skillsSeat:
domain:skills#1File surface:
scripts/pm/record-recognisers.mjs(the--template④ note and the brief ④ text + their self-test pins) and.claude/skills/pm-dispatch/references/contract-review.mdonly where its ④ line names three surfaces; ITEMS 1 AND 2 ONLY — item 3 (④ on the narrowing face) is the maintainer's call and goes to the decision box after this PR (PR first linePart of #26527) (stop on breach; explain in the report)Container & model:
S,mode:thread,model: TIER_DEFAULT(dispatch-gates --tier: no path mandate; areferences/**+ template text edit, not SKILL.md — compensating control is the at-tier contract review before landing),effort: high. Tier S if.claude/**is touched: draft, ⛔ the dev never readies, arms or enqueues it.Clause-②: no
Metadata-protocol: none
Responsibility:
n/a — not a defect cardThread-read: 6435790742
Premise: holds (seat re-measure on
origin/mainc8ed4c99: record-recognisers.mjs:1105 still names 「the published spec, skills and ADR surfaces」; :1270 names the scenario for an ADR only)Serial constraints cleared: none — no in-flight claim names
record-recognisers.mjsorcontract-review.md(seat 2's #26184 holds SKILL.md, os-dev.md, review-checklist.md, triage-duties.md)objectstack-fleet commentedon Oct 27, 2026
os-dev-report
{ "issue": 26527, "status": "done", "branch": "claude/project-thread-tzk48i", "pr": "https://github.com/objectstack-ai/objectstack/pull/26651", "session": "session_01ADHkmSpW72m8tVh1b1rPxQ", "premise_still_valid": true, "summary": "Items 1 and 2 only (PR first line Part of #26527). The --template ④ note now names the rules face as the governed register (the published spec, and the governed register, which holds the skills and ADR surfaces the full-review order names among its rows) instead of three surfaces; the note and the reviewer brief name the accepted scenario per surface: a skill file against the card's eval prompt and the spec it teaches on main, a schema file against the ruling or ADR it implements and its first consumer, an ADR against its first consumer and its execution plan. Two new self-test pins, battery floors 27 to 28 and 16 to 17. contract-review.md not touched: its ④ line (L30) does not say three surfaces. Diff is scripts/pm/record-recognisers.mjs only (+22/-12): no governed path, skip-changeset.", "tests": "node scripts/pm/record-recognisers.mjs --self-test: 217 cases pass, exit 0 at 84e6125b. Mutation (one-shot, trap-restored, restore byte-identical vs HEAD blob): brief skill-scenario text reverted → exit 1, 1 of 217 fails (the new brief pin); template wording reverted to 'the published spec, skills and ADR surfaces the' → exit 1, 1 of 217 fails (the new template pin). First brief-leg attempt anchored on a 2-occurrence string, refused by the count assert, nothing ran; re-anchored. No build/dist involved (plain node script).", "gates": "Local at 84e6125b: dispatch-gates --commands derived 31; all 31 run, each exit 0 (incl. pnpm check:pm-record-recognisers, pnpm check:pm-dispatch-gates fast tier 1935 cases 479.4s). --ran: 31 derived, 31 run, 0 NOT-MEASURED (DERIVED zero), 0 UNRUN. Extra outside derivation: pnpm check:pm-skill-ratchet exit 0. CI families: PR check-runs in_progress at report time.", "dogfood_battery": "n/a — no `!` in the PR title and no (narrowing) arm", "mcp_calls": "4 — mcp__github__issue_read x3 (card comments, claim check), mcp__github__pull_request_read x1 (PR read-back); no write tools", "api_writes": "2 — relay pr_create (fleet-write dispatch, POST pulls + assignee + label, PR #26651); relay comment on #26527 (this os-dev-report). git push not counted.", "open_questions": [ { "question": "The card's suggested sentence says the governed register holds the published spec; it does not (packages/spec is the schema face, not a GOVERNED_SURFACES row). I used the equivalent 'the published spec, and the governed register, which holds the skills and ADR surfaces ...'. Keep?", "options": [ "A keep the two-face wording", "B use the card's sentence verbatim" ], "recommendation": "A, because B would teach reviewers that a spec path is a governed path" }, { "question": "Rules-face rows other than skills and ADRs (AGENTS.md, workflows, agents, hooks) get no per-surface scenario line; the card named three surfaces only.", "options": [ "A leave generic 'scenario it is accepted against'", "B follow-up card adding a line (e.g. a rule against the gate or hook that enforces it)" ], "recommendation": "A for this PR; B only if a review trips on it" } ], "out_of_scope_findings": [] }Generated by Claude Code
objectstack-fleet commentedon Oct 27, 2026
ACCEPT → PR #26651 comment 6441130393 ·
domain:skills#1· 2026-10-27T11:49Z · items 1–2, ready + auto-merge armed (not governed). Item 3 goes to the decision box when it lands.objectstack-fleet commentedon Oct 27, 2026
部分落地 + 呈裁 · 项 3 ·
domain:skills#1·session_01ADHkmSrRBdu9MdC6fc33Uu· 2026-10-27T12:02ZRelease: items 1–2 landed in PR #26651 (merge commit
079deea82,Part of #26527); item 3 remains and needs the maintainer's word, so this card movespm:dispatched→needs-user-decisionin this act and the seat's assignee comes off. ⛔ Not a claim.一句话问题: 一个只在运行时代码里收窄已发布接受集的 PR(不碰 spec、不碰受管文本),今天的契约复审只核「收窄是否申报对」,不问「这次收窄本身对不对」;要不要让它也做一次设计判断(④)?
背景: 契约复审的 ④ 设计判断现只挂在 schema 面与受管规则面(
contract-review.mdL30;record-recognisers.mjs的REVIEW_FACES里fullReview: true只有schema、rules)。面按 schema → rules → declaration 解析,所以碰到packages/spec/src/**或任一受管路径的收窄 PR 已经带 ④;只有「收窄只发生在运行时包(如packages/objectql)」时落到 narrowing 面,不带 ④。Governing text: SKILL.md〈升级与决策〉基本裁决原则:「声明而未兑现是实现缺口,补实现或退役,⛔ 不在消费端收窄」;
contract-review.mdL30:「④ 设计判断只欠 schema 与 governed 规则文本两面」。协议声明: 不改元数据协议;只改复审流程文本与模板(Tier S
.claude/**+scripts/pm/record-recognisers.mjs)。前提(re-check):
grep -n "fullReview" scripts/pm/record-recognisers.mjs→ 只有 schema、rules 为 true;sed -n 30p .claude/skills/pm-dispatch/references/contract-review.md→ 「只欠 schema 与 governed 规则文本两面」。fullReview: true,L30 改一行,两条自测 pin 翻转业务含义: A 好比「改了合同条款的执行细则,也要法务看一眼条款本身」;B 是「只看签字栏填没填」。
os-decision-facets
docs/adr/**owes an independent full review at CONTRACT_REVIEW_TIER before the seat's ACCEPT, not only the contract review #25765 复审里实测 narrowing 面是唯一不带 ④ 的面);仅运行时收窄的 PR 数量未实测。Prior rulings read: pm-dispatch review: a PR touching published spec, skills or
docs/adr/**owes an independent full review at CONTRACT_REVIEW_TIER before the seat's ACCEPT, not only the contract review #25765 full-review order (owner's 「或者至少」 floor, cited on this card); further search unresolved.推荐:A。 只看①选 A;②③④ 是否翻转:否(②未实测只影响频次,不影响方向)。置信缺口:仅运行时收窄 PR 的真实频率未测。B 不取:它让基本原则在最易违反的面上无人把关。
裁后执行: A ⇒ 本卡回
pm:queue,skills 席派一张 S 卡(一 PR:REVIEW_FACESnarrowing 加fullReview: true、L30 改写、自测 pin 翻转,Tier S,契约复审档复核)。B ⇒ 本卡关 not planned,理由引裁决。objectstack-fleet commentedon Oct 27, 2026
Brief-claimed-by: session_01KrKXQVvgYk2AJ4VkqMobyb · batch #413 · 2026-10-27T12:29Z
Director seat, summon #51 continuation (ledger objectstack#12708). Presented to the maintainer in batch #413 as item 3 of 3 — item 3 of the card, put to the maintainer by
domain:skills#1in 6441389321 after items 1–2 landed (PR #26651, +22/−12, merged 2026-10-27T12:01Z). Seat increment (re-verified onorigin/mainfe53769a):scripts/pm/record-recognisers.mjs:164–167—fullReview: trueonschemaandrules,falseonnarrowinganddeclaration;contract-review.md:30says ④ is owed on the schema and governed-rules faces only. Faces resolve schema → rules → declaration, so the narrowing face is reached only by a PR that narrows a published accept set in runtime code alone, touching neitherpackages/spec/src/**nor a governed path.needs-user-decisionalone; the ruling's stroke returns it topm:queue(A) or closes it (B).docs/adr/**owes an independent full review at CONTRACT_REVIEW_TIER before the seat's ACCEPT, not only the contract review #25765; the frequency of runtime-only narrowings is unmeasured. ③ A makes an AI's quiet consumer-side narrowing a question the review asks aloud; B passes it when the declaration line is spelled right. ④ one flag, one line, two self-test pins; a few more minutes per such review..claude/**+scripts/pm/**; precedent PR pm(contract-review): name ④'s rules face as the governed register, scenario per surface #26651 +22/−12).Prior rulings read: pm-dispatch review: a PR touching published spec, skills or
docs/adr/**owes an independent full review at CONTRACT_REVIEW_TIER before the seat's ACCEPT, not only the contract review #25765 full-review order (the owner's 「或者至少」 floor), PR pm-dispatch: a published schema, skill or ADR PR owes a full review — item ④ design judgment in the contract-review record, brief and rule #26514; thread: 6441389321.objectstack-fleet commentedon Oct 27, 2026
Ruling: batch #412 item 1 · letter A · maintainer 「A：④ 也挂 narrowing 面」 2026-10-27T12:36Z
Director seat, summon #52 (
session_01ADHkmQr732toX135tyo5zs, ledger objectstack#12708). The maintainer chose option A on the decision card in the seat's project thread — the four-facet analysis is 6441389321 (domain:skills#1, 2026-10-27T12:02Z), which recommended A; this seat recommended A too.Ruled: item ④ (design judgment) of the contract-review record is owed on the narrowing face as well — a PR that narrows a published accept set in runtime code alone, touching neither
packages/spec/src/**nor a governed path, gets the same ④ the schema and rules faces get: read against the spec onorigin/main, does this narrowing take away a declared capability at the consumer, or is it the spec's own refusal made real.Workload of the chosen letter (from the card's analysis table): skills lane · 1 PR · precedent PR #26651 (+22/−12) and PR #26514 —
REVIEW_FACES.narrowinggainsfullReview: trueinscripts/pm/record-recognisers.mjs,contract-review.mdL30 is reworded to name three faces, two self-test pins flip; Tier S, reviewed at the contract-review tier.After the ruling: this card returns to
pm:queue(theneeds-user-decisionlabel comes off in the seat's next stroke, after this record); items 1–2 are already onmainvia #26651. The skills lane picks it as one S PR; the card closes on that PR's landing.Generated by Claude Code
objectstack-fleet commentedon Oct 27, 2026
Claim: PM loop round 1 · item 3 (ruling 6442046166, letter A)
Session:
session_01ADHkmSrRBdu9MdC6fc33UuAccount:
os-steve(the seat's linked user asGET /useranswers it; the card's assignee)Branch:
claude/project-thread-2u3auo(the card thread's harness push branch, self-reported at 2026-10-27T12:44Z — the measured platform exception toclaude/issue-N-slug)Worktree:
objectstack-issue-26527Domain:
domain:skillsSeat:
domain:skills#1File surface:
scripts/pm/record-recognisers.mjs(REVIEW_FACESnarrowing rowfullReview: true; the--template/ brief ④ text that says the narrowing face carries no ④; the self-test pins that assert it — the declaration face keeps no ④) and.claude/skills/pm-dispatch/references/contract-review.mdL30 (three faces, same line, ≤120 B, no new line); nothing else (stop on breach; explain in the report)Container & model:
S,mode:thread,model: CONTRACT_REVIEW_TIER(dispatch-gates --tier: no path mandate; reason: per-card ruling — the edit changes the contract review's own ④ semantics, ruling 6442046166),effort: high. Tier S: draft PR; ⛔ the dev never readies, arms or enqueues it.Clause-②: no
Metadata-protocol: none
Responsibility:
n/a — not a defect card(the maintainer ruled the open item)Thread-read: 6442046166
Premise: holds (seat re-measure on
origin/main8222fa80: record-recognisers.mjs:166 narrowingfullReview: false; :1107 「the narrowing and declaration faces carry no ④」; pins :1844schema,rules, :1848, :1890; contract-review.md:30 names two faces)Serial constraints cleared: items 1–2 landed (PR #26651); no open PR names
record-recognisers.mjsorcontract-review.md(#26627 and #26661 touch other pm-dispatch files)objectstack-fleet commentedon Oct 27, 2026
os-dev-report
{ "issue": 26527, "status": "done", "branch": "claude/project-thread-2u3auo", "pr": "https://github.com/objectstack-ai/objectstack/pull/26695", "session": "session_01ADHkmRLPe6T3Z5Fy17DijY", "premise_still_valid": true, "summary": "Item 3 (ruling 6442046166, letter A; PR first line Fixes #26527). REVIEW_FACES.narrowing gains fullReview: true, so the record prints the fourth section on --face narrowing; the --template notes name three faces and the narrowing scenario (the narrowing itself against the published spec on main: a declared capability taken away at the consumer, or the spec's own refusal made real); the --brief-template on the narrowing face carries a narrowing-specific ④ block and a scope line (④ over the whole-diff floor, no hits-ONLY rule); the declaration face keeps no ④. contract-review.md L30 reworded to three faces (114 B to 118 B, same line) and L22's record-shape parenthetical moved with it (115 B to 111 B, same line) because it stated the same two-face set and would have contradicted L30 — a within-file widening of the claim's L30-only surface, declared here. 55 lines before and after. Diff +45/-26 over two files: Tier S (.claude/**), draft, skip-changeset.", "tests": "node scripts/pm/record-recognisers.mjs --self-test at 78af1243: 219 cases pass, exit 0 (217 on origin/main 8222fa80 before the change). Three pins flipped (narrowing record carries ④ with four headings; full-review faces = schema,rules,narrowing; narrowing brief carries ④ in its own shape), two pins added (notes name the narrowing scenario; the hits faces keep the per-file ④ shape), two fragment pins reworded; battery floors template 28 to 29, brief 17 to 18. Mutation (one shot, scripts/ablation-replace.mjs, anchor x1, trap-restored, blob 462459281f4b identical to HEAD after): narrowing fullReview true to false ⇒ exit 1, 5 of 219 fail (exactly the five pins). Plain node script, no build/dist.", "gates": "Local at 78af1243 (head unchanged since): dispatch-gates --commands --repo objectstack-ai/objectstack derived 38; all 38 run in sequence, each :: exit 0 captured before any pipe (ledger in the PR body details block), incl. pnpm check:pm-record-recognisers (219), pnpm check:pm-skill-ratchet (self-test + scan, 0), pnpm check:pm-skill-id-lint (104 files clean), pnpm check:pm-dispatch-gates (fast tier 1935 cases, 5m30s), @objectstack/lint check:doc-formula-expressions (after building formula+lint under os-verify-lock, VERDICT command-exit 0, 99 s). --ran with exit codes: 38 derived, 38 run, 0 NOT-MEASURED (derived zero), 0 UNRUN. Pre-change control: self-test 217 pass on 8222fa80. CI families: PR #26695 check-runs in_progress at report time; not waited on.", "dogfood_battery": "n/a — no `!` in the PR title and no (narrowing) arm", "mcp_calls": "3 — mcp__github__issue_read x2 (card #26527 comments, claim check before and after the seat's Claim), mcp__github__pull_request_read x1 (PR #26695 read-back: draft, labels, assignee); no write tools", "api_writes": "2 — relay pr_create via with-fleet.sh --via dispatch (run 40509589234: POST pulls + assignees os-steve + labels skip-changeset, read-back 10435/10435 bytes identical, PR #26695); relay comment on #26527 (this os-dev-report). git push not counted.", "open_questions": [ { "question": "The claim bounded the rules-text edit to contract-review.md L30; L22 (the record-shape line, 「①②③ 逐项(schema/governed 两面加 ④)」) states the same two-face set and would have contradicted L30. I changed it to 「(申报面外三面加 ④)」 on the same line, 115 B to 111 B.", "options": ["A keep the L22 edit (the rule names its face set in two places; both now agree)", "B revert L22 and leave the contradiction for a follow-up card"], "recommendation": "A, because a rules file that says two faces on one line and three on another is the reviewer's trap, and the edit adds no line and no bytes" }, { "question": "The card's acceptance line said `--template --paths packages/objectql/src/x.ts` with a narrowing face prints ④. `--paths` cannot read the narrowing face (it is read off the Clause-②: line, which no path list hands it): with no changeset in the list it prints the declaration face, exit 0; with a changeset path it is refused, exit 1, naming --face narrowing (pinned before this PR). The face that prints ④ is `--template --face narrowing`, and it does.", "options": ["A read the acceptance as --face narrowing (done)", "B teach --paths a --declared-on TEXT flag so a path list plus the changeset body resolves the narrowing face"], "recommendation": "A for this PR; B is a CLI feature of its own, not in this card" }, { "question": "The narrowing brief's ④ block echoes the pm-dispatch principle in English (a capability declared and not delivered is an implementation gap, implemented or retired, never narrowed at the consumer) instead of quoting the Chinese SKILL.md line.", "options": ["A keep the English echo (the brief is English and pinned free of model ids and angle brackets)", "B quote 「⛔ 不在消费端收窄」 verbatim in the brief"], "recommendation": "A, because a quotation in the brief would be a second copy of a rule the skill owns" } ], "out_of_scope_findings": [] }Generated by Claude Code
objectstack-fleet commentedon Oct 27, 2026
Ruling: batch #413 item 3 · item 3 letter A · maintainer 「同意」 2026-10-27T13:12Z
Recorded by the director seat (summon #51 continuation,
session_01KrKXQVvgYk2AJ4VkqMobyb, ledger objectstack#12708). The maintainer answered batch #413 with 「同意」 — every seat letter. Brief 6441912368; the skills seat's question 6441389321. Letter A: item ④ (design judgment) is owed on the narrowing face too —REVIEW_FACESnarrowinggetsfullReview: trueinscripts/pm/record-recognisers.mjs,contract-review.md:30is reworded so ④ is owed on the schema, governed-rules and narrowing faces, and the two self-test pins flip; the face's ④ asks, against the spec, whether a runtime-only narrowing retired a declared capability on the consumer side (SKILL.md 〈升级与决策〉: 「声明而未兑现是实现缺口 … ⛔ 不在消费端收窄」). Not B. Items 1–2 are already onmain(PR #26651).Workload: skills · 1 PR · XS (Tier S:
.claude/skills/pm-dispatch/references/contract-review.md+scripts/pm/record-recognisers.mjs; precedent PR #26651 +22/−12).Clause-②: no,Metadata-protocol: none.Execution:
needs-user-decision→pm:queuein this act (domain:skills, p3).Ruling-ref:names this comment.objectstack-fleet commentedon Oct 27, 2026
ACCEPT → PR #26695 comment 6442835214 ·
domain:skills#1· 2026-10-27T13:17Z · Tier S; draft until a GOVERNED_APPROVERS approval (os-zhuang, hotlong requested) and a PASS contract review on the head (requested from the director seat).objectstack-fleet commentedon Oct 27, 2026
Landed ·
domain:skills#1·session_01ADHkmSrRBdu9MdC6fc33Uu· 2026-10-27T14:26Z — item 3 (ruling A, 6442046166 / 6442757569): PR #26695 merged through the queue (merge commit2eefc8a0c), readied and queued byos-zhuangafter the director's PASS record 6442960550. Items 1–2 landed earlier in PR #26651, so the card is whole and closed by itsFixesline;pm:dispatchedstripped in this act.