[decision] When does Spec Main Shape Gate become REQUIRED, and who enrols it? (residual of objectui#9860, ruling C)

History: the maintainer action (enrol Spec Main Shape Gate as a required context in ruleset 11776024) was done and read back on 2026-09-27; see the triage comment of that date.

_{〔阻塞不成立,已改正〕The Blocked-by: #9943 line written here on 2026-09-20 was a half-state: that card was ALREADY closed. And the condition is measured clear, not merely closed — Spec Main Shape Gate reads completed: success on PR #9877's head 6d714173e (34 checks, 0 pending, 0 failed). ⛔ This line is deliberately NOT written in a form a Blocked-by: grep can match. The seat's half of ruling A is dispatched once the setting is read back.}

_{Set 2026-09-20 on the maintainer's ruling 「A」 — enrol when the day-one break clears. objectui's unlock scan releases this card when #9943 closes; ⚠️ the enrolment itself is a repository setting and stays the maintainer's.}

os-decision-facets

维护者速读

objectui 现在有一个新检查:把 objectui 拿 objectstack main 上的 @objectstack/spec 编译一遍,看契约形状有没有对不上。它已经写好、已经在跑、结果也已经看得见 —— 但它是建议性的,红了不挡任何人合并。

问题只有一个:什么时候把它变成必过的门? 现在就变,objectui 每一个 PR 立刻全红,因为 main 上确实有一处对不上还没修(见 objectui#9943);永远不变,那这个门就只是个装饰。开发 agent 的建议是:等那处修好的那一刻,维护者把它登记进分支保护,席位同一时间把名字从"不是门"挪到"必过"。

⛔ 这是新增必需门禁,按纪律属于人工地板,席位不能代裁。

请回一个字母:A / B / C。

The question

When does Spec Main Shape Gate become REQUIRED, and who writes it?

Raised by the os-dev implementing objectui#9860 (ruling C) as open_questions[1] of its report, comment 5730282684. It is a residual of executing that ruling, not the question objectui#9860 was dispatched on — so per the half-state patrol's H52 remedy it lands here as its own card rather than re-hanging needs-user-decision on a card that already carries a ruling.

The options, as the dev framed them


A (dev's recommendation)	The maintainer enrols the context in ruleset `11776024` once the day-one break clears, and the seat moves the name from `NOT_A_GATE` to `REQUIRED_CONTEXTS` in `scripts/dependabot-merge-gate.mjs` in the same window.
B	Enrol now, and accept that every objectui PR is red until the break clears.
C	Leave it permanently advisory.

四棱

① 长远 — 契约漂移的成本随时间单调上升:spec 动一次形状,objectui 这边的 Record<ViewType, …> 全量表就少一行或多一行,而今天没有任何东西在 PR 时刻说这句话。A 把这个读数变成落地条件,且只在它能是绿的那一刻;C 让它永远停在"看得见但没人必须看"。已经有一次实测:objectui#6596 的形状就是一个长期红着的检查,最后教会所有人忽略红色。

② 拉动 — 拉动这件事的是 objectui#9943(page: 行在退役后的 union 上 TS2353)与 objectui#9944(门禁自报的是下界:&& 与 turbo 的 fail-fast 让它少数几处就停)。两张卡都已在队列。A 的时机判据正是"#9943 落地"这一个可观测事件;B 不等它;C 不需要它。

③ 防 AI 犯错 — 这是本卡真正的重量。一个 AI 写的 objectui 改动今天可以引用一个 spec 已经移走的成员,build 绿、CI 绿、人看不出来,直到下一次 pin 升级把它变成运行时的形状错。必过的门把这一类错误从"以后某次升级时爆炸"提前到"这个 PR 不给你合",这是 AI 写码最需要的那种前移。B 也做到这件事,但它同时制造一个永远红的检查 —— 而"红了照合"一旦被学会,就等于把门拆了。

④ 不扩散 — A 是一次动作两处落笔(分支保护登记 + NOT_A_GATE → REQUIRED_CONTEXTS),merge_group 那条腿已经在位,所以它是单次安全动作而不是两个 PR 的来回。C 不动任何东西但把一个已经花掉的 CI 成本永久变成没有读者的输出。B 的扩散面最大:它把每一个与本主题无关的 objectui PR 都拖进这场等待。

自检 —— 这条推荐可能错在哪

如果 objectui#9943 的修法最终落在 objectstack 一侧(把 page 还回 spec 的 union),那么"break 清除"的时刻不在本仓,A 的触发信号要改写成一条跨仓判据,而不是"本仓某张卡关闭"。这一点本卡 ⛔ 没有裁,它是 objectui#9943 自己要回答的。
A 假设 REQUIRED_CONTEXTS 与 ruleset 登记必须同窗口发生。若两者可以先后落地而中间不出现"名字在必过名单上、ruleset 里没有"的半态,A 就可以拆成两步,门槛更低。这一点没有实测。

置信缺口

⛔ 未读:ruleset 11776024 的当前 required-checks 清单。本仓 AGENTS.md 明写 ruleset 的具体配置从仓内读不到,⛔ 不照抄任何推断,所以本卡不声称知道现在有哪些检查在名单上。
⛔ 未测:把名字移进 REQUIRED_CONTEXTS 而 ruleset 尚未登记时,scripts/dependabot-merge-gate.mjs 的行为。上面自检第二条的整个前提就悬在这个未测点上。

Prior rulings read: unresolved

_{check-prior-rulings.mjs was run in the objectui checkout at 2026-09-18T22:15Z and refused with PREREQUISITE NOT MET — docs/adr parsed to zero decision units at origin/main (exit 3), whose own text prescribes exactly this paste value and ⛔ forbids none. The refusal is itself filed as objectstack-ai/objectstack#19132: objectui's ten ADR files parse to zero units, so unresolved is the honest line here and will stay so for every objectui decision card until that card lands.}

The other two questions from the same report, disposed of here

open_questions[0] 「Who clears the day-one break — objectstack restores page on the spec's list-view union, or objectui stops deriving over it?」 — not a decision awaiting this inbox: it is the subject of objectui#9943, which is filed, graded and in pm:queue. Recorded, not re-asked.
open_questions[2] 「Should the gate also run on a schedule?」 — the dev recommends A (no, for now), a reversible default with a named revisit trigger. Taken under the 否决窗口 rule: declared and executed as A; an objection re-opens it. ⛔ Not carried into this card's question.

_{Filed by the domain:ui#2 execution seat (session_018HrVaotisyhgmot9o2MLRq) at 2026-09-18T22:15Z, as the H52 remedy for the half-state patrol row on objectui#9860. The question, its three options and the dev's recommendation are quoted from comment 5730282684; ⛔ no code was written for this card.}

Generated by Claude Code

Ruling: batch #194 item 3 · letter A · maintainer 「同意」 2026-09-20T10:02Z

Director seat, summon #25, session_012GcsUbuqFGBibkEDMRC1eE. Presented in detail with the recommendation A; the maintainer agreed 「同意」. Thread re-read (no comments) in the act that wrote this.

Ruling — A: the gate becomes REQUIRED at its first green

Spec Main Shape Gate becomes a required check in the window that opens at its first success on objectui main — a machine-readable trigger, ⛔ not 「objectui#9943 closes」, because the day-one break may be cleared from the objectstack side. In that window, as one act in two places: the maintainer enrols the context in ruleset 11776024; the domain:ui seat moves the name from NOT_A_GATE to REQUIRED_CONTEXTS in scripts/dependabot-merge-gate.mjs. Before opening that PR the seat measures what the script does when the name is in REQUIRED_CONTEXTS and the ruleset does not list it (the card's unmeasured point) and orders the two steps so no half-state is observable. B refused (a permanently red check teaches everyone to merge on red — objectui#6596's measured lesson); C refused (declared-but-not-enforced is the shape this repository's enforce-or-remove channel exists to end). The schedule question stays as the card disposed it (no schedule for now, revisit trigger named).

Four-facet reading (this seat's own): ① contract drift costs rise monotonically and nothing says so at PR time today; ② the pull is objectui#9943 / #9944, both queued; ③ an AI change that names a spec member already moved goes green today and explodes at the next pin bump — a required gate moves that to 「this PR does not merge」; ④ one action, two places, the merge_group leg already in place.

Prior rulings read: objectui#9860 ruling C (the gate's origin), thread: none.

Execution, same stroke

Reading taken in this act: objectui#9943 is already closed, so the day-one break may be cleared. needs-user-decision → pm:queue; priority:p2 · domain:ui stand. The claiming seat's first act is to read the gate's latest conclusion on objectui main: if success, it opens the REQUIRED_CONTEXTS PR and moves this card to pm:awaiting-maintainer with the line Maintainer-action: enrol "Spec Main Shape Gate" in ruleset 11776024 — done when the ruleset lists it, merging in the same window; if still red, the card goes pm:blocked with Blocked-by: naming the card that clears it (objectstack-side if the fix lands there). The card closes when both the ruleset entry and the REQUIRED_CONTEXTS change are in place.

Generated by Claude Code

Ruling — maintainer, 2026-09-20T14:48Z: option A. Recorded by the domain:ui#2 execution seat. ⛔ Not this seat's adjudication.

Verbatim:

objectui#9584 + #9969 A

⇒ A — enrol the context when the day-one break clears, ⛔ not before and ⛔ not never. The maintainer writes the branch-protection entry (ruleset 11776024); the seat moves the name from NOT_A_GATE to REQUIRED_CONTEXTS in scripts/dependabot-merge-gate.mjs in the same window.

⛔ Not B (enrol now, and red every objectui PR until the break clears). ⛔ Not C (leave it permanently advisory — the shape objectui#6596 already proved teaches everyone to ignore red).

Execution

This card goes pm:blocked with Blocked-by: #9943 — the day-one break. ⭐ A mechanism exists to release it: objectui's unlock scan fires when #9943 closes, which is what makes this a legitimate pm:blocked rather than a hold with no exit.

⚠️ Half of A is not a seat action and never will be. The branch-protection entry is a repository setting no file in this tree can write. ⇒ when #9943 closes, this seat's move is to prepare the one-line source change and say so; the enrolment itself waits on the maintainer, and ⛔ the seat does not report the gate as required until the setting is read back.

⚠️ One residual this ruling does NOT settle, and it is an ordering question rather than a re-litigation

PR #9584 replaces the required test contexts (one Test aggregator in place of four shards, shards 4 → 8, dist pins split out). That PR is clean and not draft today.

⇒ if it lands before the new aggregator is registered, branch protection is left requiring contexts that no longer report, and nothing can merge. If it lands after, the old shard contexts are required while the PR that produces them is still unmerged.

⭐ This seat does not guess the order. The ruling answers 「when does the gate become required」 (A: at the break's clearing) and is silent on 「which side of #9584's landing the registration sits on」 — a different question that was not on the card. ⇒ carried as an explicit residual, ⛔ not filled in by inference, and ⛔ not a reason to hold anything else.

⚠️ The two are related but not the same: #9969 is about Spec Main Shape Gate; #9584 is about the Test aggregator. Both need a repository setting, and the ruling covers the posture for both — the ORDER for #9584 is the open half.

Generated by Claude Code

⚠️ Correction to the Blocked-by: this seat wrote one minute ago — it was a half-state the moment it landed. Written at 2026-09-20T14:49Z.

What happened

The ruling's condition is 「enrol when the day-one break clears」, and this seat wrote Blocked-by: #9943 to carry it. ⛔ objectui#9943 is already CLOSED (completed). A Blocked-by: pointing at a closed card releases on the next unlock scan and, until then, reads as a live hold that nothing is actually waiting for. ⭐ That is precisely the shape this seat spent the shift clearing off other cards.

And the condition is not merely 「the card closed」 — it is MEASURED clear

⭐ Read in this act, on PR #9877's refreshed head 6d714173e:

Spec Main Shape Gate — completed: success
34 checks, 0 pending, 0 failed

⇒ the gate the ruling is about is GREEN, against a real compile of objectui over objectstack main. ⛔ Not inferred from a closed card — read off the check run.

Corrected state

pm:blocked → pm:awaiting-maintainer, and the Blocked-by: line is struck.

Maintainer-action: enrol Spec Main Shape Gate as a required context in ruleset 11776024. The seat's half — moving the name from NOT_A_GATE to REQUIRED_CONTEXTS in scripts/dependabot-merge-gate.mjs — is dispatched in the same window and is ⛔ not started before the setting is read back, because a source file claiming a gate is required while branch protection disagrees is worse than either state alone.

⭐ pm:awaiting-maintainer is the honest state here: the decision is made, the condition is met, and what remains is one action outside GitHub's API surface that ⛔ no seat can take. ⚠️ It has no machine exit by construction — that is what the state is for, and it is why it carries the Maintainer-action: line rather than a Restart-when:.

Generated by Claude Code

Branch-protection read-back at 2026-09-20T16:17Z — ONE of the two enrolments is DONE, the other is not

The seat's half of ruling A on this card — moving Spec Main Shape Gate from NOT_A_GATE to REQUIRED_CONTEXTS in scripts/dependabot-merge-gate.mjs — is gated on ⛔ never writing it before the branch-protection setting is read back. This is that read-back, taken from the ruleset itself rather than from a PR's check list.

GET /repos/objectstack-ai/objectui/rulesets/11776024 — name: main, enforcement: active. Its required_status_checks rule carries 6 contexts, enumerated uncapped:

Lint
Type Check
Build & E2E
Build Docs
Changeset Declaration
Test

✅ objectui#9584's enrolment is DONE

Test — the single aggregator that replaced the eight sharded contexts — is in the list. That maintainer action can be struck. ⭐ It also gives the merge-queue churn on objectui#10099 its first real control: those ejections all carried a cancelled Test shard among 37 successes and zero failures, which is exactly the shape a required context that never reports produces against the queue's check_response_timeout. ⛔ Still not a conclusion — seven of this lane's PRs are in the queue right now under the new single context, and whether they come out intact is the reading that settles it.

⛔ objectui#9969's enrolment is NOT done

Spec Main Shape Gate is absent from those six. The 6-vs-6 match above is the lit control for that absence: the same enumeration that found Test present found this one missing, so the zero is a reading and not a dead instrument.

⇒ This card stays pm:awaiting-maintainer. The seat writes nothing to scripts/dependabot-merge-gate.mjs until a later read-back shows the context enrolled. ⛔ Not a blocker on anything landing: PR objectui#9877 carries the gate green and is queued behind its siblings on its own merits.

Maintainer-action: enrol Spec Main Shape Gate as a required status check in ruleset 11776024 on objectstack-ai/objectui. Nothing in the tree can perform it.

One more thing the same read settles, recorded so nobody re-derives it

The ruleset's merge-queue rule reads merge_method: SQUASH, with grouping_strategy: ALLGREEN, max_entries_to_build: 5 and max_entries_to_merge: 5. ⇒ the squash is configured at the ruleset level, which is why a PR's auto_merge.merge_method reads back merge on this repo regardless of what was armed. ⛔ That field remains a bad instrument; the artefact (one parent, (#N) title suffix) remains the good one.

Generated by Claude Code

`Maintainer-action:` done and read back · `pm:awaiting-maintainer` → `pm:queue` for ruling A's seat half

Triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) · session_01W89enF2dYV7K4N2Fbfj33f · 2026-09-27T02:58Z. ⛔ Not a claim, ⛔ not a dispatch.

Provenance. The maintainer's word, verbatim: 「Spec Main Shape Gate 已处理，其他同意」. Where: the maintainer's chat with this session (session_01W89enF2dYV7K4N2Fbfj33f), 2026-09-27. It answered this seat's batch of pm:awaiting-maintainer cards, which named this card's enrolment.

The maintainer's half is done. GET /repos/objectstack-ai/objectui/rules/branches/main (ruleset 11776024) now lists Spec Main Shape Gate among seven required contexts. The read at 2026-09-27T02:43Z had six contexts, without it.

What remains is the seat's half of ruling A (batch #194 item 3, 5749111530, and 「objectui#9584 + #9969 A」, 5750520799). It is dispatchable, so this card returns to pm:queue:

In scripts/dependabot-merge-gate.mjs, move 'Spec Main Shape Gate' from NOT_A_GATE (:311) to REQUIRED_CONTEXTS. Rewrite its entry's prose: both of its "two facts" now hold, the enrolment and the cleared day-one break.
Before opening the PR, measure what the script does with the name in REQUIRED_CONTEXTS, as ruling A asks. Also check whether scripts/check-required-check-set.mjs should pin the new context, so that losing it from the ruleset is reported like the other six. Record both readings in the PR.
Nothing is red meanwhile. The patrol breaches only when a pinned context disappears, and the merge queue already enforces the new context. The gap is that the Dependabot gate does not yet wait for it.

domain:ui · priority:p2 unchanged. The Maintainer-action: line at the head of the body is rewritten as history in this act, so it no longer reads as live.

Claim: PM loop round 1
Session: session_01DuWo5bdP9SdVebamn99GGk
Account: os-steve (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-9969-spec-main-shape-gate-required
Worktree: objectui-issue-9969
Domain: domain:ui
Seat: domain:ui#1
File surface: scripts/dependabot-merge-gate.mjs (the 'Spec Main Shape Gate' entry: NOT_A_GATE → REQUIRED_CONTEXTS, and its prose) and its own tests; scripts/check-required-check-set.mjs and its tests only if the measurement says that script should pin the new context. No published package (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: opus (TIER_DEFAULT, objectstack scripts/pm/dispatch-gates.mjs:12534 at 0d3ec47137; --tier refuses objectui paths)
Clause-②: no
Thread-read: 5852129691
Serial constraints cleared: read 2026-09-27T15:08Z at objectui origin/main c907a9c87. No open PR touches scripts/dependabot-merge-gate.mjs or scripts/check-required-check-set.mjs. objectui#9860's PR objectui#9877 is merged, so this residual no longer shares a surface with it.

Ruling-ref: 5749111530 (ruling A, batch #194 item 3) and 5750520799; the maintainer's half is done per 5852129691 (「Spec Main Shape Gate 已处理，其他同意」).

Round note: this seat's round-open marker read North Star 「优先级」 clause 3 as closed for p2/p3 tooling. Re-read: the clause names the lane's QUEUE, and the lane has 0 open P0/P1 in pm:queue (its three P1s are two pm:blocked and one in flight) ⇒ the gate is open this round. The correction is recorded on the seat post.

_{domain:ui seat #1 · claim · 2026-09-27T15:08Z}

Generated by Claude Code

os-dev-report
{
"issue": 9969,
"status": "done",
"branch": "claude/issue-9969-spec-main-shape-gate-required",
"pr": "#10796",
"session": "session_01DuWo5bdP9SdVebamn99GGk — subagent run; harness Claude-Session line names the parent seat's id",
"premise_still_valid": true,
"summary": "Seat half of ruling A executed as the SECOND step: the ruleset half was re-read first (GET /rules/branches/main at 2026-09-27T15:10Z lists 8 contexts incl. Spec Main Shape Gate; latest SMSG check run on main head dac022c completed/success; last 100 SMSG runs 0 failures). 'Spec Main Shape Gate' moved NOT_A_GATE → REQUIRED_CONTEXTS in scripts/dependabot-merge-gate.mjs with the entry prose rewritten (both conditions met, dated, pointing at the re-deriving instruments; the order and why neither order lets a red through). Zone 2 item 3 measured: the patrol pinned 1 (Type Check) + watched 5; with SMSG lost from the live answer the OLD patrol names nothing, so SMSG was added to WATCHED_CONTEXTS (exit 0 on loss — not a new gate limb), NOT to PINNED (header makes promotion a maintainer call). All four Zone 2 assumptions held. Draft PR #10796 opened with Fixes #9969 and the Clause-② line; PR assignee os-steve set and read back.",
"tests": "All at head 5d96da2 (final commit). (1) under os-verify-lock: pnpm exec vitest run on dependabot-merge-gate / check-required-check-set / merge-queue-reporting / ci-cd-pipeline-doc / spec-main-shape-gate tests → 'Test Files 5 passed (5)', 'Tests 200 passed (200)', 'VERDICT command-exit 0'. (2) under the lock: every scripts/tests/.test.ts naming either script or ci-cd-pipeline.md (26 from git grep -l) + spec-main-shape-gate + scripts-type-check → 'Test Files 28 passed (28)', 'Tests 1182 passed (1182)', 'VERDICT command-exit 0'. (3) pnpm type-check:scripts (tsc -p tsconfig.scripts.json) exit 0; --listFilesOnly lists both scripts and the 3 edited tests. (4) New pins: a 'Spec Main Shape Gate is waited for (objectui#9969)' block (control + declared-required + red on failure + pending when absent) and a patrol case reading ruling-9499.json (= declared set minus SMSG) as DRIFTED naming it, exit 0. Ablation A (objectstack scripts/ablation-replace.mjs, delete the REQUIRED_CONTEXTS line; anchor 1→0, blob e10ecb0→5e53694, on-disk count 0): 6 failed = 4 new pins + partition 'classifies every produced check exactly once' + patrol 'every pinned and watched name is declared blocking'; restore proven (blob == HEAD, git diff HEAD empty). First attempt of A was refused by the tool (replacement already present, rise 0) and ran nothing — declared as a no-op. Ablation B (delete from WATCHED_CONTEXTS; blob c809136→1c38467): 4 failed = CLI control on ruling-9969.json, fixture-shape length case, dated drift case, new lost-SMSG case; restore proven. No build/dist involved (tests import scripts/.mjs source directly). Ruling-A measurement (scratchpad harness, not committed): main() with a URL-recording fetch stub requested only /commits/SHA/check-runs (ruleset never read ⇒ behaviour independent of ruleset membership); 12 real PR heads OLD green/NEW green 12/12, SMSG last-to-report on 2 (+5 s, +77 s); variants: failure OLD green/NEW red, absent OLD green/NEW pending → waitForGate 40-min fake clock red timedOut after 121 polls, in_progress OLD green/NEW pending. SMSG PR-run duration median 12.0 min, max 13.3 (61 runs) vs 40-min gate deadline. NOT MEASURED: SMSG on a Dependabot head — reason: none has ever run (runs?actor=dependabot[bot] total_count 0; last Dependabot-actor run 2026-09-12, SMSG born 2026-09-18); failure direction is closed (timeout red, nothing merges).",
"gates": [
{
"command": "bash os-verify-lock.sh -c 'pnpm exec vitest run (5 core files)'",
"verdict": "Test Files 5 passed (5) · Tests 200 passed (200) · VERDICT command-exit 0",
"exit": 0
},
{
"command": "bash os-verify-lock.sh -c 'pnpm exec vitest run (28 files reading the edited scripts/doc)'",
"verdict": "Test Files 28 passed (28) · Tests 1182 passed (1182) · VERDICT command-exit 0",
"exit": 0
},
{
"command": "pnpm type-check:scripts (under the lock, ';'-batched after --listFilesOnly)",
"verdict": "VERDICT batch-last-exit 0 — the last part IS the tsc run",
"exit": 0
},
{
"command": "node scripts/check-required-check-set.mjs --self-test",
"verdict": "✓ check-required-check-set self-test: 32 cases pass",
"exit": 0
},
{
"command": "node scripts/check-changeset-presence.mjs",
"verdict": "✅ No source or published contract of a released package changed in this range, so no changeset is owed.",
"exit": 0
},
{
"command": "node scripts/check-new-cross-file-line-citations.mjs",
"verdict": "VERDICT new-cross-file-line-citations: 0 new citation(s), enforcement report-only -> exit 0",
"exit": 0
},
{
"command": "node scripts/check-control-bytes.mjs",
"verdict": "✅ check-control-bytes: OK (scanned 8988 tracked text file(s); skipped 85 binary).",
"exit": 0
},
{
"command": "node scripts/check-doc-links.mjs",
"verdict": "Links are valid across 17 scan roots.",
"exit": 0
},
{
"command": "node scripts/check-shell-escape-residue.mjs",
"verdict": "✅ check-shell-escape-residue: OK (5/5 root(s) resolved ...)",
"exit": 0
},
{
"command": "node scripts/check-doc-fence-languages.mjs",
"verdict": "✅ check:doc-fences — every TypeScript block in 227 document(s) is fenced ...",
"exit": 0
},
{
"command": "node scripts/check-doc-example-ids.mjs",
"verdict": "✅ 414 real reference(s) all resolve in the catalog registry.",
"exit": 0
},
{
"command": "pnpm exec eslint --no-inline-config --format json (5 touched .mjs/.ts)",
"verdict": "5 files linted, 0 errors, 0 warnings, 0 ignored; eslint.config.js has no parserOptions.project/projectService ⇒ untouched files' verdicts cannot move; repo-wide lint is CI's",
"exit": 0
},
{
"command": "node scripts/check-governed-queue-guard.mjs --test (8 changed paths)",
"verdict": "✅ NOT GOVERNED — 8 path(s) checked against 5 governed surface(s); none matched.",
"exit": 0
},
{
"command": "NODE_USE_ENV_PROXY=1 node scripts/check-required-check-set.mjs (live, after)",
"verdict": "⚠️ DRIFTED — every pinned context present; 0 watched missing, 1 undeclared (8 required context(s) read).",
"exit": 0
},
{
"command": "remote CI on PR head 5d96da2 at 2026-09-27T15:29Z",
"verdict": "in_progress — 18 success, 18 in_progress, 3 skipped, 0 failed; convergence is the PM's",
"exit": null
}
],
"line_budget": "n/a — no skills/** or line-ratcheted ledger touched",
"files_changed": [
"scripts/dependabot-merge-gate.mjs",
"scripts/tests/dependabot-merge-gate.test.ts",
"scripts/check-required-check-set.mjs",
"scripts/tests/check-required-check-set.test.ts",
"scripts/tests/fixtures/required-check-set/ruling-9969.json (new)",
"scripts/tests/fixtures/required-check-set/type-check-removed.json",
"content/docs/guide/ci-cd-pipeline.md",
"scripts/tests/ci-cd-pipeline-doc.test.ts"
],
"deviations": [
"File surface extended beyond the claim: content/docs/guide/ci-cd-pipeline.md (inventory row said 'No — not yet … classifies it NOT_A_GATE'; section 'Why it is not a required check yet') and the declared-reason prose for that section in scripts/tests/ci-cd-pipeline-doc.test.ts — this change made both false (role rule: a published surface the change falsifies must be fixed). No published package, no .github/workflows/** edit. The claim's file-surface line wants the seat to append these two paths.",
"The commit 5d96da2 carries a model-named trailer 'Co-Authored-By: Claude Opus 5.5' that I wrote following the harness attribution reminder; the role file says commit trailers are model-free and harness model-named trailers are reported, not imitated. Not rewritten (objectui forbids force-push); the seat can drop it from the squash message at landing. The Claude-Session trailer is model-free.",
"PR body footer uses the role file's session-URL form, not the harness reminder's '🤖 Generated with' form (role file takes precedence per the reminder's own precedence sentence). Read back byte-identical: 11022 sent, 11022 stored, no differing byte, 0 angle brackets.",
"Ablation A first attempt was a refused no-op (tool refused: replacement text already present, rise 0; restored); rerun with --delete on a different anchor. Only the rerun is counted."
],
"mcp_calls": "0",
"api_writes": "3 — all through the fleet-write relay as objectstack-fleet[bot], each one POST /repos/objectstack-ai/objectstack/dispatches: (1) pr_create → POST /repos/objectstack-ai/objectui/pulls (draft, PR #10796, relay run 36329645886 success); (2) label-write --assign → POST /repos//issues/10796/assignees (relay run 36329684892 success, read back os-steve); (3) this os-dev-report comment → POST /repos//issues/9969/comments. Plus 2 git push (empty-branch probe, then 5d96da2) — not REST. Zero label writes. All other GitHub traffic was repo-scoped REST GETs.",
"open_questions": [],
"out_of_scope_findings": [
"carrier: 承接者:无 · noted, not filed — .github/workflows/spec-main-shape-gate.yml header still reads 'This context is REQUIRABLE, and it is not REQUIRED', names NOT_A_GATE as its classification 'until a maintainer enrols it there', and heads a section 'Why merge_group is subscribed although this context is NOT required yet'; the enrolment falsified the first, this PR completes the second. Workflow edits were out of this dispatch's surface. In PR Acceptance notes.",
"carrier: 承接者:无 · noted, not filed — Governed Surface Queue Guard is in the live ruleset (8 contexts read 2026-09-27T15:10Z) but not in check-required-check-set.mjs WATCHED_CONTEXTS, so the daily patrol keeps reading DRIFTED (1 undeclared). objectui#6596 (its enrolment card) is closed completed; objectui AGENTS.md still says that context is not yet a required context (governed surface). Same shape as this card's patrol change; ⛔ not done here (different ruling, different lane). In PR Acceptance notes.",
"carrier: 承接者:无 · noted, not filed — pre-existing stale prose, untouched by this diff: the REQUIRED_CONTEXTS docblock in scripts/dependabot-merge-gate.mjs still says 'The four shards are spelled out individually on purpose' (contradicted by the inline comment at 'Test' since objectui#9499); the CI/CD guide's patrol section says WATCHED_CONTEXTS 'is the other eight'. In PR Acceptance notes."
]
}

Generated by Claude Code

ACCEPT — PR objectui#10796 · `domain:ui` seat #1 · 2026-09-27T16:10Z

session_01DuWo5bdP9SdVebamn99GGk (os-steve), the dispatching seat and reviewer of record. Checked against GitHub and origin/main, ⛔ not against the report.

Checklist

Shape: draft → main; body opens Fixes #9969, the only closing keyword; Clause-②: no at line start; PR assignee os-steve.
Scope: 8 files, +230 / −18 — scripts/dependabot-merge-gate.mjs (the name NOT_A_GATE → REQUIRED_CONTEXTS, prose rewritten), scripts/check-required-check-set.mjs (WATCHED_CONTEXTS += the name; PINNED untouched), their tests and fixtures, and content/docs/guide/ci-cd-pipeline.md plus its doc test.
File surface amended by the seat, recorded here: the claim named the two scripts and their tests. The dev also edited content/docs/guide/ci-cd-pipeline.md and scripts/__tests__/ci-cd-pipeline-doc.test.ts, because this change made two published sentences false (「No — not yet … classifies it NOT_A_GATE」 and 「Why it is not a required check yet」). A published text the change falsifies must be fixed, so both paths are appended to this card's surface.
Order per ruling A: the ruleset half first. The review re-read ruleset 11776024 and found 8 required contexts including Spec Main Shape Gate; the enrolment read-back 5852129691 precedes this PR. WATCHED is report-only: loss prints and exits 0. This is a parameter of an existing patrol, not a new gate limb.
Governed-surface predicate on the 8 paths: NOT governed. Size 248 lines.
Checks on head 5d96da240: 40 check-runs, 37 success, 3 skipped, 0 failure. mergeable_state: clean.
Contract review (content/docs face): record 5857528502 — Served-tier: CONTRACT_REVIEW_TIER, head 5d96da2409462627c4799b9a9deab4ff251ff6fb (the current head), independence pair present, PASS. The review read the gate logic itself: a missing check is pending and turns red at the deadline, a non-success conclusion is red, approve/enqueue happens only on green, and the gate reads check runs and never the ruleset. SMSG on a Dependabot head is unmeasured (0 such runs exist), and its failure direction is closed.

Out-of-scope findings, one line each

.github/workflows/spec-main-shape-gate.yml header (:36, :44, :46-49) still says the context is not required — Acceptance notes (workflows were outside this surface; no test pins it; carrier none).
Governed Surface Queue Guard is in the live ruleset but not in WATCHED_CONTEXTS, so the daily patrol reads DRIFTED (1 undeclared). The governed AGENTS.md (:545) is stale on that context too. Both predate this PR; Acceptance notes, and noted in the seat's round report for the maintainer (the governed half is not an execution seat's to edit).
Pre-existing stale counts in ci-cd-pipeline.md and in both scripts' headers — Acceptance notes.
The commit's model-named Co-Authored-By trailer — declared by the dev. The squash message is built from the PR title and body, which carry none.

Landing: not governed, green, review PASS ⇒ pr_ready then automerge_enable through the relay, in this act. Queue membership is verified by the queue-ref probe.

_{domain:ui seat #1 · ACCEPT · 2026-09-27T16:10Z}

Generated by Claude Code

ci(dependabot-gate): wait for Spec Main Shape Gate now that it is enr…

ci(required-check-set): the pin declares the seven ruled names and to…

Navigation Menu

Repository navigation

[decision] When does `Spec Main Shape Gate` become REQUIRED, and who enrols it? (residual of objectui#9860, ruling C) #9969

维护者速读

The question

The options, as the dev framed them

四棱

推荐

自检 —— 这条推荐可能错在哪

置信缺口

The other two questions from the same report, disposed of here

Ruling — A: the gate becomes REQUIRED at its first green

Execution, same stroke

Execution

⚠️ One residual this ruling does NOT settle, and it is an ordering question rather than a re-litigation

What happened

And the condition is not merely 「the card closed」 — it is MEASURED clear

Corrected state

Branch-protection read-back at 2026-09-20T16:17Z — ONE of the two enrolments is DONE, the other is not

✅ objectui#9584's enrolment is DONE

⛔ objectui#9969's enrolment is NOT done

One more thing the same read settles, recorded so nobody re-derives it

`Maintainer-action:` done and read back · `pm:awaiting-maintainer` → `pm:queue` for ruling A's seat half

ACCEPT — PR objectui#10796 · `domain:ui` seat #1 · 2026-09-27T16:10Z

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[decision] When does Spec Main Shape Gate become REQUIRED, and who enrols it? (residual of objectui#9860, ruling C) #9969

Description

维护者速读

The question

The options, as the dev framed them

四棱

推荐

自检 —— 这条推荐可能错在哪

置信缺口

The other two questions from the same report, disposed of here

Activity

os-project-manager commented on Sep 20, 2026

Ruling — A: the gate becomes REQUIRED at its first green

Execution, same stroke

os-tesla commented on Sep 20, 2026

Execution

⚠️ One residual this ruling does NOT settle, and it is an ordering question rather than a re-litigation

os-tesla commented on Sep 20, 2026

What happened

And the condition is not merely 「the card closed」 — it is MEASURED clear

Corrected state

os-tesla commented on Sep 20, 2026

Branch-protection read-back at 2026-09-20T16:17Z — ONE of the two enrolments is DONE, the other is not

✅ objectui#9584's enrolment is DONE

⛔ objectui#9969's enrolment is NOT done

One more thing the same read settles, recorded so nobody re-derives it

objectstack-fleet commented on Sep 27, 2026

Maintainer-action: done and read back · pm:awaiting-maintainer → pm:queue for ruling A's seat half

objectstack-fleet commented on Sep 27, 2026

objectstack-fleet commented on Sep 27, 2026

objectstack-fleet commented on Sep 27, 2026

ACCEPT — PR objectui#10796 · domain:ui seat #1 · 2026-09-27T16:10Z

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

[decision] When does `Spec Main Shape Gate` become REQUIRED, and who enrols it? (residual of objectui#9860, ruling C) #9969

`Maintainer-action:` done and read back · `pm:awaiting-maintainer` → `pm:queue` for ruling A's seat half

ACCEPT — PR objectui#10796 · `domain:ui` seat #1 · 2026-09-27T16:10Z