ci(spec-main-shape-gate): drop the merge_group trigger now the context is out of the required set (objectui#12252)

[objectstack-fleet]

Fixes #12252

Drops the merge_group trigger from .github/workflows/spec-main-shape-gate.yml now that Spec Main Shape Gate is no longer a required context. The ruling is the maintainer's answer B on objectui#12240 (ACCEPT 6125070195); this is the one-trigger follow-up it left open, sequenced after the ruleset edit and after PR #12250 merged.

What changed

• .github/workflows/spec-main-shape-gate.yml: merge_group leaves on:. pull_request (the advisory signal a human reads), the hourly schedule on main (48 * * * *) and workflow_dispatch stay. The comments that reasoned about queue builds follow: the header's bullet on merge_group, the concurrency comment, the timeout-minutes comment, and the one header sentence that listed a queue build beside a pull request as a run of its own branch's YAML.
• scripts/__tests__/spec-main-shape-gate.test.ts: the pin "subscribes merge_group, so a maintainer can enrol it without stalling the queue" becomes its inverse, "does not subscribe merge_group - re-subscribe merge_group before re-enrolling the context". Same one-line comment fix as the workflow on the artifact-trust assertion.
• content/docs/guide/ci-cd-pipeline.md: only the sentences that said the gate runs on the queue: the inventory row's trigger cell, the section's trigger line, the artifact-trust sentence, the sentence naming what the test file pins, and the last bullet of "How it became a required check, and how it left the set".

Nothing the gate verifies changes. Not touched: the gate's steps, the anti-false-green marker, the schedule, the seven-name pin in scripts/check-required-check-set.mjs, scripts/dependabot-merge-gate.mjs, ci.yml, AGENTS.md, the queue guard. No overlap with the in-flight PRs on ci.yml and turbo.json (objectui#12222) or on AGENTS.md and the queue guard (objectui#12255); origin/main was merged at 10aba37d before this PR was opened and none of its files touch this surface.

The precondition, read live before the first commit

Ruleset 11776024, read at 2026-10-12T14:19:18Z with gh api repos/objectstack-ai/objectui/rulesets/11776024 --jq '[.rules[] | select(.type=="required_status_checks") | .parameters.required_status_checks[].context]':

["Lint","Type Check","Build & E2E","Build Docs","Changeset Declaration","Test","Governed Surface Queue Guard"]

Seven names; Spec Main Shape Gate is not listed, so dropping the trigger cannot stall a queue build. node scripts/check-required-check-set.mjs run bare could not take its own reading from this container (HTTP 401 through the agent proxy, exit 2, which the script says is not a pass). The same script fed the live gh api repos/objectstack-ai/objectui/rules/branches/main answer through --fixture read: INTACT - 1 pinned and 6 watched context(s) all present (7 required context(s) read), exit 0.

Verification (head 17033fd1)

• The six test files that read this workflow or its enumerations (spec-main-shape-gate, merge-queue-reporting, ci-cd-pipeline-doc, check-required-check-set, dependabot-merge-gate, check-merge-queue-head): 6 files, 252 passed, VERDICT command-exit 0 under the verify lock. The count equals the pre-change count, since one pin was rewritten in place.
• pnpm type-check:scripts: VERDICT command-exit 0; tsc -p tsconfig.scripts.json --listFilesOnly lists the edited test file.
• Workflow YAML parse: base had four triggers (merge_group, pull_request, schedule, workflow_dispatch), head has three (pull_request, schedule, workflow_dispatch); the parsed file minus on: (jobs, permissions, concurrency, name) is identical to base's; the cron and the pull_request branches are unchanged.
• check-doc-links, check-control-bytes, check-new-cross-file-line-citations (0 new), check-action-ref-convention: exit 0 each. Control-byte grep -P over the three edited files: no match. check-governed-queue-guard --test over the three paths: NOT GOVERNED. check-changeset-presence: no changeset owed (no released package source moved). Targeted eslint over the edited test file: 0 errors, 0 warnings.
• NOT MEASURED: repo-wide pnpm lint (CI owns it; the targeted run above is the only local reading) and actionlint (not installed; the YAML parse above stands in).

Ablation (each leg committed first, run through the ablation tool: anchor hit once, blob changed, restore proven by blob equal to HEAD and an empty git diff HEAD; no build or dist involved, root scripts)

• A. The name back in REQUIRED_CONTEXTS without the trigger: merge-queue-reporting.test.ts 1 failed, 17 passed. The failing assertion is "the merge_group floor derives itself from REQUIRED_CONTEXTS (finding(ci): MUST_SUBSCRIBE_MERGE_GROUP names 6 workflows but 8 unfiltered blocking checks exist — the two doc gates subscribe merge_group today with nothing asserting they keep doing so #6160) > requires merge_group of every workflow that produces a required context": These workflows produce a context REQUIRED_CONTEXTS declares blocking, but do not subscribe merge_group: - spec-main-shape-gate.yml (produces Spec Main Shape Gate), expected [ 'spec-main-shape-gate.yml' ] to deeply equal []. Direction observed: red, as predicted. The derived floor refuses the half-restored state on its own.
• B. The trigger restored, name still not required: spec-main-shape-gate.test.ts 1 failed, 35 passed; the new pin reds with its message, expected true to be false.
• C. Control, both restored: the floor passes (merge-queue-reporting.test.ts passed) and only the new pin reds (1 failed, 53 passed over the two files). So A's red is attributable to the missing trigger and nothing else.

NOT MEASURED, owed after landing

The card's last acceptance item, "three merge groups after landing carry no Spec Main Shape Gate check run", can only be read once this lands. It is the seat's reading, not a local one. On this PR itself, Spec Main Shape Gate keeps pull_request and should still post its verdict here.

维护者速读(草稿)

• 改了什么:Spec Main Shape Gate 这条 workflow 不再订阅合并队列(merge_group),只保留 PR 上的提示性运行、main 上每小时一次的完整运行和手动触发。同时改了对应的测试钉子和 CI 文档里「在队列上运行」的几句话。
• 为什么改:该检查已不在 ruleset 的必需名单里(上面读的是七个名字,没有它),队列上跑出来的结果没有人看,却每个合并组都占一台 runner。这是 objectui#12240 裁决 B 留下的后续一步,顺序是先合并、再改 ruleset、最后才去掉触发。
• 风险与代价(含回滚):风险很小,检查本身验证的内容一个字没动,PR 上仍然会报。唯一的坑是将来如果要把它重新设为必需检查,必须先把 merge_group 订阅回来再改 ruleset,否则队列会卡在 60 分钟超时;测试钉子和 merge-queue-reporting 的派生下限都会在顺序错了时变红提醒。回滚就是把 merge_group 那两行加回 on:,并把测试里的钉子翻回去。
• 席位意见:
• 你要做的:本 PR 是 draft,按 .github/workflows/** 的约定等你批准后再由席位入队。合并后请席位读三个合并组,确认其中没有 Spec Main Shape Gate 的检查运行。

---

Generated by Claude Code

[objectstack-fleet]

维护者速读（终稿）· PR #12262 · 席位已对照 diff 校正，2026-10-12T14:50Z

改了什么：Spec Main Shape Gate 不再订阅合并队列（merge_group），只保留 PR 上的提示性运行、main 每小时一次的完整运行和手动触发；对应的测试钉子改为"不订阅队列，再次设为必需前先订阅回来"，CI 指南里"在队列上运行"的几句话随之改。3 个文件，+64/−41；检查本身验证的内容一字未动。

为什么改：你改 ruleset 后它已不在必需名单（dev 开工前读到七个名字，没有它）；队列上跑出来的结论没人看，却每个合并组占一台 runner 13 分钟，一天约 440 runner 分钟（公共仓不计费，占的是并发容量）。这是 objectui#12240 裁决 B 的收尾一步，顺序是先合并、再改 ruleset、最后去触发。

风险与代价（含回滚）：风险很小。唯一的坑是将来若要把它重新设为必需，必须先把 merge_group 订阅回来再改 ruleset，否则队列会卡到 60 分钟超时——dev 用三组消融实测：顺序错了测试钉子和 merge-queue-reporting 的派生下限都会变红提醒。回滚：把 merge_group 加回 on: 并把钉子改回。

席位意见：建议批准。队列上少一个无人读的 13 分钟作业，两个方向的误操作都有测试拦着。

你要做的：Approve；席位随后入队（按两速制它排在 #12256 主红修复之后，除非你像 #12251 那样亲自入队）。合并后席位读三个合并组，确认没有 Spec Main Shape Gate 的检查运行。

[objectstack-fleet]

Type Check is red on this head (c819b74b) and the failure is not this PR's. Director seat, 2026-10-12T21:47Z, on the maintainer's word 「12262 红了」.

• What fails: the Type Check job (115018179844, run 38311534599) at Type-check repo-root vitest setup files: tsc -p tsconfig.vitest-setup.json cannot resolve @object-ui/components, @object-ui/fields, @object-ui/plugin-dashboard, @object-ui/plugin-grid and @object-ui/i18n/locales (TS2882 / TS2307).
• Why it is not this PR's: this diff touches no workspace package (a workflow, a scripts/__tests__ file and a doc), so the scoped type-check that objectui#12222 (PR ci(type-check): the turbo cache replays unchanged packages, and pull-request and queue runs type-check only what the change affects (objectui#12222) #12251, merged 14:41Z) introduced selects zero packages, builds no dist/, and the root tsc step that follows has nothing to resolve. Before ci(type-check): the turbo cache replays unchanged packages, and pull-request and queue runs type-check only what the change affects (objectui#12222) #12251 the whole workspace was type-checked and built first; this PR's previous head 17033fd1, based before ci(type-check): the turbo cache replays unchanged packages, and pull-request and queue runs type-check only what the change affects (objectui#12222) #12251, had Type Check green. main is not red (push runs take the full scope). Every pull request without a package change fails the same way until the job is fixed.
• The fix: none exists yet to port. Filed and dispatched as objectui#12272 (a regression from this seat's own lane, priority:p1). When it lands, the seat merges origin/main into this branch and this PR's run goes green; the auto-merge the maintainer armed then takes it. No re-run of this job is useful before that (the failure is deterministic), so none is spent. ⛔ Nothing is pushed to this branch for it.