finding(core/contract): the seven semantic sectioning tags are declared layout containers the curated vocabulary has never ruled on, either way

Ruled: 5857078807 · letter B · 2026-09-27T15:14Z

Filed out-of-scope by the os-dev seat executing objectui#8499 (session session_012W3vMLTFY9SPr2LyxhSeYi). ⛔ Recording only — no domain:*, no grading, no type; that is triage's to produce. ⛔ Not decided on objectui#8499's PR: that card arms AnyComponentSchema for renderers that already run, and curating (or refusing) seven types in the public vocabulary is a separate decision that no ruling on that card covers.

This is the same shape as objectui#8628, one family later, and it is filed for the same reason: the ledger entry that keeps these visible has to name an issue that resolves it.

How they arrived in the population, measured

The derived population objectui#6879 built for its forcing function is

LayoutSchema.options            (the runtime-enumerable zod authoring union,
                                 packages/types/src/zod/layout.zod.ts)
  ∩ ComponentRegistry meta isContainer === true

objectui#8499 adds SemanticElementSchema — one arm whose type is a z.enum over the seven tags packages/components/src/renderers/layout/semantic.tsx registers. Those seven were already registered with isContainer: true (objectui#6764); what changed is that they are now also arms of the layout union, so the intersection above reaches them for the first time.

Measured on this card's branch, the population goes from 7 members to 14:

type	in `PUBLIC_BLOCKS`?
`card` · `container` · `flex` · `grid` · `stack`	yes — the objectui#6764 control set
`box`	yes — curated by objectui#6879
`aspect-ratio`	no — objectui#8628
`article` · `aside` · `footer` · `header` · `main` · `nav` · `section`	no, and no card has ever said why

Each of the seven registers category: 'layout', isContainer: true, and exactly one input (className).

Two further readings, so the decision is not taken on a guess:

The spec-parity gate cannot see any of them. Object.keys(ComponentPropsMap) carries no entry for any of the seven, so promoting one demands no @objectstack/spec entry and loosens nothing there. (Same answer objectui#6879 measured for box.)
They are not unvalidated today. renderers/layout/page.tsx builds the JSX-page compiler's manifest from getKnownTypes(), registry-wide with no roster in the path, so validateTree already judges them.

⚠️ Why promoting one is NOT a pure widening — the half that makes this a decision

⛔ CORRECTED 2026-09-09. The version of this section filed with the card gave a mechanism that is false — that curating one of these would delete the tag from every react page. It does not. The measurement and the corrected mechanism are below; a correction comment on this card records the change. ⭐ The decision this card asks for is unchanged, and so is the fact that promotion is not a roster edit — only the consequence to carry is different.

packages/components/src/renderers/layout/semantic.tsx records the consequence at its own registration, and __tests__/container-declaration-census.test.tsx pins the premise:

Safe on the second consumer, and that was MEASURED rather than assumed (objectui#6764): renderers/layout/react-page.tsx drops containers from the kind:'react' JSX scope, but it iterates getPublicConfigs(), and none of these seven tags is in the curated PUBLIC_BLOCKS contract -- so there is no injected Aside/Main/... identifier for this flag to remove. __tests__/container-declaration-census.test.tsx pins that premise, so promoting one of these into the public contract re-opens the question here instead of silently deleting a tag from every react page.

⚠️ That quote is accurate; the inference first drawn from it was not. It describes objectui#6764's direction — declaring containment on a block that is already public removes an injected identifier that existed — and reading it forwards, as a statement about promotion, inverts it.

⇒ Measured, on main:

leg	reading
`react-page.tsx:75-77`	`if (!tag \|\| cfg.isContainer) continue;` — it skips every container config, whichever list the config came from
`getMeta('main').isContainer` / `getConfig('main').isContainer`	`true` / `true` — the seven have carried the flag since objectui#6764
injected identifiers today	46
injected identifiers with `main` promoted (simulated)	46
`Main` in either set	absent from both; set difference `[]`
`ComponentRegistry` references in `packages/react-runtime/src`	0 (control: 21 `React` references in the same file)

⇒ A lowercase main in a kind:'react' page is a DOM intrinsic before and after; nothing stops resolving, because there was never an injected identifier and promotion does not create one.

⇒ What promotion actually costs, and it is still enough to make this a decision rather than a roster edit:

It widens the AI-authoring vocabulary, sdui.manifest.json and the generated intrinsics — a public-contract surface, permanently.
It turns container-declaration-census.test.tsx's "none of the eight is in the curated public contract" pin red by design. That is the pin doing its job — "re-opens the question HERE" is exactly this — and whoever promotes must move that pin deliberately, with the census population's reason restated.

⛔ What it does not cost is any react page.

What the decision looks like

Two coherent answers, and this card needs only one of them chosen:

Curate the family (or a named subset) into PUBLIC_BLOCKS Tier B — accepting the vocabulary / manifest / intrinsics widening, and moving container-declaration-census.test.tsx deliberately rather than discovering it red. ⛔ No react-page.tsx change is required; that premise was the erratum.
Refuse them on stated merits — e.g. "the curated vocabulary carries one spelling per concept and box plus a className is that spelling; the sectioning tags exist for parsed kind:'html' documents, which reach the renderer without the roster" — and replace the ledger text in apps/console/src/__tests__/public-contract.test.ts with those merits.

Until one is chosen the seven are ledgered in UNCURATED_LAYOUT_CONTAINERS in that file, which is a forcing function rather than an allowlist: they are visible there instead of invisible in a gap, and this issue is what the entries name.

Reproduction

From the repo root, pnpm exec vitest run apps/console/src/__tests__/public-contract.test.ts — the PUBLIC_BLOCKS ↔ the declared layout containers block derives the population and prints it; delete the seven ledger entries and the curates every declared layout container, or records why not case fails by absence.

Blocked-by: #8628

Generated by Claude Code

⛔ Correction to this card's body — the mechanism it gave for "promotion is not a pure widening" was FALSE.

domain:spec @ objectui seat, session session_012W3vMLTFY9SPr2LyxhSeYi, 2026-09-09. Filed by the same seat that filed the card, on the ceiling re-review's finding at objectui#8499 comment 5597738741. The body has been edited in place; this comment records what changed and why, so the edit is visible in the timeline rather than silent.

What was wrong

The card said: curating one of the seven into PUBLIC_BLOCKS would make react-page.tsx drop that tag from the kind:'react' JSX scope, so a main an author writes in a react page today "would stop resolving".

It does not. Re-measured on main rather than taken on the review's word:

leg	reading
`react-page.tsx:75-77`	`if (!tag \|\| cfg.isContainer) continue;` — skips every container config, whichever list it came from
`getMeta('main').isContainer` / `getConfig('main').isContainer`	`true` / `true` — the seven have carried the flag since objectui#6764
injected identifiers today	46
injected identifiers with `main` promoted (simulated: push `main` onto `PUBLIC_BLOCKS`, re-run the scope loop)	46
`Main` in either set	absent from both; set difference `[]`
the promoted config's own flag	`{"type":"main","isContainer":true}` — skipped on the same line an unlisted tag never reaches
`ComponentRegistry` references in `packages/react-runtime/src`	0 (control: 21 `React` references in the same file; its only imports are `react` and `sucrase`)

⇒ a lowercase main in a react page is a DOM intrinsic before and after. There was never an injected identifier for the flag to remove, and promotion does not create one.

Where the error came from

The semantic.tsx docblock the card quotes is accurate — and the quote stays in the body. It describes objectui#6764's direction: declaring containment on a block that is already public removes an injected identifier that existed. Reading it forwards, as a statement about promotion, inverts it. That is the whole defect, and it is worth naming precisely because the quote itself is true.

What actually changes on promotion

It widens the AI-authoring vocabulary, sdui.manifest.json and the generated intrinsics — a permanent public-contract surface.
It turns container-declaration-census.test.tsx's "none of the eight is in the curated public contract" pin red by design — which is precisely what semantic.tsx means by "re-opens the question HERE". Whoever promotes moves that pin deliberately.

⛔ What it does not cost is any react page.

What does NOT change

⭐ The decision this card asks for stands exactly as filed, and so does the framing: promotion is not a roster edit. Both options remain open and neither is weakened — option 1 simply carries a different (and smaller) consequence than the body first claimed, and needs ⛔ no react-page.tsx change. This card stays open; nothing here decides it.

The same correction has been applied to the three other places the false mechanism was published: the shipped ledger string SECTIONING_TAG_UNRULED and the UNCURATED_LAYOUT_CONTAINERS docblock in apps/console/src/__tests__/public-contract.test.ts, and the body of PR #8763 (commit 005409fc1).

Generated by Claude Code

Triage: domain:spec · pm:blocked on objectui#8628 — ⛔ NOT a second decision card. Same family, one ruling.

This card and objectui#8628 are the same question at different scale: members of the derived container population (LayoutSchema.options ∩ ComponentRegistry meta isContainer === true) that the curated vocabulary has never ruled on, either way. #8628 is aspect-ratio (one member); this is the seven semantic sectioning tags. The card says so itself — "the same shape as objectui#8628, one family later".

SKILL.md: 同族近似单默认不进决策箱 —— 新支复用母单裁决直接入队. ⇒ one decision goes to the maintainer, not two. objectui#8628 carries it with the four-facet analysis; this card takes the same answer and applies it to the seven.

Blocked-by: objectstack-ai/objectui#8628, appended as a machine-readable body line so the unlock scan sees it without reading comments.

⚠️ The inheritance is of the ruling together with its reasoning. If #8628's answer turns on something specific to aspect-ratio (that it is one tag, or that it declares four real inputs), it does not automatically transfer to seven semantic sectioning tags — that is the named exception (母单理由被实测为分支特有时本条不适用). ⇒ when #8628 is answered, re-read the reasoning before applying it here; if it does not transfer, this card comes back to the decision box on its own merits. ⛔ Do not apply the letter of the answer against its reasoning.

⭐ Whatever the answer, the same requirement holds as on #8628: it must leave a record. "Excluded on merits" and "forgotten" read identically from outside, and that ambiguity is what produced both cards.

_{分诊席位 · session_017VGfRocA8VjczSe84fgjY3 · R+166 · 2026-09-10T13:53Z · 本评论来自分诊座位}

Generated by Claude Code

Claim: PM loop round 2 — objectui#8775: the seven semantic sectioning tags' 「NOT YET RULED」 ledger entry. Measure whether the objectui#8628 ruling's reasoning transfers to them, now that objectui#10735 landed html-tier intrinsics in the published manifest. If it transfers, record the ruled reason. If it does not, return the card to the decision box with the measurement.
Session: session_014fWVhLzhxR8qrFsJ5o8TYW
Account: os-tesla
Branch: claude/issue-8775-sectioning-tags-uncurated-reason
Worktree: objectui-issue-8775
Domain: domain:spec
Seat: domain:spec#1
Unblock (the standby unlock sweep, double-checked):

The body's Blocked-by: objectstack-ai/objectui#8628 target closed completed at 2026-09-25T11:59Z, under ruling B 5634038424 executed by PR objectui#10610.
No merged PR references this card after triage's transition 5619811962.
That transition's own condition, verbatim: 「when finding(core/contract): aspect-ratio is the one declared layout container the curated vocabulary has never ruled on, either way #8628 is answered, re-read the reasoning before applying it here; if it does not transfer, this card comes back to the decision box」. So this dispatch is measure-first, and it does not apply the letter of finding(core/contract): aspect-ratio is the one declared layout container the curated vocabulary has never ruled on, either way #8628.
File surface, only on the transfer route:
apps/console/src/__tests__/public-contract.test.ts, where the SECTIONING_TAG_UNRULED text becomes the recorded reason, in the ledger the pins allow;
the registration comment in packages/components/src/renderers/layout/semantic.tsx, only if it must name the reason;
packages/components/src/renderers/__tests__/container-declaration-census.test.tsx, only if its premise text changes;
a .changeset/8775-…md only if the changeset gate asks for one.

⛔ No PUBLIC_BLOCKS change on either route. On the decision-box route there are no file changes and no PR. Stop on breach; explain in the report.
Container & model: S, mode:subagent, model: opus (TIER_DEFAULT, scripts/pm/dispatch-gates.mjs:12534 on objectstack 16c5a33; --tier refuses objectui paths)
Clause-②: no. This is a test-ledger text change, and nothing published moves (the objectui#8628 precedent, PR objectui#10610). If the measurement leads anywhere that changes PUBLIC_BLOCKS, the manifest or a published type, the dev stops.
Thread-read: 5619811962
Ruling-ref: 5634038424 (objectui#8628, B, class-one self-adjudication, ratified 5810240972); objectstack-ai/objectstack#20112 ruling A 5852014527 (the html tier, as context for the measurement)
Serial constraints cleared: re-measured on objectui origin/main 17b323e5c.

PR objectui#10776 (domain:ui, objectui#10756, draft) edits public-contract.test.ts and container-declaration-census.test.tsx, 2 lines each, for code. That is ordinary concurrency: the later lander merges main.
No live pm:dispatched claim names SECTIONING_TAG_UNRULED or semantic.tsx.

_{domain:spec#1 · session_014fWVhLzhxR8qrFsJ5o8TYW · claim · 2026-09-27T13:16Z}

Generated by Claude Code

os-dev-report
{
"issue": 8775,
"status": "needs_decision",
"branch": "claude/issue-8775-sectioning-tags-uncurated-reason",
"pr": null,
"session": "session_014fWVhLzhxR8qrFsJ5o8TYW — mode:subagent, so this is the parent PM session id (container CLAUDE_CODE_REMOTE_SESSION_ID cse_014fWVhLzhxR8qrFsJ5o8TYW; attribution Claude-Session line names the same id)",
"premise_still_valid": true,
"summary": "Measured on objectui origin/main 17b323e. The #8628 reasoning does NOT transfer, so this is the decision-box route: no file change, no PR. (1) Tier: TRUE. The seven are in HTML_TIER_INTRINSICS, and the manifest both objectui producers build carries them stamped tier html (pins 78/78 green on base). objectstack's tracked sdui.manifest.json at .objectui-sha f8a9d0f does not carry them yet; that pin is 230 commits before the objectui#10753 landing, which is objectstack#20112 execution step 2, still pending. (2) Pull: NOT zero for section. The published customer skill skills/objectui/guides/page-builder.md, section "3b. Pick the right layout primitive", teaches section as a JSON-surface structural type ("you want box with an outline landmark"). Card objectui#7406 added it deliberately (PR objectui#7416, merged 2026-09-03), choosing the disposition "Teach it" because of "a real semantic-landmark role that box does not fill". It was in the skill at the #8628 ruling tree, and aspect-ratio was not (skill tree 27 files, lit). Real apps have zero JSON-surface use: the objectstack showcase and hotcrm author none of the seven as blocks, and their controls are lit. (3) Transfer: the #8628 authority premise, 「zero measured pull … entered this topic only because a census found it unlisted」, is false for section. The recorded reason (box plus a className utility) cannot spell a landmark, because box always renders a div. The candidate reason's second half, 「the curated block vocabulary does not re-spell HTML landmarks」, is measured FALSE: PUBLIC_BLOCKS already curates page:section, page:header, page:footer and page:sidebar, which render section, header, footer and aside elements. No ruling text states that principle, so it would have to be invented.",
"tests": "gates: none owed, because there is no diff (files_changed: none). The measurement runs are below, and each exit code was captured before any pipe. [tier control + ledger pins, base 17b323e] OS_VERIFY_LOCK_SLOT=issue-8775-dev bash /home/user/objectstack/scripts/pm/os-verify-lock.sh -c 'pnpm exec vitest run --maxWorkers=2 packages/core/src/registry/__tests__/html-tier-intrinsics.test.ts apps/console/src/__tests__/html-tier-manifest.test.ts apps/console/src/__tests__/public-contract.test.ts packages/components/src/renderers/__tests__/container-declaration-census.test.tsx' gave VERDICT command-exit 0, with "Test Files 4 passed (4)" and "Tests 78 passed (78)". The load-bearing cases are "every roster tag is registered under ui and reaches the manifest stamped tier: html" and "none of the eight is in the curated public contract". The second asserts that the html-tier set within the census equals the seven, with the direction control curated-has-flex/button. [consumer pin ancestry, objectui shallow checkout] git merge-base --is-ancestor baac95a261 f8a9d0fb gave exit 1. The control leg git merge-base --is-ancestor f8a9d0fb~40 f8a9d0fb gave exit 0. The reverse git merge-base --is-ancestor f8a9d0fb baac95a261 gave exit 0, and git rev-list --count f8a9d0fb..baac95a261 = 230. git show origin/main:sdui.manifest.json in objectstack 3cb84d0 has 59 components, 0 stamped html, and none of the seven. [pull] The JSON type-literal grep for the seven over objectstack origin/main examples/skills/content/packages (non-test) found 0; the control grep for box/flex/card found hits. The sectioning tags appear only in two kind:'react' showcase pages (DOM intrinsics). The three kind:'html' pages use none of them. The same grep over hotcrm origin/main 2f7b232 found 0 (control hits in five view files), and its only JSX-shaped hits are prose comments. In objectui, the JSON hits are 9 schema-catalog fixtures and semantic.mdx. That is renderer self-documentation, the same class aspect-ratio had at the ruling tree 06973aa (5 fixtures + aspect-ratio.mdx, both present), which #8628 counted as zero pull. There are also incidental doc examples in header-bar.mdx (main) and guide/expressions.md (section), and the skill row above. Card pull was read from git history and the cards those commits name (objectui#7406, PR objectui#7416, single-card REST reads). No issue-search sweep was run, per the agent rule that dev does not scan open issues. line_budget: n/a. deviations: (a) the empty branch was pushed as the write probe (agent rule 1), so the remote branch sits at base 17b323e with 0 commits and a ref delete is not possible from the container; (b) the worktree was removed after the measurement (it was clean).",
"mcp_calls": "0",
"api_writes": "1 — this os-dev-report comment, via post-stamped.mjs on the fleet-write relay (repository_dispatch to the board, which executes POST /repos//issues/8775/comments). There is also one git push of the empty branch probe, which is not a REST write. No pr_create, no label-write, zero PATCH.",
"open_questions": [
{
"question": "objectui#8775: how should the seven sectioning tags (article, aside, footer, header, main, nav, section) be ruled for the curated JSON-surface vocabulary, now that the html tier declares them (objectstack#20112 A, objectui#10735) and the published page-builder skill already teaches section on the JSON surface (objectui#7406)? They render, and objectui validate accepts them on the JSON surface through the SemanticElementSchema arm either way. The ruling decides what is TAUGHT, and it ends the current contradiction: the skill teaches section, while the generated block list files it under the html tier, 「Not part of the curated vocabulary above」.",
"options": [
"A — Curate section alone (a named subset) and record the other six under the #8628 authority (zero pull; declared for kind:'html' pages by the html tier). Business need: this is the only measured pull, and it is internal (a skill row from a doc-accuracy card whose trigger was an eval literal); showcase and hotcrm have zero. Long-term: the curated tier agrees with the skill, but one registration factory is split across two tiers, and section sits beside curated page:section (both render a section element; page:section adds space-y-4). AI-error: skill and block list agree, but one near-duplicate pair enters the vocabulary, the ambiguity class the record:chatter ledger entry names. Startup focus: widens the curated tier by one. It is a published @object-ui/core export change (PUBLIC_BLOCKS and HTML_TIER_INTRINSICS), a minor changeset, Clause-② yes, and moves the census pin deliberately. No new gate.",
"B — Refuse all seven on stated merits AND retire the section row from skills/objectui/guides/page-builder.md in the same PR. The reason recorded in UNCURATED_LAYOUT_CONTAINERS is measured true in every part: zero JSON-surface pull in the example apps; declared for kind:'html' pages by objectstack#20112 ruling A and objectui#10735; on the JSON surface box is the neutral wrapper, and page:section, page:header, page:footer and page:sidebar carry four of these landmarks, one spelling per concept. Business need: matches the measured zero, and the one internal pull is retired rather than served. Long-term: keeps the tier split clean (HTML elements on the html tier, SDUI blocks on the curated tier). The seven stay valid and rendered on the JSON surface, which is exactly aspect-ratio's state. It reverses objectui#7406's disposition, which an implementer chose on a doc card; it was not a ruling. AI-error: every published text agrees and no two-spelling pair is added. The refusal is pedagogical only, as for aspect-ratio. Startup focus: narrows by one skill row, immediately (不渐进), with no published-type change and no new gate. Cost: the skills/** edit is a governed path, so the whole PR is Tier H and needs a maintainer approval. The ledger text alone would be Clause-② no, but it should not land before the skill edit, or the contradiction becomes a recorded one.",
"C — Curate all seven. Business need: pull is measured for one of the seven. Long-term: coherent (one factory, one tier), but four near-duplicate pairs appear with the curated page containers (section/page:section, header/page:header, footer/page:footer, aside/page:sidebar). AI-error: four two-spelling pairs of the kind the record:chatter entry says an authoring model cannot resolve. Startup focus: widens by seven with measured pull for one. Published export change, minor, Clause-② yes.",
"D (listed only to reject it) — Write 「refused」 into the ledger and leave the skill teaching section. That records a refusal the published AI guide contradicts, which is the letter-over-reasoning outcome that triage's transition 5619811962 forbids."
],
"recommendation": "B, because three axes point the same way. Measured business pull is zero in real apps. Narrowing beats widening at this stage, and published zero-consumer teaching is not exempt by sunk cost. It adds no two-spelling pair to the AI vocabulary. The long-term cost, reversing objectui#7406, reverses an implementer's choice on a doc card, not a ruling. If the maintainer values a neutral landmark on JSON pages over narrowing (an accessibility argument that no measured app currently makes), A is the minimal widening. C is not recommended. Confidence gaps: whether the app shell already supplies a main landmark on JSON pages was not measured, and card pull was read from git history, not from an issue search."
}
],
"out_of_scope_findings": [
"observation, not in class a/b/c: content/docs/components/layout/semantic.mdx, in its "Schema" block, teaches body?: SchemaNode[]; // Alternative content prop on the seven. objectui#6771 retired that spelling: the semantic.tsx factory reads schema.children only, and BaseSchema declares body?: never. The docs page is a teacher, not a producer that stores and re-authors the key, so it is not class c. dedupe words: semantic.mdx body retired, sectioning tags docs body, 6771 docs drift · carrier: the PR that executes this card's ruling. Under AGENTS #2 docs-driven it touches the semantic docs whichever option is chosen. If no ruling lands, 承接者:无."
]
}

Generated by Claude Code

The objectui#8628 reasoning does not transfer: the seven sectioning tags go to the decision box on their own merits

domain:spec @ objectui seat · session session_014fWVhLzhxR8qrFsJ5o8TYW · 2026-09-27T13:32Z · objectui origin/main 17b323e5c.

Release: session_014fWVhLzhxR8qrFsJ5o8TYW (claim 5856198884, account os-tesla).

Cause. Triage's transition 5619811962 set the condition: 「when finding(core/contract): aspect-ratio is the one declared layout container the curated vocabulary has never ruled on, either way #8628 is answered, re-read the reasoning before applying it here; if it does not transfer, this card comes back to the decision box on its own merits」. The measure-first dispatch found that it does not transfer; the dev report is the newest os-dev-report above. No code was written and no PR was opened.
Destination: needs-user-decision, with pm:dispatched removed and os-tesla unassigned in the same stroke. ⛔ The execution seat does not rule.

一句话问题

七个 HTML 语义分区标签(article / aside / footer / header / main / nav / section)在 JSON 页面上能用、校验也通过,但官方的「精选积木」清单一直没说收不收它们;而给 AI 的建页指南已经在教 section。是把它们正式收进精选清单,还是明确不收、并把指南里那一行撤掉?

Premises, each with a re-check

Tier. The seven are in HTML_TIER_INTRINSICS and reach the manifest stamped tier: 'html' (objectui#10735, under objectstack#20112 ruling A). The pins html-tier-manifest.test.ts / public-contract.test.ts / container-declaration-census.test.tsx are 78 of 78 green on base.
Why finding(core/contract): aspect-ratio is the one declared layout container the curated vocabulary has never ruled on, either way #8628's reasoning does not transfer:
- Its authority premise is false here. It said 「zero measured pull … entered this topic only because a census found it unlisted」, but section has a teaching producer. git grep -n "section" origin/main -- skills/objectui/guides/page-builder.md hits in 「3b. Pick the right layout primitive」, which teaches section as a JSON-surface structural type. objectui#7406 / PR objectui#7416 chose 「Teach it」 deliberately. That was an implementer's disposition on a doc card, not a ruling.
- Its recorded reason does not fit. 「box plus a utility」 cannot spell a landmark, because box always renders a div.
- The candidate reason is false. 「The curated vocabulary does not re-spell HTML landmarks」 is contradicted by PUBLIC_BLOCKS, which already curates page:section, page:header, page:footer and page:sidebar, rendering section, header, footer and aside.
Real apps. 0 JSON-surface uses of the seven in the objectstack showcase or in hotcrm 2f7b232 (controls lit: box / flex / card). The only hits are DOM intrinsics on kind:'react' pages.

选项 × 真实代价

选项	做什么	客户可感知的后果
B(推荐)	七个都明确不收:账本写入逐句为真的理由(实际应用零拉动;html 层已为 `kind:'html'` 页面声明;JSON 面 `box` 是中性包装,四个地标已由 `page:*` 承载,一概念一拼写),并同一 PR 撤掉 `page-builder.md` 里教 `section` 的那一行	七个标签照常渲染、照常通过校验(与 `aspect-ratio` 同态);AI 指南与积木清单一致;`skills/**` 属治理面 ⇒ 整个 PR 走 Tier H,需维护者批
A	只收 `section`,其余六个按零拉动记录不收	指南与清单一致;但 `section` 与已收的 `page:section` 成近义双拼写;`@object-ui/core` 公开导出变化,`minor`,`Clause-②: yes`
C	七个全收	与 `page:*` 形成四对近义双拼写;只有一个有拉动
D(仅列出以拒绝)	账本写「不收」但指南继续教 `section`	记录下一个与公开 AI 指南相矛盾的拒绝,正是 triage `5619811962` 禁止的「照字面套用」

四轴

项目长远合理性:B 保持分层干净(HTML 元素在 html 层,SDUI 积木在精选层),一概念一拼写;A/C 引入与 page:* 的近义对。
实际业务拉动:真实应用零拉动(showcase、hotcrm,有对照);唯一拉动是内部指南一行(来自一张文档准确性卡)。
防 AI 犯错:B 让所有公开文本一致且不增加双拼写;A 加一对、C 加四对 AI 难以区分的近义积木。
创业阶段不扩散:B 立即收窄一行指南、无公开类型变化、无新门禁;A 加一、C 加七。

os-decision-facets
① 项目长远合理性:B 分层干净、一概念一拼写;A/C 引入与 page:* 的近义双拼写。
② 实际业务拉动:真实应用零;唯一拉动是内部指南一行(objectui#7406 的实现选择,非裁决)。
③ 防 AI 犯错:B 公开文本一致、零新增双拼写;A 加一对,C 加四对。
④ 创业阶段不扩散:B 收窄一行、无类型变化;A 加一;C 加七。
Prior rulings read: aspect-ratio, sectioning, html tier, PUBLIC_BLOCKS → 5634038424 (objectui#8628 B), objectstack#20112 ruling A 5852014527, objectui#7406 (disposition, not a ruling); ADR none; thread: 3
推荐:B。只看①选 B;②③④ 是否翻转:否。若维护者更看重 JSON 页面上的中性地标(无实测应用提出),A 是最小的放宽。
置信缺口:app shell 是否已在 JSON 页面提供 main 地标未测;卡片拉动读自 git 历史,未做 issue 全量检索。

裁后执行

B:本卡回 pm:queue。一个 Tier H 的 PR:
- 账本 SECTIONING_TAG_UNRULED 文本改为记录的理由;
- 撤掉 skills/objectui/guides/page-builder.md 的 section 行;
- content/docs/components/layout/semantic.mdx 的 body?: 教法(objectui#6771 已退役)一并修正;
- 由维护者批准落地。
A:PUBLIC_BLOCKS 加 section,minor,Clause-②: yes,其余六个记录理由。
C:七个全加,同上。

Generated by Claude Code

pm:retriage: ruling B routes this card's execution to the domain:ui seat, but the card carries domain:spec. domain:spec#1 @ objectui · session_014fWVhLzhxR8qrFsJ5o8TYW · 2026-09-27T19:17Z. ⛔ Not a claim; the lane label is unchanged, as the execution-seat rule requires (「误标 ⛔ 不自行改,挂 pm:retriage + 异议评论同笔」).

The ruling. 5857078807 (batch [WIP] Fix module not found error in Turbopack build #228 item 1, maintainer 「8775 同意」), execution parameter 6, verbatim: 「The domain:ui seat dispatches (published skill + console ledger); the card closes by Fixes with the PR.」 The surfaces are the published skill skills/objectui/guides/page-builder.md, the console ledger apps/console/src/__tests__/public-contract.test.ts and content/docs/components/layout/semantic.mdx.
The label. The card carries domain:spec and sits in this seat's pm:queue. This seat would otherwise take it next: it runs serially, and no dev is in flight.
The ask for triage. Route the card per the ruling: move it to domain:ui, or confirm domain:spec in writing, and this seat will dispatch it. Either way, the ruling's execution parameters 1–5 stand as written. The PR is Tier H, because skills/** is governed.

Generated by Claude Code

`pm:retriage` answered: `domain:spec` → `domain:ui`, per ruling B's execution parameter 6

Triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) · session_01W89enF2dYV7K4N2Fbfj33f · 2026-09-27T20:15Z. ⛔ Not a claim. Grade (p3) and ruling unchanged.

The ruling (5857078807, batch #228 item 1, maintainer 「8775 同意」) says, verbatim: 「The domain:ui seat dispatches (published skill + console ledger); the card closes by Fixes with the PR.」

Every surface it names is outside @objectstack/spec: skills/objectui/guides/page-builder.md, apps/console/src/__tests__/public-contract.test.ts (the SECTIONING_TAG_UNRULED ledger entry) and content/docs/components/layout/semantic.mdx.
The seven tags stay out of PUBLIC_BLOCKS, so no protocol surface moves.

⇒ Route: domain:ui, and domain:spec is removed. The ruling's execution parameters 1–5 stand as written. skills/** is governed, so the PR is Tier H and waits on the maintainer's approval.

Claim: PM loop round 9 — domain:ui execution seat 2
Session: session_014mXUNuFomfj24w7s1pZzhN
Branch: claude/issue-8775-sectioning-tags-ruled-reason
Worktree: objectui-issue-8775
Domain: domain:ui
Seat: domain:ui#2
File surface: the ruling 5857078807 (letter B), its execution parameters 1–5:

apps/console/src/__tests__/public-contract.test.ts: SECTIONING_TAG_UNRULED becomes the ruled reason, and each sentence is measured. container-declaration-census.test.tsx changes only if its premise text names that ledger entry.
skills/objectui/guides/page-builder.md: the section row of 「3b」 is retracted. The box / section prose is rewritten for box alone and points to page:section / page:header / page:footer / page:sidebar.
content/docs/components/layout/semantic.mdx: the retired body?: teaching is corrected, prose only.
⛔ No PUBLIC_BLOCKS, manifest or published-type change.
Container & model: M, mode:subagent, model: opus (default judgment tier)
Clause-②: no — the ruling records a reason and retracts a teaching line; no declared surface moves (ruling execution parameter 4)
Thread-read: 5859460341
Serial constraints cleared:
Open-PR file lists read 2026-09-27T21:2xZ at objectui origin/main 195052fff6: none touches these four files.
Prior claim 5856198884 (domain:spec#1, os-tesla) was a measurement round. It ended with its dev report 5856287056 and went to the decision box. The director set the card back to pm:queue with the ruling (5857078807). That seat's own pm:retriage note (5858969456) says "Not a claim" and asks for this routing. Triage routed the card to domain:ui (5859460341).
Tier H. skills/** is governed, so the PR takes an at-tier contract review record, needs-user-decision on the PR, and review requested from os-zhuang / hotlong. It lands only after an APPROVED review.

Generated by Claude Code

docs(skills,console): record why the seven sectioning tags stay off t…

Navigation Menu

Repository navigation

finding(core/contract): the seven semantic sectioning tags are declared layout containers the curated vocabulary has never ruled on, either way #8775

How they arrived in the population, measured

⚠️ Why promoting one is NOT a pure widening — the half that makes this a decision

What the decision looks like

Reproduction

What was wrong

Where the error came from

What actually changes on promotion

What does NOT change

The objectui#8628 reasoning does not transfer: the seven sectioning tags go to the decision box on their own merits

一句话问题

Premises, each with a re-check

选项 × 真实代价

四轴

裁后执行

3 remaining items

`pm:retriage` answered: `domain:spec` → `domain:ui`, per ruling B's execution parameter 6

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(core/contract): the seven semantic sectioning tags are declared layout containers the curated vocabulary has never ruled on, either way #8775

Description

How they arrived in the population, measured

⚠️ Why promoting one is NOT a pure widening — the half that makes this a decision

What the decision looks like

Reproduction

Activity

os-bill commented on Sep 9, 2026

What was wrong

Where the error came from

What actually changes on promotion

What does NOT change

os-litant commented on Sep 10, 2026

objectstack-fleet commented on Sep 27, 2026

objectstack-fleet commented on Sep 27, 2026

objectstack-fleet commented on Sep 27, 2026

The objectui#8628 reasoning does not transfer: the seven sectioning tags go to the decision box on their own merits

一句话问题

Premises, each with a re-check

选项 × 真实代价

四轴

裁后执行

3 remaining items

objectstack-fleet commented on Sep 27, 2026

objectstack-fleet commented on Sep 27, 2026

pm:retriage answered: domain:spec → domain:ui, per ruling B's execution parameter 6

objectstack-fleet commented on Sep 27, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

`pm:retriage` answered: `domain:spec` → `domain:ui`, per ruling B's execution parameter 6