[pm:seat] domain:ui — objectui execution seat · 🟢 session_01CSoz9uGhaaSgiq3hshtN7L · R38 dispatched 12:45Z 08-24 · in flight 5 · 1 rework · 5 awaiting a human

[os-sales]

Seat registration post for the domain:ui lane (objectui execution seat), created under the maintainer's 2026-08-21 ruling that split objectui cards three ways (domain:devx / domain:spec / domain:ui, the last being the only new label).

Body is authoritative; title and assignee are derived views. Single writer: the sitting seat PM. Comments are audit only and never carry state.

---

🟢 SEAT OCCUPIED — R38 dispatched 2026-08-24T12:45Z, running

⛔ THIS IS NOT A ROUND-BOUNDARY LINE. R38 is mid-flight with 5 cards out. The mutex anchor for a next session is the R38 dispatch cluster at 12:42Z–12:45Z, not a close-out.

• Sitting PM: session_01CSoz9uGhaaSgiq3hshtN7L, took the seat 09:50Z 08-24 from a clean VACANT. Held continuously through R35 → R36 → R37 → R38.
• In flight: 5 (the full batch). See §3.
• ⛔ The round boundary is not a stopping point. Maintainer, 2026-08-24, verbatim: 「怎么空了？继续啊，待人类决策的整成清单，不要停掉活」. 收工点 = in-flight zero ⇒ dispatch the next wave immediately. ⛔ Do not end a round on a report and wait for a human gate. Items needing a human go on the §6 list and the loop keeps running.

⚠️ RUN THE FOURTH MUTEX READ. The three prescribed readings are structurally blind to a session that does not use pm:dispatched — its cards go pm:queue → closed. Also read the lane's most recently CLOSED card and check its provenance. It fired in R35: three readings clear, then an eight-card write cluster at 08:08Z, 38 minutes after the previous seat closed. It resolved to the triage Routine's hourly unlock scan, not a competing seat — but the three prescribed readings could not tell those apart, which is the whole point. Cost: one list_issues + one issue_read. Filed as objectstack#11604.

Corroborate with liveness, not absence: git ls-remote --heads origin 'refs/heads/claude/issue-*' and the open-PR list.

---

1. Current PM

• session_01CSoz9uGhaaSgiq3hshtN7L, seat domain:ui @ objectstack-ai/objectui (execution seat).
• Batch size: 5 — maintainer, 2026-08-23 (「任务很多，并发加到5」), reaffirmed 2026-08-24 (「并发数加到5个吧，默认好像是3个最多」). The skill's own batch:3 default is superseded.
• Default tier is Opus — 「Fable 5 配额已耗尽，可以用 Opus 继续，并发还是保持 5」. A quota exhaustion is a tier fact, never a density fact. ⛔ Do not cut density on resource grounds.
• Dispatch autonomously — 「你应该自主派发」. Only genuine manual-floor items go up.
• This seat may dispatch above its own served tier — 「你虽然是 opus，但是可以派 fable 的卡」.
• Clause-② at-tier dispatch is the route here — 「直接入队」. ⚠️ needs:contract-review does not exist in this repo, so the compensating-label route is unavailable by construction. Clause-② cards landed under the recorded exemption so far: 文档教作者写的 ActionSchema.condition = { expression, then, else } 全仓零消费者:谓词被忽略、then/else 永不派发、零诊断 #3917, Console/app-shell Sentry: a live DSN with sendDefaultPii: true is committed in apps/console/.env.production and baked into every production build — the VITE_SENTRY_ENABLED switch is build-time and absent from the inlined env, so a shipped bundle cannot be silenced #5522, ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293, and 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927 is out now. See §5 for what makes that safe.

2. Mutex lesson (standing, do not compact away)

See the ⚠️ block above. A stale seat post plus an unused status label is otherwise indistinguishable from a free seat.

3. In flight (R38) and the hot-file serial queue

Dispatched 12:42Z–12:45Z, all claimed atomically (assign + pm:queue → pm:dispatched, then claim comment, then race re-read — all three re-reads clean).

card	surface	size	Clause-②	state
#5631	packages/types/layout + renderers/basic (~29 files)	L	yes	🟡 running, carried from R37
#5293	packages/plugin-view — prose-only REWORK on PR #6008	S	yes	🟡 running
#5824	packages/types/src/__tests__/phase2-schemas.test.ts	S	no	🟡 running
#4730	packages/i18n/src/locales/** (console.objectView.* only)	M	no	🟡 running
#5927	packages/types/src/zod/{navigation,objectql,form,data-display,views}.zod.ts + zod-mirror-parity.test.ts	L	yes	🟡 running

Serial fences currently held:

• ⚠️ packages/types/** carries three concurrent cards (ui:icon reads the SDUI identity key name as its lucide icon name, so any icon node that authors name renders nothing at all #5631, finding: phase2-schemas.test.ts imports 9 schemas no test uses — unused-import warnings overstate the file's pin coverage #5824, 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927) on disjoint files. Each dispatch names the other two and forbids straying. ⛔ Do not add a fourth packages/types card until one lands.
• ⚠️ finding(types): TableColumn.type disagrees three ways — interface declares 8 literals, zod mirror accepts any string, renderer's live read set handles values outside the union via an as any cast #5853 is NOT in flight — this seat escalated it back to needs-user-decision (see §6). The serialization that held 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927 behind it is therefore cleared, but 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927's dispatch forbids touching what finding(types): TableColumn.type disagrees three ways — interface declares 8 literals, zod mirror accepts any string, renderer's live read set handles values outside the union via an as any cast #5853 covers and forbids treating its escalated ruling as settled law. When finding(types): TableColumn.type disagrees three ways — interface declares 8 literals, zod mirror accepts any string, renderer's live read set handles values outside the union via an as any cast #5853 is re-ruled, it serializes behind 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927 (data-display overlap).
• ⚠️ burn-down: 119 ledgered DOM-attribute leaks in packages/components/src/renderers/**, grouped by mechanism #5632 must keep excluding ui:icon's row until ui:icon reads the SDUI identity key name as its lucide icon name, so any icon node that authors name renders nothing at all #5631 lands.
• ⚠️ Reverse i18n sweep: ~384 confirmed-dead locale keys across 58 namespaces (workflow/BPMN designer entirely unwired, console.objectView oversized, more) #4730 holds the locale packs. Any card adding or renaming locale keys serializes behind it.

Pre-registered for R39 — read the comments before dispatching any of these:

• A third teaching site still writes the retired condition: { expression, then } branch shape — guide/schema-overview.md, outside #3917's fence #5980 — ⛔ still not dispatchable: it carries NO labels. domain:* is triage's single-producer output, and 未打标签的 issue 任何人不得认领. An earlier body of this post promised it as "front of R36"; that promise was wrong and is withdrawn. It needs a triage first touch, not a PM decision.
• 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927 Groups B/C — the dispatch requires the dev to return eight per-key renderer readings as one new decision-inbox issue. That issue is R39 input to the maintainer, not to this seat.
• Published skill objectui/guides/data-integration.md teaches a QueryParams shape the adapters silently drop — unprefixed keys plus a direction sort key #6006 — published skill skills/objectui/guides/data-integration.md teaches a QueryParams shape the adapters silently drop (direction spelling + unprefixed query options vs $select/$filter/$orderby/$top), with an unbounded-read consequence on a dropped limit. ⚠️ skills/** is a governed surface and a customer-published package under the maintainer's 2026-08-21 expansion-budget ruling — dispatching it needs budget attached and stops at draft.
• The Reverse i18n sweep: ~384 confirmed-dead locale keys across 58 namespaces (workflow/BPMN designer entirely unwired, console.objectView oversized, more) #4730 key-level trims — search (9, namespace measured partially live), table (10), report (10), form (8), appDesigner (20/34, heavy indirect-reference tier). Explicitly held out of the current Reverse i18n sweep: ~384 confirmed-dead locale keys across 58 namespaces (workflow/BPMN designer entirely unwired, console.objectView oversized, more) #4730 dispatch. The umbrella stays open for them.
• The toSortItems finding (filed by the ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293 rework): a dormant tolerant dual-read on a published export, packages/plugin-view/src/config/view-config-utils.ts:328. Removing it is its own Clause-② break needing its own adjudication.

4. Landing ledger

⛔ Counting restarted at the 04:00Z takeover; the predecessor's 138/134 cannot be extended across the gap.

round	window	verdicts	landed
R34	04:08Z → 07:23Z	—	3 cards closed / 4 PRs merged
R35	09:50Z → 10:43Z	5/5 ACCEPT	2 cards closed / 4 PRs merged / 1 awaiting a human
R36	10:43Z → ~11:40Z	5/5 ACCEPT	all merged
R37	~11:40Z → 12:45Z	3 ACCEPT · 1 REWORK · 1 ESCALATE	#6005 merged, #6003 merged, #6007 enqueued
R38	12:42Z → running	—	—

R37 detail:

• The withdrawn dist/index.d.ts byte-identity sentence survives verbatim in five released CHANGELOGs — decide whether released records get amended #5913 → PR docs(scripts): record the withdrawal of the vacuous byte-identity sentence #6005 → ✅ MERGED (docs(scripts): record the withdrawal of the vacuous byte-identity sentence).
• finding(fields): every lazy widget import in packages/fields/src/index.tsx is defeated by a static import of the same module — 13+ INEFFECTIVE_DYNAMIC_IMPORT warnings per console build #5325 → PR chore(console): pin the 43 ineffective dynamic imports to a checked ledger #6007 → ✅ ACCEPT, all 22 checks green (19 success + 3 skipped, zero failures), marked ready and auto-merge enabled 12:43Z.
• finding(types): TableColumn.type disagrees three ways — interface declares 8 literals, zod mirror accepts any string, renderer's live read set handles values outside the union via an as any cast #5853 → ⛔ ESCALATED back to needs-user-decision (see §6).
• ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293 → PR fix(plugin-view): spell the views-prop sort direction key order, not direction #6008 → ⚠️ REWORK (narrow, prose-only); the code half is accepted and stays as pushed (see §5).
• ui:icon reads the SDUI identity key name as its lucide icon name, so any icon node that authors name renders nothing at all #5631 → carried into R38.
• Also verified merged this window: feat(types): close the dashboard widget type vocabulary and gate the plugin-dashboard catalog #6003 (feat(types): dashboard widget vocabulary closed + plugin-dashboard catalog gated, from schema-catalog: all 9 plugin-dashboard examples are refused by the spec's DashboardSchema — the AI few-shot corpus teaches the pre-ADR-0021 widget shape #4600).

⚠️ Blind-spot-5 keeps firing: R35 closed 2 cards but merged 4 PRs. ⛔ "PRs = cards − no-PR cards" is false. The five blind spots are unchanged:

1. ⛔ Do not count from the merged-PR list. 2. ⛔ A taken-over card may not carry domain:ui. 3. ⛔ The query returns cards closed by ANY seat — intersect first. 4. ⛔ Count by closed_at in THIS round's window. 5. ⛔ A card can produce more than one PR, or a PR without closing.

5. Notes

• ⛔ get_comments IS NOT OPTIONAL — R37'S PROTOCOL VIOLATION WAS SKIPPING IT. finding(fields): every lazy widget import in packages/fields/src/index.tsx is defeated by a static import of the same module — 13+ INEFFECTIVE_DYNAMIC_IMPORT warnings per console build #5325 was added late as a substitute and this seat read the body but never called get_comments, so it dispatched a card carrying a 2026-08-22 maintainer ruling (option C) without carrying the ruling. Owned publicly on the card. ✅ The rule is now absolute: every dispatch reads the card's full comment thread first, no exceptions for "small" or "obvious" cards. R38's three new dispatches each did — and each turned up a recorded ruling that the labels alone could not show (Reverse i18n sweep: ~384 confirmed-dead locale keys across 58 namespaces (workflow/BPMN designer entirely unwired, console.objectView oversized, more) #4730 ruled (b) 08-19, 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927 ruled the grouped framework 08-24, finding: phase2-schemas.test.ts imports 9 schemas no test uses — unused-import warnings overstate the file's pin coverage #5824 carrying a triage instruction to re-measure).

• ⛔ A CLAIM IN A CARD BODY IS A CLAIM, NOT A READING — R35'S THREE DISPATCH ERRORS ALL CAME FROM RELAYING ONE. The stale-premise check was run on the action face (git log/git show) and on Blocked-by: cards — not enough. Three dispatch words shipped a card's own assertion as fact:

  1. [finding] 「同一 en 字符串必须同一译文」不能做成门禁:281 组共享 en 值里 164 组至少有一个包刻意分开译 —— 记下分母,并点出其中两组像是笔误 #3880 — called [finding] en 包 32 个值用 ASCII "..." 结尾、113 个用 U+2026 "…" —— 同一屏上省略号两种字形,且九包各自跟着抄 #3878 "a separate, still-held card" and issued two clauses naming the same empty set; [finding] en 包 32 个值用 ASCII "..." 结尾、113 个用 U+2026 "…" —— 同一屏上省略号两种字形,且九包各自跟着抄 #3878 had landed as PR fix(i18n): one ellipsis glyph across the ten packs, es usted single-form, pt de-contraction (#3878 #3875 #3877) #4378.
  2. ObjectMap is the only view renderer that does not resolve marker titles through getRecordDisplayName — it reads a hard-coded 'name' key instead #5953 — asserted a test "pins the undefined behaviour". It does not; that is prose in a comment at line 138 with no assertion.
  3. Console/app-shell Sentry: a live DSN with sendDefaultPii: true is committed in apps/console/.env.production and baked into every production build — the VITE_SENTRY_ENABLED switch is build-time and absent from the inlined env, so a shipped bundle cannot be silenced #5522 — named "pin lag" as the live risk. There is no pin; the real risk is mirror drift.
     ✅ The fix, and it is cheap: the stale-premise check has a third face — every assertion the card makes about the tree, and every sibling card it references, gets re-read like a Blocked-by:. All three were one grep or one issue_read away. R38's dispatches each carry an explicit "the card's numbers are stale, re-derive and report the delta" clause for exactly this reason.

• ✅ A DEV MAY FALSIFY PART OF A MAINTAINER RULING — VERIFY EVERY LEG, THEN CORRECT THE RULING ON THE CARD. ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293's ruling ordered plugin-view/src/index.tsx:177 corrected as a views-prop producer. It is not one. This seat re-derived all three legs on origin/main rather than taking the report's word: the line sits inside defaultProps (opens :170) of ComponentRegistry.register('sort-ui', SortUI, { (opens :158); direction there is type-correct per SortUISchema.sort (packages/types/src/views.ts:862-871, interface opens :844); and it is read at runtime by SortUI.tsx:54/:62/:114/:116/:150/:152/:188/:207. The ruling's producer list and its own do-not-touch note were in direct conflict, and the note was the correct half. ✅ The dev obeyed the note and reported the conflict instead of silently picking a side — the behaviour the three-zone dispatch format exists to buy. ⛔ The PM corrects the enumeration on the card; it does not re-rule the decision.

• ⛔ AN ABSOLUTE NEGATIVE IN A CUSTOMER-PUBLISHED CHANGESET IS A REWORK TRIGGER EVEN WHEN THE FIX IS RIGHT. ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293's changeset said "No reader in the repo has ever known the word". One grep in the same package falsifies it: view-config-utils.ts:328, order: (s.order || s.direction || 'asc'), inside the published export toSortItems. It has no in-repo production caller and was never on the views-prop path, so the fix is untouched — but the changeset promises the break is "loud and searchable", and a migrating host greps direction, lands on a published export that still quietly accepts it, and concludes the retirement is partial. ✅ The scoped sentence in the same paragraph ("all three consumers of the resolved activeView.sort read order") was exactly right and carries the whole argument. The rule: a repo-wide negative assertion needs a green counter-probe, and prose that ships to customers is held to it.

• ✅ AN HONESTLY-REPORTED WRONG PREDICTION IS WORTH MORE THAN A GREEN TEMPLATE. ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293's dev predicted the runtime reverse-verification leg red, observed it green (12/12), reported it as observed, gave the sound reason (a type-only declaration is erased at runtime, so reverting it cannot move runtime behaviour), and then moved the falsifying mutation to the fixture to show the runtime assertion was not vacuous. ✅ The type leg's error was inverted — TS2578: Unused @ts-expect-error directive — which is the strongest possible evidence that the guard suppresses a real error rather than sitting inert. ✅ Ask for the inverted form.

• ✅ RULED CARDS ARE INVISIBLE FROM LABELS — THE DECISION RE-READ IS WHAT FINDS THEM. R35 found two cards in pm:queue with rulings already in hand (文档教作者写的 ActionSchema.condition = { expression, then, else } 全仓零消费者:谓词被忽略、then/else 永不派发、零诊断 #3917 ruled 08-09, schema-catalog: all 9 plugin-dashboard examples are refused by the spec's DashboardSchema — the AI few-shot corpus teaches the pre-ADR-0021 widget shape #4600 ruled 08-14, whose ruling comment itself says "is now unblocked"). R38 found two more (Reverse i18n sweep: ~384 confirmed-dead locale keys across 58 namespaces (workflow/BPMN designer entirely unwired, console.objectView oversized, more) #4730 ruled 08-19, sat 5 days; 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927 ruled 08-24). A ruled card and an unruled one are label-identical.

• ✅ A CLAUSE-② CARD CAN CARRY ITS OWN COMPENSATING EVIDENCE. What made 文档教作者写的 ActionSchema.condition = { expression, then, else } 全仓零消费者:谓词被忽略、then/else 永不派发、零诊断 #3917 safe was not the exemption but the evidence: a verdict flip measured in both directions (a throwaway worktree at origin/main parsed the same fixtures through the old schema), pinned in tests, plus a 32-package downstream tsc sweep against the new dist/*.d.ts. ✅ Ask for the flip, not just the fix. 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927's dispatch demands exactly that per widening (parses after / fails before / reverts red).

• ⛔ A GATE THAT COMPOSES CORRECTLY CAN STILL BE STUCK SHUT — READ ITS CALL SITE. Console/app-shell Sentry: a live DSN with sendDefaultPii: true is committed in apps/console/.env.production and baked into every production build — the VITE_SENTRY_ENABLED switch is build-time and absent from the inlined env, so a shipped bundle cannot be silenced #5522's initSentry() ran at module-eval time, before initRuntimeConfig() had started; with a fail-closed permission that reads DENIED and memoizes, the merged result would have been a silent permanent removal of telemetry for every deployment. A suite full of green absence assertions cannot see this. ✅ Every negative assertion needs a counter-probe that must stay green; absence is the shape a broken test reproduces perfectly.

• ⛔ A SUBAGENT DEATH IS NOT A VERDICT — TAKE IT FROM GITHUB. ✅ Probe set: remote branch · open PR · report comment. ⚠️ A branch ref resolving to a main commit is a stale ref carrying no work — check what the sha is.

• ⛔ THE HTML-COMMENT REPORT MARKER DOES NOT SURVIVE THIS REPO'S SANITIZER. Accept a plain-text first line; ⛔ never conclude "no report" from its absence. ⚠️ The sanitizer also ate a short <…> fragment inside backticks in a report comment and in a PR body. Every dispatch tells devs to avoid angle-bracket placeholders in prose. Filed: objectstack#11604.

• ⛔ THE REST CHANNEL IS DEAD IN THIS SESSION. curl https://api.github.com/... → 403 (the agent proxy's refusal). Everything via MCP.

  • ⚠️ MCP list_issues's labels is an OR, not an AND — a result wider than either input is the tell. ✅ Working method: one label, fields with labels/assignees, paginate with after, intersect client-side.
  • ⚠️ search_issues with label: returns total_count: 0 for a non-empty answer. ⛔ Not for lane scans.
  • ⚠️ Native issue type is NOT selectable in list_issues but IS returned by issue_read get as issue_type. ⛔ The bug label is not a proxy for the type (perf(console): 注册后 3-6s 白屏无 loading;每次硬导航都过全屏「正在初始化应用程序」 #2628, 文档教作者写的 ActionSchema.condition = { expression, then, else } 全仓零消费者:谓词被忽略、then/else 永不派发、零诊断 #3917, 审批界面(Approval Center)优化评估:徽章竖排 / 决策确认 / 进度可视化 / 动作标签 i18n #2762, ObjectMap is the only view renderer that does not resolve marker titles through getRecordDisplayName — it reads a hard-coded 'name' key instead #5953 are all native Bug carrying no label).
  • ⚠️ pull_request_read get_status returns only the legacy commit-status API (one Vercel row here) — it is not the gate read. ✅ Use get_check_runs.
  • ⚠️ actions_list on a busy workflow overflows the tool-result limit; extract fields from the saved file with python instead of re-requesting.

• ⛔ READ GATES BY NAME, ONCE, ~11 MIN AFTER THAT PR'S OWN RUN started_at. ⚠️ This seat read early three times now (R35 twice; once more on chore(console): pin the 43 ineffective dynamic imports to a checked ledger #6007 at 8m33s, finding 4 checks in_progress). An early read is a discipline failure, not a judgement failure — check the wall clock before the request; a background until timer is the cheap fix and it worked on the re-read.

  • Judge: every run completed, no failure/cancelled, mergeable_state: clean. chore(console): pin the 43 ineffective dynamic imports to a checked ledger #6007 read 19 success + 3 skipped = 22.
  • ⛔ 入队资格 = every check green, NOT the required subset. ⛔ mergeable_state: unknown is not clean; in_progress is not a verdict.
  • ✅ Build Docs green in ~13 s is a correct path-filtered skip when the diff is outside content/docs/**. Duration is not the discriminator; the executed step is.

• ⚠️ cancelled on main's push lane is NORMAL here, not a red main. Known domain:devx defect ci.yml's concurrency cancels the push-lane coverage job mid-run, so Codecov gets nothing on a merge that is followed by another merge #5422 (concurrency cancels push-lane coverage). The merge queue's own merge_group checks are what gate the merges. ⛔ Do not re-file; ⛔ do not read cancelled as damage.

• ⛔ Also still red on main, not this seat's: Test (coverage) is red on every push to main because CODECOV_TOKEN is unset — the four shards are green, the upload is rejected #5436 (CODECOV_TOKEN unset). domain:devx. ⛔ Do not re-file.

• ⛔ A MEASURED GRAPHQL BUDGET IS NOT A RESERVATION. Landing calls run on GraphQL (5000/hr, shared); reads on REST core. ⛔ Never retry a rate-limited call.

• ⛔ A CHANGESET IS NOT ALWAYS OWED — run node scripts/check-changeset-presence.mjs. Empty frontmatter is the "releases nothing" declaration. check-changeset-no-major.mjs refuses major; a public-type breaking change is minor by repo convention (AGENTS.md 版本号策略) — feat(types)!: retire the ActionCondition branch shape from ActionSchema.condition #5981 and ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293 are the precedents.

• ⚠️ objectui has NO scripts/pm/. dispatch-gates.mjs names gates that do not exist here; its --tier line is a floor only. Heavy legs route to the container-global os-verify-lock.sh in the sibling checkout — established practice.

• ⚠️ Brief app-shell cards to run path-filtered — whole-package vitest is ~784 s and holds the shared lock.

• ⚠️ This repo's vitest guard REFUSES the --filter forms (it silently drops args / runs the wrong package). ✅ Correct spelling: pnpm exec vitest run <paths> from the repo root.

• ⚠️ The type-check script is spelled type-check, hyphenated, and runs tsc --noEmit && tsc -p tsconfig.test.json, so test files are covered.

• ⚠️ A first-run RED reading Cannot find module @object-ui/... is the unbuilt-closure signature, not a real failure — build the dependency closure and re-read. ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293's dev hit it on both app-shell and console.

• ⛔ A subagent's send_later fires into the PM's session. Brief devs not to schedule their own check-ins.

• ✅ A claim from a different session id blocks, UNLESS demonstrably spent — report returned, round closed, card re-queued, and no branch on the remote. Burden is evidence, and it goes in the new claim comment.

• ✅ A GOVERNED-SURFACE CARD STOPS AT DRAFT. docs/adr/**, .claude/**, skills/**, AGENTS.md, CLAUDE.md. ⛔ Never ready/enqueued/auto-merged. Human merge IS the review record. Request review from os-zhuang — ✅ that works when the PR author is not os-zhuang; the 422 caveat applies only when the author is os-zhuang, where you assign instead.

• Stale-claim suspects, still not acted on: 除 content 外没有任何文本键既被求值又被读回 —— statistic.value / card.title / button.label 无法绑定表达式 #4795 and examples/schema-catalog: 48 个示例仍用已废弃的 div 组件类型,docs 站每张缩略图都刷一条 deprecation warning #3965 carry assignee yinlianghui — this seat's own GitHub identity, so earlier sessions' claims under the shared identity, not "non-agent assignees". Both >24h, no branch on the remote. Reclaim candidates; ⛔ needs the enquiry-and-silence window.

• Mislabels reported to triage, never changed here: [Decision]-titled cards in pm:queue ([Decision] Should component-registry lookup normalise case? Two measured instances of "every node renders Unknown component type" #5247, [Decision] schema.onNavigate is a function value read off the grid schema — the fifth key #5091's ruling did not cover #5234, [Decision] ObjectMapConfigSchema 要不要 .strict() —— 类型面的保证只保护 TS 作者,而 AI 产出的 metadata 恰恰不带类型 #5157, [决策] objectui record:details 渲染器的双形状接缝:合成器产 title/descriptor-fields 与 spec 区块形状经裸 ?? 缝合,往哪个方向收 #4018, [决策] System Hub bespoke 卡片墙(SystemHubPage)是否按 #3655 选项 C 退场 #3743, [Decision] Should @object-ui/types ship a generated JSON Schema? — the contract an AI author can read before writing #5392); 台账燃尽批次 8/8 · 被 objectstack#4171 阻塞的 any 擦除簇,3 符号——上游解除后处理(objectstack#4115) #3162 (blocked per its own title). plugin-form: navigateOnSuccess is mount-blind and says nothing when its destination is refused — the key has no ruling and its own contract question is still open #5034 carries assignee qq9340100 — ⛔ not available. 审批界面(Approval Center)优化评估:徽章竖排 / 决策确认 / 进度可视化 / 动作标签 i18n #2762 is a 20-item UX basket with a framework-side dependency — tracking-shaped, needs splitting. Console AI chat (ADR-0057) — post-cleanup UX follow-ups #2477 needs re-scoping: its item 2 appears already landed (studioChatDockPersistence.test.tsx exists on main), and it is the oldest card in the lane, so on age order it would buy a no-op.

6. ⏸ Awaiting a human — the standing list

Per the maintainer's 「待人类决策的整成清单，不要停掉活」: this list is the parking lot, not a stop condition. The loop keeps dispatching around it.

A. Needs a maintainer ruling (in the decision box):

1. ⏸ finding(types): TableColumn.type disagrees three ways — interface declares 8 literals, zod mirror accepts any string, renderer's live read set handles values outside the union via an as any cast #5853 — re-ruled request, escalated by this seat 08-24. The recorded ruling fixed a 12-value canonical vocabulary on NUMERIC_EDIT_TYPES evidence. Honest end-to-end measurement gives ~55: packages/plugin-grid/src/ObjectGrid.tsx:2206 does if (next.type == null && fieldDef.type) next.type = fieldDef.type;, forwarding the spec FieldType (49 options) verbatim into TableColumn.type. Tightening to 12 would refuse 42 live spellings — the exact failure the ruling rejected, at ~10× scale. Four-facet block posted with options A/B/C/D; non-binding lean A (import SpecFieldTypeEnum.options; precedent verified at ui-action.ts:28/:190/:211, and packages/types already depends on @objectstack/spec ^17.0.0). ⛔ This seat does not re-rule it.
2. ⏸ 17 zod mirrors refuse spellings their TypeScript declarations declare — each needs a widen-the-mirror / narrow-the-declaration ruling #5927 Groups B/C — eight per-key widen/narrow questions come back to the inbox with the renderer reading attached, per the 08-24 ruling. Not yet filed; the in-flight dispatch produces them.
3. ⏸ Published skill objectui/guides/data-integration.md teaches a QueryParams shape the adapters silently drop — unprefixed keys plus a direction sort key #6006 budget — dispatching the skills/** guide fix needs an expansion budget attached under the 2026-08-21 ruling.

B. Needs a maintainer hand on a button (no ruling required):

4. ⏸ PR docs(adr-0056): reconcile Consequences with the P2/P4 pure model #5970 (ADR-0056 的 Consequences 自相矛盾:「Capabilities 在 Setup 里作者化」与 P2/P4「在 Studio 设计、Setup 里没有 facet 可编辑」冲突 #3794, ADR-0056) — awaiting a human merge. Governed path, draft, CI clean (19 success + 3 skipped), review requested from os-zhuang. Two Consequences bullets reconciled; the second was found by the mandated audit and is a separate self-contained hunk a reviewer can drop alone.
5. ⏸ Console/app-shell Sentry: a live DSN with sendDefaultPii: true is committed in apps/console/.env.production and baked into every production build — the VITE_SENTRY_ENABLED switch is build-time and absent from the inlined env, so a shipped bundle cannot be silenced #5522 — pm:awaiting-maintainer. Two operator actions, neither of them code:
   • Rotate the Sentry DSN committed in apps/console/.env.production before console/app-shell: stop shipping an un-disableable Sentry DSN, and make sendDefaultPii opt-in (#5522) #5559 removed it — still in git history, presumably still live. ⛔ Rotating a committed credential is a maintainer action; this seat does not invent or rotate a DSN, and ⛔ deleting the value is not sufficient if a shipped bundle can still emit.
   • Grant the permission on the hosted console (OS_TELEMETRY_CLIENT_ERROR_REPORTING_ENABLED, or RuntimeConfigPlugin's allowClientErrorReporting) — otherwise SaaS error reporting goes quiet on deploy. Deliberate and documented; related SaaS/demo console deploy env must inject VITE_SENTRY_DSN — otherwise the hosted console ships with error reporting off after #5546 #5550.

C. Needs triage, not the maintainer:

6. A third teaching site still writes the retired condition: { expression, then } branch shape — guide/schema-overview.md, outside #3917's fence #5980 — carries no labels, therefore not claimable by anyone. Needs a triage first touch.
7. [finding] 「同一 en 字符串必须同一译文」不能做成门禁:281 组共享 en 值里 164 组至少有一个包刻意分开译 —— 记下分母,并点出其中两组像是笔误 #3880 stays OPEN as the denominator record — 281 shared en strings, 164 legitimately divergent; triage ruled this "knowledge, not dispatchable work". Carries domain:ui with no pm-state, so it sits in sweep case ③. → the right grade is tracking. ⛔ Not applied here; grading is not this seat's to produce.
8. Unassigned, awaiting a triage first touch: [finding] After #3878 merged the two ellipsis groups, Loading… is one 10-key group and de renders it four ways (ko two, ar two) — #3880's "two-key typo" no longer describes it #5972, ObjectMap's marker useMemo never memoizes — getMapConfig runs unmemoized in the render body, so mapConfig has a fresh identity every render #5976, ObjectMap.listViewMapConfigReach.test.tsx:138 narrates a getMapConfig literal in the present tense that no longer exists #5977, A third teaching site still writes the retired condition: { expression, then } branch shape — guide/schema-overview.md, outside #3917's fence #5980, Published skill objectui/guides/data-integration.md teaches a QueryParams shape the adapters silently drop — unprefixed keys plus a direction sort key #6006, plus [finding] aggregate-capability.test.ts's header claims analytics.query() never throws on a non-2xx — the installed client throws, and the tests below depend on it #5954 from R34.

D. Resolved since the last revision — do not re-file:

• ✅ ui:icon reads the SDUI identity key name as its lucide icon name, so any icon node that authors name renders nothing at all #5631's item 1 was ruled — maintainer, 08-24, verbatim 「5631 A′，按一次正经的契约迁移立项。」 The card is now in flight as a proper contract migration. The earlier escalation is discharged.

• ✅ The decision inbox was drained by the maintainer between ~09:50Z and 10:34Z on 08-24 in a live PM chat. Every card this lane had escalated up to that point (ObjectViewProps.views[].sort declares direction, but every sort consumer reads order — a host-supplied desc view silently sorts ascending #5293, finding(types): TableColumn.type disagrees three ways — interface declares 8 literals, zod mirror accepts any string, renderer's live read set handles values outside the union via an as any cast #5853, The withdrawn dist/index.d.ts byte-identity sentence survives verbatim in five released CHANGELOGs — decide whether released records get amended #5913, ui:icon reads the SDUI identity key name as its lucide icon name, so any icon node that authors name renders nothing at all #5631) was ruled. ⚠️ finding(types): TableColumn.type disagrees three ways — interface declares 8 literals, zod mirror accepts any string, renderer's live read set handles values outside the union via an as any cast #5853 has since been escalated again, on new measurement — see A.1.

• Open finding to the skills lane: objectstack#11604 — the mutex blind spot, the dead REST channel, and the sanitiser's appetite for short angle-bracket fragments. ⛔ This seat does not grade or route it.

[os-sales]

Audit probe — angle-bracket survival through this write path

Not lane state. Posted to measure, precisely, which angle-bracket forms survive a body written through the GitHub MCP tools, because the seat post currently asserts a mechanism ("the GitHub body sanitizer strips angle-bracket-delimited content") that a counter-observation contradicts: dependabot's PR bodies (#4945, #4958) carry hundreds of literal HTML tags intact. If those survive and mine do not, the variable is the write path, not GitHub. This comment will be read back and the result recorded; it can be deleted afterwards.

Five forms, each on its own line, each labelled so a missing one is identifiable by its label alone:

• A · HTML comment:
• B · JSX tag in prose:
• C · TS generic in prose: Array<{ a: string }>
• D · bare comparison in prose: a < b and c > d
• E · pre-escaped entity: <div>

Fenced block, same five:

A <!-- os-dev-report -->
B <div className="x">
C Array<{ a: string }>
D a < b and c > d
E &lt;div&gt;

Inline code spans: A <!-- os-dev-report --> · B <div className="x"> · C Array<{ a: string }> · D a < b > c

Control (must survive, proves the comment itself landed): ZZCONTROL-SENTINEL-5560.

---

Generated by Claude Code

[os-sales]

Audit note — round 7, cross-repo finding raised for the maintainer's attention. Seat post body is authoritative; this comment records one item ahead of the round-close rewrite because it names a production-affecting shape in the other repo, outside this seat's lane.

objectstack-ai/objectstack#10928, filed by the #4765 dev while measuring the installed spec. Unassigned and unlabelled — deliberately, because it is an objectstack card and neither the labelling nor the grading is this seat's to do.

The shape: UniqueScopeSchema's rejection message calls 'organization' "the explicit spelling of true" on both surfaces. Measured against installed GA 17.0.0, that is false on one of them — on a declared index, bare true is the positional spelling of 'global', not 'organization'. The helper readout is unambiguous:

value	isUniqueDeclared	isGlobalUnique	isOrganizationUnique
true	true	false	false
'global'	true	true	false
'organization'	true	false	true

Bare true sets neither scope flag, so no organization key part is prepended — which is precisely what makes it 'global' on that surface.

Why it is worth the maintainer's eye rather than the queue's. An author refused on that key is told by the error message itself to write 'organization'. Doing so silently changes materialization — the driver prepends the NULL-safe organization key part — on an index that may already exist on deployed databases. That is the unannounced reinterpretation the #8323 ruling and the #5082 sequencing exist to prevent, arriving through the error text rather than through a schema change.

Duplicate-searched before filing (two searches, objectstack repo). Nearest prior art is #8463 / PR #8512, which corrected the same field-vs-index divergence on a different artifact — isOrganizationScopedUnique's JSDoc in driver-sql — not on the shared error map.

Relevance to this lane, and why it changed a fix here. The same divergence is why #5614 does not uniformly rewrite unique: true: at field level bare true is not deprecated at all ("stays valid indefinitely … no trap" per the spec's own prose), so harmonizing both occurrences would have shipped a fresh piece of false guidance on a card whose entire purpose is stopping exactly that. The dev caught this by measuring rather than by following the card, and the card is better for it.

---

Generated by Claude Code