[PM seat] domain:spec @ objectui — 🟢 os-litant · session_01VhxTqo · R1 done (5 landed) · batch 5 · in flight: #9787

[objectstack-fleet]

_{Refreshed 2026-10-01T10:28Z, mid-R2 (round-open marker 5922723714).}

① Current PM

🟢 os-litant · session session_01VhxTqosz7wn54ahqyxgERT · seat id domain:spec#1 · seated 2026-09-30T22:27Z (round-open marker 5920819526), on the maintainer's summon /pm-dispatch spec@objectui.

• Wake Routine: trig_01GewDCCbCa71aH8xVxdGJCt (hourly, self-bound).
• Batch: 3 from here, on the maintainer's word in this seat's session at 2026-10-01T07:41Z, verbatim: 「当前任务处理完，后续并发降到3」.
  • Before that, the batch was 5 (「任务很对，并发可以加到5」).
  • The five tasks in flight run to MERGED, reviews included.
  • No new card is dispatched until fewer than three subagents are running. The count includes reviews.
• Contract review: CONTRACT_REVIEW_TIER runs as an isolated at-tier subagent; this seat is below that tier.

② Ledger

Landed this shift: 15. Each was verified on main by git patch-id --stable against the PR's net diff, with an at-tier PASS on the landing head.

card	PR	commit	review rounds
#10940	#11288	92970c4d6	1
#8945	#11286	969d4f291	3 (a FAIL on the at-rest reason, then fixed)
#8946	#11290	48401689f	1
#9409	#11284	02f1813f8	2 (a CI red: a doc version literal)
#8652	#11292	d79f525d9	2 (a docs round, AGENTS.md #2)
#9787	#11301	3d0994887	1 (three dev rounds before review)
#8831	#11308	50583364a	1 (two dev rounds)
#11168 slice 2	#11313	cd5b19a7e	1; the card stays open (Refs)
#11117	#11312	0e6e76bc4	1 (five dev rounds, one of them a merge of main after objectui#10859 batch 5)
#8654	#11324	95bd23c90	1
#11168 slice 3, PR B (object-form.layout; objectui#7759 group C)	#11351	17dc16793	1
#10224	#11329	fd6f5da44	1 + 2 deltas (the last on the merge that picked up objectui#11330's repair)
#11326 (domain:ui, the maintainer's direct dispatch 「11326 插队」)	#11367	d0fba91aa	1
#11354 (objectui#8347 preparation: plugin-form)	#11370	eccdf7baa	1
#11347 (objectui#8347 preparation: components)	#11360	c82ff391f	3 (list.title kept and declared by the seat's ruling 5927981915; then the types level corrected to minor). The card was closed by the seat. Content is verified by changed lines, because main moved the context.

Closed without a PR: 2. #8367 and #9111. PR #11086 (the 17.5.0 bump) had already delivered each one's acceptance.

In flight: 3 (batch 3).

• spec parity: the ten blocks @objectstack/spec 17.5.0 newly carries, with their unpublished inputs, refused arms and member pins — declare each by measurement (objectui#11111 decision 3 = B) #11168 slice 3, PR A (objectui#11335), round 5. The record 5929481787 judged round 4 RIGHT except one object-tree clause: data.object stands in for a missing objectName. Round 4 had reworded navigation to the objectui#11293 behaviour (the host navigator opens page).
• spec(react): SchemaRenderer's 37 un-cast reads off its working copy of a node block objectui#8347's index-signature removal (the interpreter's copy is renderer state; authored keys get the per-key verdict) #11349 (react): PR objectui#11374 is under at-tier review. The working copy is typed internally, and react reads 0 / 0 under the P1 mutation.
• spec(plugin-dashboard,plugin-designer): twenty un-cast reads of undeclared widget keys block objectui#8347's index-signature removal (layout, title, colorVariant, filter, pageSize) #11348 (dashboard + designer), dispatched at 10:09Z.

#8347 re-planned. Premise P1 failed: 303 errors once the signature goes, 95 of them in shipped source across 13 packages, almost none ruled.

• The card is pm:blocked behind its preparation cards. spec(components): twelve un-cast reads of undeclared keys block objectui#8347's index-signature removal (list, loading, input-otp, sidebar, drawer, dropdown-menu, popover, sheet) #11347 and spec(plugin-form): ten object literals the form variants author with keys their target type does not declare block objectui#8347's index-signature removal (fields on drawer/modal, layout on split/tabbed/wizard) #11354 have landed. Still open: spec(plugin-dashboard,plugin-designer): twenty un-cast reads of undeclared widget keys block objectui#8347's index-signature removal (layout, title, colorVariant, filter, pageSize) #11348, spec(react): SchemaRenderer's 37 un-cast reads off its working copy of a node block objectui#8347's index-signature removal (the interpreter's copy is renderer state; authored keys get the per-key verdict) #11349, spec(app-shell,plugin-detail,plugin-kanban,cli,plugin-gantt,plugin-map,tests): the small P1 sites that block objectui#8347's index-signature removal, one bundle #11355, and spec(types,react): the spec's public blocks, page kinds and the action:button node have no TypeScript authoring type, so objectui#8347's index-signature removal leaves their TS authors no face (29 literals measured in components' tests) #11364 (the TypeScript faces for the spec's public blocks, page kinds and the action:button node, measured from 29 component test literals).
• The removal lands last on spec(types)!: remove BaseSchema's index signature from the authoring face, declare the riders, pin the compile-fail — executes the #7927 removal ruling (governed: AGENTS.md #0.1) #8347 itself. It is governed, so it waits for a human merge.
• Q1 (prepare per package, remove last) is answered by the objectui#7927 ruling's step 1 and objectui#8327's 5587717012.

Filed: 12.

• objectui#11293 (inert navigation on a standalone kanban or calendar).
• objectui#11302 (MetadataFieldsPage renders only the refusal headline).
• objectui#11330 (Spec Main Shape Gate red after objectstack d830d71f re-branded the retired-key type). The merge queue bounced PRs from 05:12Z. domain:ui seat 1 claimed it at 05:38Z, and its fix PR objectui#11343 entered the queue at 06:21Z.
• objectui#11347, spec(plugin-dashboard,plugin-designer): twenty un-cast reads of undeclared widget keys block objectui#8347's index-signature removal (layout, title, colorVariant, filter, pageSize) #11348, spec(react): SchemaRenderer's 37 un-cast reads off its working copy of a node block objectui#8347's index-signature removal (the interpreter's copy is renderer state; authored keys get the per-key verdict) #11349, spec(plugin-form): ten object literals the form variants author with keys their target type does not declare block objectui#8347's index-signature removal (fields on drawer/modal, layout on split/tabbed/wizard) #11354, spec(app-shell,plugin-detail,plugin-kanban,cli,plugin-gantt,plugin-map,tests): the small P1 sites that block objectui#8347's index-signature removal, one bundle #11355, spec(types,react): the spec's public blocks, page kinds and the action:button node have no TypeScript authoring type, so objectui#8347's index-signature removal leaves their TS authors no face (29 literals measured in components' tests) #11364 (the spec(types)!: remove BaseSchema's index signature from the authoring face, declare the riders, pin the compile-fail — executes the #7927 removal ruling (governed: AGENTS.md #0.1) #8347 preparation cards).
• objectui#11365 (input-otp's documented separator draws nothing; filed bare, for triage).
• objectstack#20997 (the reserved-name refusal is reported at fields).
• objectstack#21005 (action:button / action:icon accept endpoint, which ActionSchema refuses with endpoint → target).

Held by another seat: #11070 (domain:devx#2). Not dispatchable: #11334 (pm:on-hold), and #11228 / #7927 (pm:blocked).

Queue (domain:spec + pm:queue):

• p2: spec(app-shell,plugin-detail,plugin-kanban,cli,plugin-gantt,plugin-map,tests): the small P1 sites that block objectui#8347's index-signature removal, one bundle #11355 (the small-sites card). Its map and gantt items are serial behind PR A and objectui#10859 batch 6.
• p2: spec(types,react): the spec's public blocks, page kinds and the action:button node have no TypeScript authoring type, so objectui#8347's index-signature removal leaves their TS authors no face (29 literals measured in components' tests) #11364 (public-block TS faces), after spec(react): SchemaRenderer's 37 un-cast reads off its working copy of a node block objectui#8347's index-signature removal (the interpreter's copy is renderer state; authored keys get the per-key verdict) #11349 lands.
• p2: 27 keys across 18 pairs have a zod mirror that ACCEPTS a spelling its TypeScript declaration refuses — each needs a narrow-the-mirror / widen-the-declaration ruling #7759's remaining groups (B, D, F), after PR A.
• p3: 121 declared-but-unmirrored keys across 16 schema pairs — the lane #6058's new UnmirroredDeclared ledger made visible #6152's next round, after PR A.
• spec parity: the ten blocks @objectstack/spec 17.5.0 newly carries, with their unpublished inputs, refused arms and member pins — declare each by measurement (objectui#11111 decision 3 = B) #11168 slices 4 (object-gantt) and 5 (object-timeline, plus booking (c), the kanban/calendar openNewTab wording and the ObjectTimelineProps.navigation JSDoc).

Waiting elsewhere on #11168:

• endpoint ×2 waits on objectstack#21005.
• action:button.undoable waits on the maintainer's ruling: B (a runtime path so the block captures Undo) or C (narrow the spec row).

③ Hot-file serial queue

region	state
apps/console/src/__tests__/registry-inputs-spec-parity.test.ts	Held by PR A (owner 5 by census at 24c5172f9; rounds 4 and 5 touch only MEMBER_PINS rows).
packages/types/src/__tests__/zod-mirror-parity.test.ts	Free. PR B's row has landed. Next: whichever preparation card opens a row, then #6152, then #7759.
packages/react/src/SchemaRenderer.tsx	Held by #11349 (PR #11374). #11364 follows it for the published schema prop type.
packages/types/src/base.ts (BaseSchema)	Reserved for #8347's last PR, after its five preparation cards.

④ Notes

• H4 worklist with no card: PR fix(types): a gate-wrapped arm waives objectName when dataSource.object binds the node (objectui#11117) #11312's body lists 21 retired-orthography dataSource.filter test fixtures to convert and 3 to keep. It is hygiene with no public-door reach, so it has no card. Whoever next edits those test files takes it.
• Writes go through the fleet-write relay. ⚠️ This container's write pace (40/hour) was exhausted at R1's peak, with 5 devs and 4 reviews plus their patch rounds. Writes are queued in order and drained by the reopen time, and dev and review briefs carry the "exit 10 ⇒ hand the file back" rule.
• An account-wide API session limit stopped all five subagents at about 07:5xZ (it reset at 08:20Z). Each resumed in its own worktree, and none lost work.
• This seat's own misses this shift, each corrected on its card:
  • dataSource.filter's "three shapes legitimately reach a renderer" note is now half true — upstream converged the authoring door on the ViewFilterRule array #8945: a card-body premise was quoted as a ruling in the dispatch.
  • [finding] objectui still declares page.assignedProfiles and describes it as "Profiles that can access this page" — objectstack is retiring the key under ADR-0090 D2, which deleted the Profile concept #9409: a single-line grep was reported as "0 hits".
  • decision(plugin-kanban): navigation is read off the kanban node and declared on neither arm — while the grid schema declares it (5 of 7, objectui#8327 class (a)) #8652: a grep truncated by head -20 was reported as "0 hits" on the timeline.
  • fix(components,types): eight renderers stop reading undeclared node keys; input-otp and loading read their declared spellings (objectui#11347) #11360: an addendum assumed a flagged changeset address sat in the dev's appended note. It sat in the append-only prefix, and the dev was right to leave it.
  • Spec Main Shape Gate red: objectstack d830d71f gives a retired key a branded [REMOVED] Key retired type, and four objectui type pins plus LayoutRenderer.tsx do not compile against objectstack main #11330: filed with no labels, while the standing ruling's readers find work through domain:* lane probes. The merge-queue red sat for about 45 minutes before a domain:ui seat picked it up.
• The director's reminder 5903839371 still holds: a review record covers the LANDING head.

[objectstack-fleet]

Round-open marker — seat founded, R1

domain:spec @ objectui, session session_01RJpZ7i6X3hdxouUgKaqiQg, marker written 2026-09-22T16:57Z.

writes via fleet-write relay as objectstack-fleet[bot] · session session_01RJpZ7i6X3hdxouUgKaqiQg · route read 2026-09-22T16:55Z · first write = this post's parent issue #10217, author objectstack-fleet[bot], relay run 35757308315

Preflight — the three checks, before any write

#	check	reading
①	objectstack checkout at main	git pull --ff-only origin main clean; HEAD 16d090ed — ⚠️ equal to the named commit, ⛔ not strictly after it. The substantive condition holds: .github/workflows/fleet-write.yml and all four of scripts/pm/fleet-write/{dispatch,execute,ops,validate}.mjs are present. Recorded as equality rather than reported as 「在之后」.
②	session id	session_01RJpZ7i6X3hdxouUgKaqiQg (from get_session → ccr.id)
③	route	"transport":"dispatch" ✅ exit 0

⭐ The route's own reason independently corroborates a measurement this seat took before the instruction arrived: 「a cloud seat container, whose proxy replaces the Authorization header」. This seat had measured exactly that — a deliberately bogus bearer token still answered 200 and still resolved to a user login, and every /app* path answers 403 "Access to this GitHub API path is not permitted through this proxy". ⇒ the App token can neither be minted in this container nor supplied to it from outside; the relay is the only route to the bot identity. ⛔ The first relay run did not fail at 「Mint the fleet App token」 ⇒ the App installation scope does include objectui.

Round-mutex — four readings, taken before any write

reading	value
① last stand-down brief	none — no seat post for this seat existed; this post is its first
② newest round-open marker after it	none
③ newest non-self Claim: bearing this seat id on the lane's pm:dispatched ∪ pm:queue	none
④ newest CLOSED lane card bearing this seat id	none

Enumeration behind ①/②: all pm:seat in objectui, open+closed, = 4 (#9800, #9771 domain:ui; #5748 domain:devx; #5560 domain:ui) ⇒ none is this seat. ⇒ mutex clear. domain:spec seats 1–5 on objectstack were read, ⛔ not touched — this is the per-repo split, ⛔ not a takeover of any of them.

Charters re-read this round (git-history.mjs touch)

charter	touch
SKILL.md	7e0bfce40
references/core-rules.md	7e0bfce40
references/lanes/spec.md	4d1d9d9f9

⚠️ The newest sibling marker on the lane recorded 37ed9ae04 for the first two ⇒ they moved ⇒ re-read this round by construction, ⛔ not inherited.

check-harness-current: CURRENT — every harness-loaded path on origin/main is in the shared HEAD 16d090ede0. ⛔ No swap owed.

Tier — a re-measurement that reverses the newest sibling reading

origin/main:scripts/pm/dispatch-gates.mjs:12282:export const CONTRACT_REVIEW_TIER = 'claude-fable-5-1';

The newest domain:spec marker on the fleet recorded this constant as claude-opus-5 and concluded 「this seat is AT TIER」. That reading was true when taken; commit 372931e (「the contract-review tier is fable — revert the opus constant」, #19684) reverted it. This session is served claude-opus-5 (read from the session record, ⛔ not assumed) ⇒ BELOW TIER.

⇒ ⛔ no in-seat Clause-② review this round; a Clause-②: yes card needs an isolated at-tier subagent. This round therefore takes Clause-②: no work only. ⭐ Same shape the sibling seat itself named as its shift's lesson: a measurement authorises a conclusion only under the condition it was taken in — including that seat's own.

[objectstack-fleet]

⛔ Published correction — two figures in this post were wrong, and the corrected finding is STRONGER

Written 2026-09-22T17:00Z. This corrects the seat post body (§④) and the round-open marker 5780565247, both written by this seat minutes ago.

What was wrong, and the single cause

Both errors came from one flawed instrument: I counted comments with grep -c '"id":', which counts nested ids (comment.id, user.id, …), ⛔ not array members.

published	actual	verdict
「control: #9837 returns 4」	#9837 returns 1	⛔ wrong — 4 was the "id": occurrence count of a single comment
「MCP get_comments under-reports independently — #9837 gave 1 of 4」	MCP returned 1, REST returns 1 — they agree	⛔ withdrawn in full — there is no MCP under-reporting; that was an artefact of the same bad counter

⭐ The control word discipline catches the shape of a zero but ⛔ not a miscounted positive. The re-measure used a JSON-aware count (array length), ⛔ not a grep.

The finding itself — unchanged in direction, and now much better evidenced

The 11 cards still return zero comments, and the control still passes (#9837 returns 1, ⛔ not 0) ⇒ the instrument can see comments ⇒ the zero is a real absence. But the corrected re-measure found the thing the bad counter had hidden:

All 12 cards — the 11 from 2026-09-20 AND the control #9837 — carry a metadata comments counter that exceeds the actual list by EXACTLY 1.

card	meta	actual
#9837 (control)	2	1
#10145 #10084 #10083 #10078 #10074 #10073 #10072 #10071 #10069 #10068 #10066	1 each	0 each

⇒ exactly one comment was deleted from each of twelve cards. GitHub's counter does not decrement on delete, which is why the two readings disagree — the counter is the fossil of the deleted comment.

Named, verified instance. #9837's surviving ruling cites its triage comment by id: 「Facts (this card; triage 5727829764)」.

GET /repos/objectstack-ai/objectui/issues/comments/5727829764  -> 404 Not Found
GET /repos/objectstack-ai/objectui/issues/comments/5729875984  -> 200   (control: the ruling itself, alive)

⇒ the instrument distinguishes deleted from alive, and the triage measurement underpinning a maintainer-agreed ruling is gone.

What this changes for dispatch — one card separates from the other eleven

• The 11 stay undispatchable. Each says 「the comment governs and this paragraph does not」 and declares itself CONFIRMED by the deleted comment ⇒ no surviving authority, and the body disclaims its own wording. ⛔ Not dispatchable on their own terms.
• finding(plugin-grid): object-grid 的 pagination 与 selection 用相反的规则读「自己在不在」—— 写下对象,一个生效、一个无效 #9837 remains dispatchable. Its deleted comment was the triage one, but the surviving ruling (5729875984, maintainer 「同意」) restates the facts it relied on in its own body ⇒ it is self-contained; the card's authority does ⛔ not depend on the deleted comment.

⇒ the separation is 「does a surviving authority state the facts」, ⛔ not 「how many comments are left」.

⚠️ For the triage seat and the maintainer

Twelve cards in one lane each lost exactly one comment, and the deleted ones are the triage re-measurements — including one a director-seat ruling cites by id. ⛔ This seat does ⛔ not know the cause (deliberate deletion, a retracted bulk write, or platform-side) and ⛔ does not guess. Reported, ⛔ not diagnosed. The evidentiary base of the 2026-09-20 filing batch is gone and needs re-measuring at source before any of those 11 can be dispatched.

[objectstack-fleet]

Stand-down brief — seat released (final write)

Authority, verbatim, maintainer in this seat's session: 「你可以下班了」 · and on the open findings: 「你刚才提的问题总监都接手了，不用再管」.

Stand-down = state flush plus explicit placement of what is in flight. This comment is the seat's last write: the title, the body and #9837's state label were changed in the same relay request, immediately before it. Read at 2026-09-23T03:09Z.

① Landed this shift — 1

card	PR	merge	verified by content
#9837 (ruling batch #162 item 2 · A′)	#10218	62597c5, an ancestor of objectui main	ObjectGrid.tsx:4404-4408 checks showPagination === false before pagination presence, and the changeset @object-ui/plugin-grid: minor is on main

⚠️ Honest note on the instrument: my first dark control, the snippet pagination !== undefined → ? true, still matched on main, because A′ keeps that branch and only puts the explicit off ahead of it. The discriminator is the order of the two checks, ⛔ not the snippet. It was re-read on that basis.

② In flight — 0

• finding(plugin-grid): object-grid 的 pagination 与 selection 用相反的规则读「自己在不在」—— 写下对象,一个生效、一个无效 #9837 was closed by Fixes. pm:dispatched was removed in this request (a closed card carries no pm state label); the assignee stays as the record.
• The dev subagent returned its report (5781323644) and was accepted (5787900889). No other dispatch was made.
• The lane's only other pm:dispatched card, objectui#8068 的过渡背书：58 个 array/object 类 input 仍无 per-block 成员 pin（逐条转 pin，删豁免） #8071, belongs to hotlong — ⛔ not this seat's. Its staleness is among the items handed to the director.

③ Handed to the director — ⛔ nothing here is left to a successor seat

The post-stamped fail-open fallback, the relay's automerge_enable permission, the classifyHttp 403 misclassification, the 12 deleted triage comments, the spec selection.type default seam, and #8071. Source of the handover: the maintainer's words above. ⛔ This seat filed no card and made no pm:retriage write for any of them.

④ This shift's own misses, stated plainly

• No wake Routine was created at seating (the rule is 「入座即建唯一自绑 cron Routine」). The shift ran on maintainer turns only ⇒ there is nothing to delete, and ⛔ no timer points at this session.
• A miscount went onto the board and was corrected publicly (5780598365): grep -c '"id":' counted nested ids, ⛔ not comments.
• A claim draft carried a bare timestamp inside a quoted span. post-stamped refused it (exit 2, zero writes), so the atomic pair was never split.

⑤ Closing readings

• Writes this shift: every one went through the relay as objectstack-fleet[bot] — ⛔ zero direct writes.
• One relay run stopped part-way: run 35810462718 stopped at automerge_enable. The two actions before it landed and were recorded on finding(plugin-grid): object-grid 的 pagination 与 selection 用相反的规则读「自己在不在」—— 写下对象,一个生效、一个无效 #9837 (5787926803); the maintainer then merged by hand.
• Zero suggestions, per the shift-report default. Everything that might have become one is with the director.

[objectstack-fleet]

Round-open marker: R1 of this shift

domain:spec @ objectui · session session_01877XiBYSaRCk2CU7cMSg3S · 2026-09-23T23:49Z

Round mutex. ① Newest stand-down brief: 5788305956 (the predecessor, session_01RJpZ7i6X3hdxouUgKaqiQg). ② No round-open marker after it. ③ Lane pm:dispatched ∪ pm:queue shows no Claim: newer than the brief: the newest card updated_at in the lane is 2026-09-22T03:51Z, which is before the brief. ④ Nothing on the lane has closed since then. The brief is the release marker, so the seat is taken directly.

Charter touches (git-history.mjs touch, objectstack origin/main fdeeea0): SKILL.md 0bc46c6d1, references/lanes/spec.md f151ef2c9, references/execution-duties.md f151ef2c9. All three were read this fire. check-harness-current: CURRENT.

Round outcome: 0 dispatched. The reasons are on the body (§②): the batch is held on deleted evidence, #9849 is waiting on pm:retriage, and the North-Star gate is closed for the remaining p2/p3 hygiene and tooling cards. The Routine re-checks every hour. It stays silent if nothing has changed.

[objectstack-fleet]

Seat-post body corrected at 2026-09-24T01:48Z. Session session_01877XiBYSaRCk2CU7cMSg3S found the body out of date during an hourly check. Triage rerouted #9849 to domain:ui, and #9747 was closed as a duplicate. The P0/P1 counts were re-read, and the North-Star gate is still closed. The lane still has 0 dispatchable cards and 0 in flight.

[objectstack-fleet]

Seat-post body corrected at 2026-09-24T03:48Z. The hourly check found #8346 closed not_planned, so it is off the gated candidate list. Nothing else changed: the North-Star gate is still closed (objectstack 4 P0 / 29 P1, objectui 0 P0 / 7 P1). The lane still has 0 dispatchable cards and 0 in flight.

[objectstack-fleet]

Seat-post body updated at 2026-09-24T05:50Z. On the maintainer's instruction, this seat re-measured the 2026-09-20 batch at source, one comment on each of the 11 cards. The ledger in §② now classifies them. No card was claimed or dispatched in this act.

[objectstack-fleet]

Round-open marker: R1 of this shift · 2026-09-27T01:26Z

domain:spec @ objectui · session session_014fWVhLzhxR8qrFsJ5o8TYW · account os-tesla. Summoned by the maintainer (/pm-dispatch spec@objectui).

Round mutex (four readings, taken before any write):

reading	value
① last stand-down	this post's body revision at 2026-09-25T04:25Z (「Unseated」, predecessor session_01877XiBYSaRCk2CU7cMSg3S, Routine disabled)
② round-open marker after it	none
③ newest non-self Claim: bearing domain:spec#1 on the lane's pm:dispatched ∪ pm:queue, or a branch push after ①	none. #7469's claim 5827036862 is domain:ui#5 (another seat; released at 5827639834), and its branch sits at BASE. claude/issue-7759-* last moved 2026-09-09.
④ newest CLOSED lane card bearing this seat id	#10289, predecessor's session (before ①)

⇒ The stand-down is the newest event, so the seat is taken directly.

Charters re-read this fire (git-history.mjs touch, objectstack origin/main 16c5a33): SKILL.md 7b068877c · references/execution-duties.md 7a13e0562 · references/lanes/spec.md f151ef2c9 · references/seat-lifecycle.md f151ef2c9 · references/landing-operations.md 4502bfb59 · references/cross-repo-coordination.md 3f3c0f765. check-harness-current: CURRENT.

Tier. CONTRACT_REVIEW_TIER = 'claude-fable-5-1', and this seat is below it. Every contract review therefore runs as an isolated at-tier subagent.

This round's batch (3), by the take order. Contract face first, then priority:p2, then Bug and card age:

• finding(types): the 「record id is a string」 rule has THREE unruled survivors — CommentSearchResult.recordId and RecordSubscription.recordId (both REQUIRED, published) plus DataSourceMutationEvent.id #10078 (p2): the inherited record-id ruling (see its claim);
• finding(types,plugin-kanban): ColumnWidthConfig / ColumnWidthConfigSchema are published with zero readers and no schema key that references them (objectui#8522 removed the last reader) #10582 (p3 bug): retire ColumnWidthConfig;
• finding(types,plugin-charts): ObjectChartSchema.xAxisField / yAxisFields / aggregation are declared on the object-chart node, but ObjectChart reads none of them; a node written with them draws no category axis #10608 (p3 bug): retire the three inert ObjectChartSchema keys.

Serial behind them: #10611 after #10078 (views.ts), and #10584 after #10608 (ObjectChartSchema block). #7469 (p3 enhancement, ruling C) is next in line.