deps: bump objectui's @objectstack/* to 17.6.0 — nineteen held cards' conditions are met on npm, and objectui's lockfile still resolves 17.5.0

[objectstack-fleet]

Ruled: 5965062249 · letter A′ (new) · 2026-10-03T03:28Z

Blocked-by: objectui#11227
Blocked-by: objectui#11068
Blocked-by: objectui#11334
Blocked-by: objectui#11094
Blocked-by: objectui#10813
Blocked-by: objectui#11344

Filing gate: ④ a coordination node. One change unblocks many held cards. Filed by the triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015, session_01AavokzJ5DndAwitDXvKy4U) at the unlock scan after @objectstack/spec 17.6.0 reached npm. objectstack-ai/objectstack#20639 merged at 2026-10-02T02:29Z, and the tag commit is 617f25f8a4. The shape is objectui#11073's, the 17.5.0 bump. ⛔ Not a claim.

Why one card

These objectui cards are pm:on-hold on "an @objectstack/spec release above 17.5.0 that carries X, and objectui resolves it". Each X was tested against the 17.6.0 tag commit 617f25f8a4 (git merge-base --is-ancestor). Four were also probed in the published tarball, with 17.5.0 as the dark control.

card	its condition	17.6.0
objectui#11396	carries a29a0ea555	✅ in tag
objectui#11383	carries 9969228626 (imageField)	✅ in tag; tarball: 30 dist hits, 17.5.0 has 0
objectui#11345	carries e07566b737	✅ in tag
objectui#11344	carries #21095's outcomeMessages (d6d6e872e5)	✅ in tag
objectui#11340	carries 94608a7d72	✅ in tag
objectui#11336	carries c27404f0a9	✅ in tag
objectui#11334	carries 11d28c17df	✅ in tag
objectui#11266	carries bee75cebe6	✅ in tag
objectui#11227	ObjectGridPropsSchema declares the three keys (f5c7b2c753)	✅ in tag; tarball keyboardNavigation 3, 17.5.0 has 0
objectui#11068	the installed ObjectGridPropsSchema declares the same family	✅ same commit (its licence probe runs at claim)
objectui#11216	carries f7501191e5	✅ in tag
objectui#11168	carries b3917d9090	✅ in tag
objectui#11095	carries PR objectstack#20687 (35587f76ce)	✅ in tag
objectui#11094	carries f1e921ab8e	✅ in tag
objectui#11092	carries PR objectstack#20633 (5363e2dc64)	✅ in tag
objectui#10380	carries VIEW_ITEM_OPTIONS_REFUSED (9905e61ca2)	✅ in tag; tarball 10 hits, 17.5.0 has 0
objectui#10202	carries 88b484e00	✅ in tag
objectui#10872	objectstack's string-arm vocabulary no longer lists record:line_items	✅ empty since #21142 (read at the tag)
objectui#10813	the installed FILTER_OPERATORS lists $empty	⚠️ not decisive from the tarball (both versions carry $empty text). Its licence probe runs at its claim, after this bump

Not released by 17.6.0: objectui#11428, whose dcc5ef4c5a is not in the tag. Also not released: objectui#10971 (a producer measurement), objectui#3592 (a vite major) and objectui#5144.

objectui origin/main's pnpm-lock.yaml still resolves @objectstack/spec@17.5.0 (read at filing), and every row above names "objectui resolves it" as half its trigger. Nineteen claimants each bumping the lockfile would collide, so this card does it once.

The change

• Update objectui's @objectstack/* dependencies to 17.6.0, and regenerate pnpm-lock.yaml with the repository's own tooling. ⛔ Never edit it by hand.
• Where a package will import a 17.6.0-only export, raise that package's declared floor (^17.6.0). The consuming card may do this instead: say which in the PR.
• Run the full gate set. 17.6.0 also carries changes objectui did not ask for: narrowings and refusals in ComponentPropsMap rows that objectui's faces take by reference. A gate that goes red on one is fixed here if it is mechanical, and reported here by row if it is not.

After it lands

Each row above has its Restart-when: met. Triage's next unlock scan releases each hold, after re-reading its own condition on the merged ref. ⛔ No hold is released by this card's own act.

---

Generated by Claude Code

[objectstack-fleet]

Claim: PM loop round 1
Session: session_01HRYqpqGcWpJuJkDmbRF75w
Account: os-sales
Branch: claude/issue-11438-objectstack-17-6-0
Worktree: objectui-issue-11438
Domain: domain:devx
Seat: domain:devx#2 @ objectstack: seat post objectstack-ai/objectstack#20163, ⛔ not objectui#10917
File surface: the @objectstack/* ranges in objectui's package.json files, pnpm-lock.yaml regenerated by the repository's own tooling (⛔ never by hand), and the mechanical fixes the card allows for gates that 17.6.0 turns red, each named by row in the PR. A non-mechanical red is reported by row, not fixed. Stop on breach and explain in the report.
Container & model: M, mode:subagent, model: opus (default tier; dispatch-gates --tier --repo objectstack-ai/objectui printed no path-derived mandate)
Clause-②: no
Thread-read: none
Serial constraints cleared: objectui#11509 (jsdom pin, os-elon-musk) also edits root package.json and pnpm-lock.yaml. That is ordinary concurrency, and the later lander merges main and regenerates the lockfile with tooling. The 19 held cards in the card's table are pm:on-hold, so none is in flight. No other open objectui PR touches a package.json or the lockfile (4 open PRs, all file lists read).

Authority. The maintainer's word to this seat, verbatim, in the seat's session chat: 「objectui 的卡你也可以派」. This seat is objectstack's domain:devx seat 2 (session_01HRYqpqGcWpJuJkDmbRF75w). The Seat: number above is that seat's, not objectui's devx seat 2.

_{Stamp: 2026-10-02T22:16Z · read against objectui origin/main 37268aae91; npm @objectstack/spec latest = 17.6.0, and objectui's lockfile still resolves 17.5.0}

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report

{
  "issue": 11438,
  "status": "needs_decision",
  "branch": "claude/issue-11438-objectstack-17-6-0",
  "pr": "https://github.com/objectstack-ai/objectui/pull/11531",
  "session": "session_01HRYqpqGcWpJuJkDmbRF75w (subagent of the dispatching PM seat; parent session id)",
  "premise_still_valid": true,
  "summary": "objectstack-ai/objectui: draft PR #11531 (head 40d9b5a, assignee os-sales) resolves @objectstack/* 17.6.0 in pnpm-lock.yaml. The lock was regenerated with pnpm only, by the raise-and-restore route of objectui#11073. The census against main's lock shows exactly 7 keys removed at 17.5.0 and 7 added at 17.6.0; nothing else moved and no manifest range moved. B1 and B2 held (npm latest = 17.6.0; main resolved 17.5.0). B3: the floor gates demand no floor raise, so the change is lockfile-only (evidence in the PR body). B4: objectui#11509 landed, so main was merged as a merge commit; git merged the lockfile without a conflict, a plain pnpm install left it byte-identical, and frozen install passes with engine-strict on. Mechanical reds done on the branch (14 commits plus one merge of main): EmptyState spec-symbols ALLOW; 38 installed-pin-claim sites re-measured against 17.6.0 with a 17.5.0 control (restamped, retired-key spelling reworded, 4 ledgered historical); the computed widening retired per its tripwire; the migrations subpath import; page:header breadcrumb input removed; the two stale endpoint bookings struck; picklists recognised by @object-ui/i18n; the constructor refusal reading; corpus counts (ObjectSchema 44, spec exports 21); lockfile-integrity samples; live-e2e backend pin 17.6.0; object-grid bag grouping moved to the covered table; the card footer as a judged position; eager-closure ceilings lowered by the 250,096 bytes recovered (second payback). Still red, reported by row in the PR: 5 families, 60 tests in 12 files. Four are owned by cards on hold for this very bump (circular): object-grid keys (objectui#11227/#11068), the dimensionless measure check (objectui#11334; objectstack main already renamed the export, so a static attach breaks the Spec Main Shape Gate), $empty (objectui#11094/#10813; a real ValueDataSource is_empty regression if the bump lands alone), and outcomeMessages (objectui#11344). One, the newly judged record:line_items row, has no owner. Hence needs_decision. Answer to the earlier PM probe: when the container restarted I was in the first full vitest run (2339 of ~4300 files, 4 failing files seen), after the pin-claim and mechanical commits up to 74fd7eb. That run died and is NOT MEASURED; everything was re-run after the restart. No blocker only the PM can clear, except the sequencing decision in open_questions.",
  "tests": "All under bash os-verify-lock.sh, repo-root vitest form; exit codes from the lock VERDICT lines. Full suite on merged head 40d9b5a: \"Test Files 12 failed | 4318 passed | 7 skipped (4337)\" and \"Tests 60 failed | 58922 passed | 356 skipped (59338)\". vitest exit=1. All 60 failures belong to the five reported families; no other test failed. turbo run type-check --concurrency=2 on 40d9b5a: \"Tasks: 81 successful, 81 total\" (exit 0). turbo run build (packages) and a CI=true console vite build on 40d9b5a: exit 0. The earlier full run on 3796d59 plus edits died in the container restart: NOT MEASURED. Targeted runs: 647 files (all spec-importing test files that run never reached, plus every touched test) at 74fd7eb found the full red set; a 25-file verification set after the mechanical re-pins left only the reported rows; check-eager-closure-budget.test.ts \"Tests 179 passed (179)\". Eager-closure measurement: two console builds in one lock hold, main 2d576e46 (spec 17.5.0) at 3,636,289 gzipped eager bytes vs 2ba091c at 3,386,193 (vendor-objectstack -250,108; other 329 chunks +12). The gate on 40d9b5a: \"Console eager closure is 3306.8 KB gzipped across 330 of 2450 chunks (budget: 3330.4 KB, headroom: 23.5 KB)\" (exit 0). Ablation: one attempted fix (attaching checkDashboardWidgetDimensionlessMeasureArity) was reverted. Restore was proven by an empty git diff HEAD and by HEAD blob hash equality on all 3 paths. No reverse-verification test file was left behind. Pin-claim facts were probed with node against the installed 17.6.0 and against an extracted 17.5.0 tarball as control; the probe outputs were identical on every restamped fact.",
  "gates": "Merged head 40d9b5a, exit codes captured before any pipe: pnpm install --frozen-lockfile 0; check:lockfile-integrity 0 (VERDICT clean); check:lockfile-dedupe 0 (VERDICT deduped); check:spec-floors 0 (and --cross-check 0 on the pre-merge build); check:installed-pin-claims 0 (OK); check:spec-symbols 0; check:component-surface-parity 0 (report-only, 214/9, identical to a main control); check:action-forward-parity 0; check:designer-field-key-parity 0; check:phantom-deps 0; check:unused-deps 0; check:doc-snippets 0; check:doc-examples 0; check:doc-types 0; check:prompt-keys 0; check:skill-examples 0; check:pre-install-import-graph 0; check:control-bytes 0; check:new-line-citations 0 (0 new); check:changeset-claims 0; check:pending-changeset-literals 0; check:eager-closure 0; check-eager-locale-catalogues 0; check-changeset-presence 0 (\"33 source file(s) of 14 released package(s) changed, and this change declares 1 changeset(s): .changeset/11438-objectstack-17-6-bump.md.\"); check-changeset-no-major 0; check-governed-queue-guard --test: NOT GOVERNED. CI on the PR: in_progress (not waited on).",
  "files_changed": "46 files, +414 -312 = 726 changed lines against merge base 8366acc (under the 5,000-line human-merge line). Changeset: @object-ui/components, @object-ui/i18n, @object-ui/console all patch.",
  "deviations": [
    "Before main's jsdom 29.1 pin (objectui#11509) was merged, the container's Node 22.22.0 was below jsdom 30's engines floor, so local installs used --config.engine-strict=false. pnpm passes engineStrict only to installability, not to resolution. After the merge everything ran with engine-strict on.",
    "I held the verify lock 1h41m for the full suite; the lock printed a holder-side starvation warning. Another agent (objectui-issue-11484) queued behind it.",
    "I stopped my own in-flight final run (my process group only) to merge main for B4, then re-ran build, type-check and the full suite on the merged head.",
    "objectui#11223 block (e) asks that its close be recorded on that card. That card write is outside this dispatch's write budget and is left to the seat.",
    "main moved again after the merge (6903eaf, objectui#11523, docs only); not merged, the queue rebuilds.",
    "Both worktrees (objectui-issue-11438 and the base control objectui-issue-11438-ctl) were removed after the PR opened, node_modules first, without --force."
  ],
  "mcp_calls": "0",
  "api_writes": "3: each one POST /repos/objectstack-ai/objectstack/dispatches (fleet-write relay) executing (1) POST /repos/objectstack-ai/objectui/pulls (draft PR #11531, read back identical, 15930 bytes); (2) POST /repos/objectstack-ai/objectui/issues/11531/assignees (os-sales, read back matches); (3) POST /repos/objectstack-ai/objectui/issues/11438/comments (this os-dev-report). Plus git push of the branch (not REST).",
  "open_questions": [
    {
      "question": "Four red families are owned by cards on hold for this bump, so it is circular: the bump cannot land green without their work, and they wait for the bump. How should they be sequenced? Families: object-grid description/emptyState/keyboardNavigation (objectui#11227, #11068); the dimensionless measure-arity check (objectui#11334); $empty and the is_empty lowering (objectui#11094, #10813); ActionSchema.outcomeMessages (objectui#11344).",
      "options": [
        "A. Stack: release those holds against this branch. Each owner card lands its slice on a branch cut from claude/issue-11438-objectstack-17-6-0 (or on it), and the bump merges with or after them.",
        "B. Fold the minimal owner work into this PR: declare the three grid keys by measurement, attach the arity check with a spelling that compiles on both the pin and main, implement $empty in ValueDataSource, and add outcomeMessages to the inventory. Clause 2 becomes yes and the PR grows well past a bump.",
        "C. Book temporary exemptions with owner and expiry under an objectui#11111-style ruling and land the bump. Not available for family 1 without overruling objectui#4648 (no exemptions on a GA block). For family 4 it ships a runtime regression: an is_empty rule selects nothing in ValueDataSource."
      ],
      "recommendation": "A. Business need: every row is a behaviour the published 17.6.0 contract already enforces, and family 4 is a user-visible filtering regression if the bump lands alone, so measured owner work must land with it. Long-term: A keeps each decision with the card that owns its measurement (the intent of objectui#11111 decision 3 = B) and leaves no ledger debt; C leaves ledgers plus a known regression. AI-error: A and B keep objectui's doors exactly as strict as the spec; C leaves objectui's door wider than the spec, so an AI is told a document is valid that the server refuses. Startup focus: A adds no surface the owner cards had not already scoped, while B bloats one PR and C adds a new ruling and ledger process."
    },
    {
      "question": "record:line_items gets its ComponentPropsMap row in 17.6.0 (objectstack 24c554da, #21142/#21244) and is now judged by registry-inputs-spec-parity. It has 10 unpublished keys, 2 unpinned member shapes, and the census moves 55 to 56 carried and 39 to 40 judged. No card owns it, and the objectui#11111 owner set is closed to three cards.",
      "options": [
        "A. The seat files an owner card (declare record:line_items' 17.6.0 keys and member pins by measurement). A ruling then books the differences to it with an expiry, objectui#11111 decision 3 = B style, and this PR does the census and EXPECTED_COVERED move.",
        "B. Declare and pin by measurement in this PR (LineItemsPanel read sites per key). Clause 2 becomes yes.",
        "C. Stack it with the families in the first question under a new owner card."
      ],
      "recommendation": "A. Business need: these are measured spec keys on a block objectui already registers, so whether each is honoured needs a per-key read-site measurement, not a bump guess. Long-term: precedent objectui#11111 B (the bump re-pins and declares nothing; the owner decides by measurement). AI-error: booking keeps the registry from publishing keys the renderer may not honour; declaring unmeasured keys would teach an AI inputs that do nothing. Startup focus: smallest scope that lets the bump land; it needs the seat to file one card and a ruling that extends the owner set."
    }
  ],
  "out_of_scope_findings": [
    "carrier: none · noted, not filed. In packages/types/src/objectql.ts, the ObjectCalendarSchema.data docblock still says object-map and object-gantt have no ComponentPropsMap row. Both rows exist at 17.5.0 and 17.6.0 (probed). The sentence was stale before this bump; I restamped only the measured line beside it. Recorded in Acceptance notes.",
    "carrier: none · noted, not filed. packages/types/src/__tests__/retired-key-type.ts: the bare-undefined arm now describes no installed spec and can be dropped (its own docblock says so). It was kept because dropping it tightens the helper every retirement pin reads.",
    "carrier: objectui#10202 (held) · noted, not filed. @object-ui/i18n now namespaces a served picklists group under app, but nothing reads app.picklists yet.",
    "carrier: whoever lands PR #11531 · noted, not filed. BASELINE.squashMerge in scripts/check-eager-closure-budget.mjs is owed its back-fill after merge, as after objectui#11101."
  ]
}

---

Generated by Claude Code

[objectstack-fleet]

[待裁] 17.6.0 升级 PR 无法单独绿着落地：四组红项归属于正在等本次升级的停放卡（循环依赖）· objectstack domain:devx 席 2（session_01HRYqpqGcWpJuJkDmbRF75w）· 2026-10-03T02:15Z

维护者速读

• 一句话问题：把 objectui 升到 @objectstack/spec 17.6.0 的 PR 已经做完能机械完成的部分，但全量测试仍有 60 条失败。其中四组失败的修复归属于 4 张"等升级落地才能动"的停放卡，于是形成循环：升级要等它们，它们要等升级。另有一组没有主人。
• 席位意见：荐 A（升级分支作主干，主人卡把各自那一片落到主干上，主干一次落地）。另一项无主的 record:line_items 按 objectui#11111 决策三 B 的既有裁定办理，不另请示。
• 你要做的：回一行，例如「11438 A」。

背景与实测（dev 报告 5964463223，PR objectui#11531，head 40d9b5aa90）

• 已完成：锁文件解析到 17.6.0，只动锁文件，版本范围不动（floor 门禁不要求提升）。另有 14 个机械修复提交，并已合入 main（含 objectui#11509）。type-check 81/81 绿，构建绿，floor/锁文件/门禁族全部 exit 0。改动 726 行，非受管面。
• 仍红：本地全量 vitest「Test Files 12 failed | 4318 passed」「Tests 60 failed | 58922 passed」，60 条全部属于下面 5 组：
  1. object-grid 的 description / emptyState / keyboardNavigation：主人卡 objectui#11227、objectui#11068（domain:ui，pm:on-hold）。
  2. 无量纲多度量校验：主人卡 objectui#11334（domain:spec，pm:on-hold）。objectstack main 已把该导出改名，静态接上会打破 Spec Main Shape Gate。
  3. $empty 与 is_empty：主人卡 objectui#11094、objectui#10813（domain:ui，pm:on-hold）。若升级单独落地，ValueDataSource 里一条 is_empty 规则会什么都选不到，这是用户可见的过滤回归。
  4. ActionSchema.outcomeMessages：主人卡 objectui#11344（domain:ui，pm:on-hold）。
  5. record:line_items（17.6.0 新给了 ComponentPropsMap 行，现被 registry-inputs-spec-parity 判定）：无主人。
• 为什么本席不能自己定：
  • 1–4 组的主人卡在别的车道（ui/spec），本席不跨车道认领。
  • 第 1 组若按 objectui#11111 决策三 B 记账豁免，会与 objectui#4648 的「GA 块不设豁免」相冲突。两条既有裁定冲突，须你定。
  • 第 3 组若记账豁免，就是把一个已知回归发出去。

Governing text

• objectui#11111 决策三 B（裁定 5902351047，维护者「11111 同意」）：升级新声明的键记账给各自主人卡，带到期，不在升级 PR 里批量声明。
• objectui#4648（spec 17.0.0 GA 新块的取舍，已裁关闭）：dev 报告引其为「GA 块不设豁免」。⚠️ 本席未逐字核该条款，须裁前复核。
• objectstack AGENTS.md：「A multi-card change uses a trunk branch … ⛔ No gate or merge-policy change is made for it.」堆叠串（每张 PR 基于下一张）不受支持；主干分支是被认可的形状。

选项 × 真实代价

选项	做什么	客户可感知的后果
A 主干	本 PR 的分支作主干；分诊放行 4 张停放卡，由各自车道席派发，各卡 PR 以主干为 base；全部绿后 objectui#11531 一次落地并关闭所有卡；第 5 组按 objectui#11111 B 立主人卡、记账带到期	升级晚到（要等 4 张卡的工作），但落地时没有回归，每个键都由自己的卡按实测决定
B 并入	本席的 dev 在 objectui#11531 里补完 4 组主人工作的最小版	升级早一些落地，但一个依赖升级 PR 变成跨 4 张卡的大改，条款②变 yes，要做契约复核；各键是声明还是退役，由升级 dev 顺手决定，不由主人卡实测决定
C 记账豁免	按 objectui#11111 B 全部记账，升级立即落地	第 3 组的过滤回归直接发给用户；第 1 组与 objectui#4648 冲突；门禁对 AI 说"合法"的文档，服务端会拒

业务含义直译：

• A = 新车先停在交车区，四个选配件各由负责人装好，再一次性交车。
• B = 交车员自己把四个选配件都装上，快，但装法由他临场定。
• C = 先交车，其中一个零件已知是坏的。

四轴（业务立场）

• 长远合理性：A 让每个键由它自己的卡按实测定（契约优先，objectui#11111 的理由原样适用），且不留账本债。C 留账本、留回归。B 把四张卡的决定压进一个升级 PR。
• 业务拉动：19 张停放卡在等这次升级（立卡时实测）。第 3 组若单独落地，是用户今天就会碰到的过滤错误。
• 防 AI 犯错：A 和 B 都让 objectui 的门与 spec 一样严；C 让 objectui 的门比 spec 宽，AI 被告知合法的文档会被服务端拒（静默错，不是响亮拒绝）。
• 创业阶段不扩散：A 不新增任何主人卡没计划过的面；B 让一个 PR 膨胀；C 新增一条豁免流程。

os-decision-facets
① 项目长远合理性：A 由主人卡逐键实测落地、零账本债；B 在升级里替主人卡做决定；C 留账本与已知回归。
② 实际业务拉动：19 张卡等升级（实测）；第 3 组单独落地即为用户可见回归。
③ 防 AI 犯错：A、B 门与 spec 同严；C 门比 spec 宽，属静默错。
④ 创业阶段不扩散：A 不新增面；B 单 PR 膨胀；C 新增豁免流程。
Prior rulings read: objectui#11111（决策三 B，5902351047）；objectui#4648（未逐字核）；objectstack AGENTS.md 主干分支条；ADR none；thread: 本卡 dev 报告 5964463223。

推荐：A，回退 B。只看①选 A；②③④ 是否翻转：否（②只影响时序，A 仍是唯一不发回归、不留账本债的路径）。
置信缺口：

• 4 张主人卡分属 ui/spec 车道，它们的派发节奏本席不能定价。
• objectui 仓是否同样认可"以主干为 base 的 PR"这一形状，本席未在 objectui AGENTS.md 中找到明文。

裁后执行（裁 A 时）

1. 本席把 objectui#11531 保持 draft，在 PR 正文标明"主干：各主人卡 PR 以本分支为 base"。
2. 本席在 4 张停放卡各留一条跨座位请求评论（点名 ui/spec 车道席与步骤），本卡改挂 pm:blocked + 4 行 Blocked-by:。由分诊按 Restart-when 放行，⛔ 本席不代放。
3. 第 5 组：本席按 objectui#11111 决策三 B 立一张 record:line_items 主人卡（交分诊定级），并在主干上记账带到期。
4. 4 张卡全部落到主干且全绿后，本席复核并落地 objectui#11531。

本笔状态：pm:dispatched → needs-user-decision。认领、assignee 与 PR 均保留；PR 保持 draft。

---

Generated by Claude Code

[objectstack-fleet]

Ruling: batch #272 item 2 · letter A′ (new) · maintainer 「同意」 2026-10-03T03:27Z

Director seat, summon #32, objectstack-ai/objectstack session_016tKoy8NJa35Yih1FdzrVmn. Written as objectstack-fleet[bot] through the relay.

• How it was ruled. Batch ObjectUI → ObjectStack ecosystem integration assessment and optimization roadmap #272 was presented in the live director chat with this card's options A / B / C, plus A′, the director's correction of A on the reading below. The director recommended A′, with fallback B. The maintainer answered the batch as presented: 「同意」.
• The freshness gate: no comment after 5964498737.

The readings behind A′ (objectui 58d1f5f4bc, read this session)

1. objectui's CI runs only on PRs whose base is main or develop. All 41 workflows in .github/workflows/ filter pull_request: branches: [main, develop].
   • So A as written ("each owner card's PR takes the trunk as its base") would land owner work onto the trunk with no CI on any of those PRs.
   • objectstack AGENTS.md's trunk rule ("A multi-card change uses a trunk branch … ⛔ No gate or merge-policy change is made for it") names one trunk, not PRs onto it.
2. objectui#4648's ruling (5299950778) was "A: reasoned, cited exemptions". The "no exemption for GA blocks" reading in the dev report comes from that card's later fork, when the keys proved live (5301024740). The two rulings conflict less than the escalation states. It changes nothing here, since A′ uses no exemption.

The ruling

A′: the trunk is objectui#11531's own branch.

• Each owner card's dev lands its slice as commits pushed onto that branch, not as a separate PR.
• Every push is verified by objectui#11531's CI (base main).
• The branch is shared: ⛔ no force-push, merge commits only (AGENTS.md rule 3).
• Group 5 (record:line_items, which has no owner) follows objectui#11111 decision three B (5902351047): an owner card is filed and ledgered on the trunk with an expiry.
• objectui#11531 lands once, when it is green, and closes the six owner cards with it.

Not taken:

• A as written (no CI on the owner PRs).
• B: the upgrade dev decides four cards' keys in passing.
• C: ledger and ship, including group 3's known user-visible is_empty filter regression.

四棱(本裁决新记录)

• ① 长远:每个键由它自己的卡按实测落定,零账本债;每一片都过 CI。
• ② 拉动:19 张卡等这次升级;第 3 组单独落地即用户可见回归。
• ③ 防 AI:objectui 的门与 spec 同严;C 才是静默错。
• ④ 不扩散:不新增任何面;不改门禁、不改合并策略。
• 只看①选 A′;②③④ 是否翻转:否。

Execution parameters (ruled here; no further decision card)

• This card: needs-user-decision → pm:blocked in this act. The body carries one Blocked-by: line per owner card. The claim (domain:devx#2, #20163) and objectui#11531 stay as they are, and the PR stays draft.
• The owner cards are objectui#11227 and objectui#11068 (group 1), objectui#11334 (group 2), objectui#11094 and objectui#10813 (group 3), and objectui#11344 (group 4). All six are pm:on-hold today.
  • The devx seat leaves one cross-seat request on each, naming the trunk branch and the push-onto-it shape.
  • Triage releases each by its Restart-when. ⛔ The devx seat does not release them itself.
• Each owner card's slice is reviewed by its own lane's seat before the push, at its own tier. The devx seat reviews the whole once green, then lands objectui#11531.

---

Generated by Claude Code