[objectstack-fleet]

Requested by Tesla · project thread

Fixes #26553

Clause-②: no
Metadata-protocol: none

维护者速读

• 改了什么：ADR-0050（表单布局 vs 呈现）的状态行从 Proposed 改为 Accepted，写明你在 [Decision] ADR-0050 form layout vs presentation — the maintainer's own call: schedule the spec-major narrowing of FormView.type to layout only, defer it, or withdraw the record (set aside from ADR batch 2 on #26553) #27032 的 A′ 裁决、裁决记录号 6464210455 和两张切换卡（spec(view): narrow FormView.type to layout only (simple | tabbed | wizard) — retire split, drawer, modal with tombstones (ADR-0050 Accepted, #27032 letter A′) #27155 spec、ui(plugin-form): ObjectForm drops the split / drawer / modal branches and SplitForm retires; showcase task views re-spelled (ADR-0050 cutover, renderer half of #27155) #27159 renderer），2026-10-20 的修订文字原样保留在裁决句之后；docs/adr/README.md 由 pnpm adr:index 重新生成（Proposed 5 → 4，Accepted 130 → 131）。
• 为什么改：你 2026-10-28 已裁 ADR-0050 接受，两张切换卡都已完成关闭，但记录和索引仍写着 Proposed；读索引的人和从这份 ADR 切卡的人读到的都是过期状态。这是 docs(adr): the 19 stale Proposed records (status date before 2026-07-26) are ruled by the maintainer in four batches of five — carrier for the rulings and the status-line PRs #26553 十九份记录里最后一份未收尾的。
• 风险与回滚：只改一行状态文字加生成的索引，不动任何 schema、代码或 ADR 正文；回滚即 revert 本 PR。
• 先看：docs/adr/0050-form-layout-vs-presentation.md 第 3 行（裁决原句在前，原修订文字在后）；docs/adr/README.md 第 20–21 行的计数与第 123 行 ADR-0050 那一行。
• 两分钟验证：node scripts/adr-index.mjs --check，预期打印 adr-index: OK — docs/adr/README.md is current and every record parsed，退出码 0。
• 需要你做的：本 PR 触及 docs/adr/**（Tier H），席位复审后由你批准合并。

What changes

docs/adr/0050-form-layout-vs-presentation.md, line 3 only. The status line read:

**Status**: Proposed (2026-06-15) · **Amended** (2026-10-20, letter B′ on #24692, ruling record 6281834942, maintainer 「同意」) — the list add-record open mode is no longer declared on addRecord; see the note under Decision 2.

It now opens with the ruled sentence, verbatim from the pointer on the card (issue 26553, comment 6464456446), and keeps the amendment text after it:

**Status**: Accepted (2026-10-28, letter A′ on #27032, ruling record 6464210455) — cutover card #27155 (spec) + #27159 (renderer); the 2026-10-20 amendment stands · **Amended** (2026-10-20, letter B′ on #24692, ruling record 6281834942, maintainer 「同意」) — the list add-record open mode is no longer declared on addRecord; see the note under Decision 2.

The amendment's links (#24692 and its ruling record) are kept, not dropped: the ruled sentence says the amendment stands, and the links are the only place the record names where that amendment was decided. Dropping them would leave a standing amendment nobody can trace (Prime Directive 13). No other line of the record asserts a Proposed state: a case-insensitive grep for proposed over the file hits line 3 alone before the edit and nothing after it.

docs/adr/README.md is regenerated by node scripts/adr-index.mjs (pnpm adr:index), never hand-edited: the ADR-0050 row moves from Proposed to Accepted, its date column reads the first ISO date of the new status value (2026-10-28), and the summary counts move to Accepted 131 / Proposed 4. The generator reads Accepted as the keyword because it is the earliest vocabulary entry in the value; the bold Amended sits well past the 80-character keyword window and would lose on position anyway.

No changeset. docs/adr/** ships in no released package (no files[] under packages/ names docs/adr), so the PR publishes nothing and carries skip-changeset, declared on the create.

Tests

All readings on head ac204886 in a fresh worktree off origin/main 1aeeafd9.

• node scripts/adr-index.mjs wrote docs/adr/README.md (237 lines); node scripts/adr-index.mjs --check exit 0: adr-index: OK — docs/adr/README.md is current and every record parsed.
• pnpm check:adr-index exit 0 (part of the derived gates below).
• git diff --shortstat origin/main...HEAD: 2 files changed, 4 insertions, 4 deletions.
• The edit was applied by a script asserting exactly one hit of the old line at line 3 and exactly one hit of the new line afterwards (old hits 0, new hits 1).

Acceptance notes

• Neighbours. Four open PRs touch docs/adr/ records on their heads against their merge-base with origin/main 1aeeafd9: docs(adr): amend ADR-0142 D5 — an organization authors its own assignment rules, evaluated before the declared ones; answer merge question 11 #30828 (ADR-0142), ci(release): start the next snapshot preflight on a daily schedule, stopping at the maintainer approval (ADR-0125 amendment) #30862 (ADR-0125), docs(adr): ADR-0146 amended in place — an upgrade plan may report a .custom override that hides a vendor change, as advice only #30899 (ADR-0146) and feat(spec,security): a controlled_by_parent object may declare an optional master #30940 (ADR-0055). None of them touches docs/adr/README.md. Whichever of them and this PR lands later merges main and re-runs pnpm adr:index if a status line moved.
• Claim surface held. The record's body (Decision, Migration, Consequences) is untouched: the claim names the status line alone, and the cutover's completion is recorded by the two closed cards the line names, not by rewriting the record.
• scripts/adr-index.mjs --check is the gate's spelling; CI's lint ADR step calls --report, which warns on a stale index and never reds. The local --check reading above is the one that would.

Verification

Derived gates: 21 derived, 21 run, 0 NOT MEASURED, 0 UNRUN (head ac20488)

node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 21 commands from the committed change set (2 paths vs merge-base 1aeeafd9; changed lines 8, under the 3000 human-merge threshold). Each ran in the foreground with stdin closed, its exit code landed on disk before any pipe; every one exited 0:

• node scripts/check-adr-links.mjs and its --self-test
• node scripts/check-closing-keyword-parity.mjs and its --self-test
• node scripts/check-comment-mask-corpus.mjs
• node scripts/ci/plan-test-core-shards.mjs --self-test
• pnpm --filter @objectstack/lint run check:doc-formula-expressions (the lint closure was built first under scripts/pm/os-verify-lock.sh: VERDICT command-exit 0, 96 s held)
• pnpm check:adr-anchors, pnpm check:adr-index, pnpm check:adr-symbol-anchors
• pnpm check:ci-filter-parity, pnpm check:closed-ceremony-text, pnpm check:cross-package-test-inputs
• pnpm check:doc-authoring, pnpm check:driver-memory-census, pnpm check:gitlink-declared
• pnpm check:nul-bytes, pnpm check:pm-governed-merges, pnpm check:pm-prior-rulings
• pnpm check:refd-timer-probe, pnpm check:watch-hint-literal

Reconciliation: dispatch-gates.mjs --repo objectstack-ai/objectstack --ran printed 21 derived famil(ies) accounted for — 21 run, 0 NOT-MEASURED. Tier: dispatch-gates --tier names CONTRACT_REVIEW_TIER as mandatory from docs/adr/**; check-governed-merges.mjs --test lists both paths as governed.

Tier H by path (docs/adr/**). It stays a draft: the spec seat runs the contract-tier review, then hands the PR to the approver; no seat readies, queues or arms it.

---

Generated by Claude Code

[objectstack-fleet]

domain:spec#1 · os-tesla · session_01L9qibzzqHZdAVqgBAvxXKB

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: ac204886c44d6e03cc7bb775094e5145ae0a58c7
Local-runs: none
Scope: docs/adr/0050-form-layout-vs-presentation.md, docs/adr/README.md + scripts/adr-index.mjs (the generator the README claims to be the output of), packages/spec/src/ui/view.zod.ts (FormViewSchema.type, the ADR's first consumer), ui/packages/plugin-form/src/ObjectForm.tsx + ObjectFormContainer.tsx (the renderer half), .github/workflows/lint.yml + package.json (check:adr-index is report-only), card #26553 comments 6458958232 / 6464456446 / 6688244732 / 6688746866, ruling 6464210455 on #27032, cards #27155 / #27159 and their PRs #27229 / #27305

① Derived judgments

• Accept set: unchanged — right. The net diff against merge-base 1aeeafd9 is two files under docs/adr/** (+4/-4); no packages/** path, no exports map, no .d.ts. Clause-②: no on the PR body is right.
• Governed register (ADR surface): ADR-0050 moves Proposed → Accepted — right. The maintainer's ruling 6464210455 on [Decision] ADR-0050 form layout vs presentation — the maintainer's own call: schedule the spec-major narrowing of FormView.type to layout only, defer it, or withdraw the record (set aside from ADR batch 2 on #26553) #27032 (A′, 「其他同意」, 2026-10-28) accepts the record; the director seat's pointer 6464456446 on docs(adr): the 19 stale Proposed records (status date before 2026-07-26) are ruled by the maintainer in four batches of five — carrier for the rulings and the status-line PRs #26553 prescribes the exact sentence for this closeout PR. Line 3 of the record opens with that sentence verbatim: **Status**: Accepted (2026-10-28, letter A′ on #27032, ruling record 6464210455) — cutover card #27155 (spec) + #27159 (renderer); the 2026-10-20 amendment stands.
• Index keyword and counts — right. scripts/adr-index.mjs statusKeywordOf strips the markdown emphasis, backtick and blockquote characters, then takes the vocabulary entry with the lowest index within KEYWORD_WINDOW = 80; Accepted sits at index 0, and the retained Amended sits past index 150, outside the window and later anyway. Summary table Proposed 5 → 4, Accepted 130 → 131, Total 149 unchanged: exactly one record moved keyword and no other record's status line differs between the merge-base and head, nor between the merge-base and main acd190b2 (empty git diff on docs/adr/*.md status lines and on README.md).
• Index row — right, derived by hand from the generator. No **Date**/Decided header field in the first 40 lines, so date falls back to the first ISO date in the status value: 2026-10-28. statusCell drops the keyword, unwraps the leading parenthesis and cuts the detail with cell(detail, 64) = 63 chars + …; the 63-char prefix of 2026-10-28, letter A′ on #27032, ruling record 6464210455 — cutover … ends at — cut, giving the committed cell Accepted — 2026-10-28, letter A′ on #27032, ruling record 6464210455 — cut…. Supersedes/Superseded-by stay — (no such fields, no superseded word in the value). Line 123 of the head README is byte-for-byte that row.
• CI reading of the same fact. pnpm check:adr-index is --self-test && --report and never reds (lint.yml L697–700, continue-on-error: true), so its green is not currency evidence on its own; but the step ADR index is current (report-only …) in check-run Lint · repo & tooling gates (127983466962) completed with zero ::warning annotations on this head (the only annotation is the ubuntu-latest notice), which is the --report verdict that the committed index equals the rendering and every record parsed. The dev's --check exit 0 is the PR body's account and is not relied on.
• No other surface is implied. The record's Decision / Migration / Consequences text, the Consumers line and the 2026-10-20 note under Decision 2 are untouched.

② Semver level

skip-changeset label is on the PR (documentation, size/xs, skip-changeset); Check Changeset check-run success. docs/adr/** ships in no released package, so nothing is published and no changeset is owed — matches the diff. Clause-②: no — right; this is not the narrowing face (the narrowing itself shipped under #27229 / #27305 with their own declarations).

③ Boundary flags

• open_questions: [] — none to answer.
• Dev deviations (3): git fetch --deepen for the neighbour check, PR-event subscription instead of polling, worktree kept until CI is read — none touches the committed tree. Answered: no effect on the review.
• Neighbour flag: four open PRs touch docs/adr/ records (docs(adr): amend ADR-0142 D5 — an organization authors its own assignment rules, evaluated before the declared ones; answer merge question 11 #30828, ci(release): start the next snapshot preflight on a daily schedule, stopping at the maintainer approval (ADR-0125 amendment) #30862, docs(adr): ADR-0146 amended in place — an upgrade plan may report a .custom override that hides a vendor change, as advice only #30899, feat(spec,security): a controlled_by_parent object may declare an optional master #30940), none touches README.md; main has not moved a status line since the merge-base, so no regeneration is owed today. Whichever lands later re-runs pnpm adr:index (the lint step only warns on a stale index). Answered.
• Author: PR user.login is objectstack-fleet[bot] — no login note owed.
• Tier H by path (docs/adr/**): the PR is a draft, mergeable: true; the approver's hand is the maintainer's. Not this record's to arm.
• Check-runs on the head at record time: every completed run is success or skipped (Governed Surface Queue Guard, PR-body guards, Check Changeset, Check PR Size, Check Documentation Links, every Lint and Type Check job, Test UI, Dogfood Regression Gate); Test Core (1/8) was still in_progress when this record was rendered — a core test shard that no docs line can red; landing rule ② (all checks green) is read again at queue time, not here.

④ Design judgment

docs/adr/0050-form-layout-vs-presentation.md against its first consumer and its execution plan.

• The status is true of the tree. Spec half (spec(view): narrow FormView.type to layout only (simple | tabbed | wizard) — retire split, drawer, modal with tombstones (ADR-0050 Accepted, #27032 letter A′) #27155 → PR feat(spec,app-shell)!: a form view's type is its layout only — simple, tabbed or wizard; split, drawer and modal are retired with pane, splitSize and splitResizable #27229, merged 2026-10-28T12:16Z, merge commit 14e9685fc9): on main, view.zod.ts L5151–5156 declares type: enumWithRetiredValues(['simple','tabbed','wizard'], FORM_VIEW_TYPE_RETIRED).default('simple'), with tombstones for split / drawer / modal naming ADR-0050 and @objectstack/spec 18.3.0 (L4997–5016), and section.pane, splitSize, splitResizable retired with split (L3919, L4455, L5020–5029). Renderer half (ui(plugin-form): ObjectForm drops the split / drawer / modal branches and SplitForm retires; showcase task views re-spelled (ADR-0050 cutover, renderer half of #27155) #27159 → PR feat(plugin-form,spec)!: ObjectForm lays the form out, the object-form block opens drawer/modal, split retires #27305, merged 2026-10-29T09:40Z, a33fce29): SplitForm.tsx is gone from ui/packages/plugin-form/src, ObjectForm.tsx has no split / drawer / modal branch, and the drawer/modal arms live in ObjectFormContainer.tsx as caller-supplied containers — the shape ruling A′ ordered. Both cutover cards are closed completed. "Accepted" describes the live tree, not a plan.
• The line parses and the ruled sentence is intact. The dev kept the old amendment clause (· **Amended** (2026-10-20, letter B′ on #24692, ruling record 6281834942, maintainer 「同意」) — …) after the ruled sentence. Judged right, not a deviation: the ruled sentence itself says "the 2026-10-20 amendment stands", and the retained clause is the only place the record links where that amendment was ruled (Prime Directive 13: a standing amendment stays traceable). The keyword read is unaffected (Accepted at index 0; Amended outside the 80-char window), the first ISO date is the ruling date, and the generator's row shows the ruled sentence, not the amendment. The redundancy ("amendment stands · Amended (…)") is cosmetic.
• Should-fix (outside this PR's claim, for a follow-up card, not a FAIL): the body's Migration ("spec-major: the removed values are a breaking surface change") and Consequences / Sequencing ("implementation is a spec-major change") still describe the cutover as spec-major, while ruling A′ (6464210455) shipped it as an 18.3.0 (narrowing) minor because this repository has no spec-major window (AGENTS.md L1084–1090). The status line cites that ruling record, so the fact is reachable, and the card (docs(adr): the 19 stale Proposed records (status date before 2026-07-26) are ruled by the maintainer in four batches of five — carrier for the rulings and the status-line PRs #26553 done-when) and the claim's file surface (6688244732: "status line :3 only, ⛔ any other ADR body text") forbid touching the body in this PR. A one-line dated note under Migration, in the shape of the 2026-10-20 note under Decision 2, would close the gap; it is not this PR's to add.
• Tried to break: a status value with the keyword past the window, a second vocabulary word earlier than Accepted, a link whose text changes the cut — none applies; Accepted is the first token after **Status**: and every link sits beyond the 64-char cut.

docs/adr/README.md against its generator. Generated output, judged only for being the generator's: the three hunks (two summary counts, one row) are exactly what renderIndex produces for the one changed record (derivation in ①), the banner and every other row are untouched, and the head's report-only CI step raised no stale-index warning. Holds.

No must-fix.

Implemented-by: claude/project-thread-uz8dqb
Reviewed-by: session_01L9qibzzqHZdAVqgBAvxXKB

VERDICT: PASS

[objectstack-fleet]

domain:spec#1 · os-tesla · session_01L9qibzzqHZdAVqgBAvxXKB

维护者速读

• 改了什么：ADR-0050 状态行从 Proposed 改为 Accepted，开头逐字使用 docs(adr): the 19 stale Proposed records (status date before 2026-07-26) are ruled by the maintainer in four batches of five — carrier for the rulings and the status-line PRs #26553 指针 6464456446 规定的句子（裁决 A′，[Decision] ADR-0050 form layout vs presentation — the maintainer's own call: schedule the spec-major narrowing of FormView.type to layout only, defer it, or withdraw the record (set aside from ADR batch 2 on #26553) #27032 裁决记录 6464210455；切换卡 spec(view): narrow FormView.type to layout only (simple | tabbed | wizard) — retire split, drawer, modal with tombstones (ADR-0050 Accepted, #27032 letter A′) #27155 + ui(plugin-form): ObjectForm drops the split / drawer / modal branches and SplitForm retires; showcase task views re-spelled (ADR-0050 cutover, renderer half of #27155) #27159）。原有 2026-10-20 修订说明及其链接保留在后面。docs/adr/README.md 由 pnpm adr:index 重新生成：Proposed 5→4，Accepted 130→131。
• 为什么改：19 条过期 Proposed ADR 已全部裁决，只剩 ADR-0050 的状态行没落地。读 ADR 索引的人和从 ADR-0050 切出的卡看到的还是 Proposed。
• 风险与回滚：只改文档两行状态，不发布任何包（skip-changeset）。回滚即 revert。
• 先看：docs/adr/0050-form-layout-vs-presentation.md 第 3 行。
• 两分钟验证：node scripts/adr-index.mjs --check，预期 exit 0。
• 需要你做的：Tier H（docs/adr/**）。请以 os-zhuang 批准 PR docs(adr): ADR-0050 status line reads Accepted (letter A′ on #27032); regenerate the index #30986，批准后席位落地。契约复审 PASS 记录 6688923854，Head-sha ac204886c44d6e03cc7bb775094e5145ae0a58c7。

席位意见：可以落地，待授权批准。

席位核对

ACCEPT · seat domain:spec#1 · head ac204886 · 2026-11-06T15:47Z

1. Shape: Fixes #26553, draft, base main, 2 files (docs/adr/0050-form-layout-vs-presentation.md :3, regenerated docs/adr/README.md), +4/−4. Label skip-changeset, Clause-②: no / Metadata-protocol: none, matching claim 6688244732.
2. Ruling: pointer 6464456446 (ADR-0050 ruled A′ on [Decision] ADR-0050 form layout vs presentation — the maintainer's own call: schedule the spec-major narrowing of FormView.type to layout only, defer it, or withdraw the record (set aside from ADR batch 2 on #26553) #27032, ruling 6464210455, maintainer 「其他同意」). The ruled sentence comes first verbatim. The dev kept the 2026-10-20 amendment clause after it; the ruled sentence says the amendment stands, so the seat accepts this.
3. Contract review: at CONTRACT_REVIEW_TIER by an isolated reviewer. Scope = hits (both ADR files) plus scripts/adr-index.mjs, view.zod.ts FormViewSchema.type, and the plugin-form renderer. PASS 6688923854 on ac204886, no must-fix. Its findings:
   • The index row and counts are byte-for-byte the generator's output.
   • Keyword Accepted parses first.
   • Both cutover halves are on main (PR feat(spec,app-shell)!: a form view's type is its layout only — simple, tabbed or wizard; split, drawer and modal are retired with pane, splitSize and splitResizable #27229, PR feat(plugin-form,spec)!: ObjectForm lays the form out, the object-form block opens drawer/modal, split retires #27305), so "Accepted" is true of the tree.
   • One should-fix outside this PR's file surface: the ADR body's Migration and Consequences text still says "spec-major", while A′ shipped an 18.3.0 (narrowing) minor. The seat will carry it as a follow-up and not widen this PR.
4. Gates / check-runs: dev: 21 of 21 derived gates exit 0, --ran reconciled (report 6688746866). On head ac204886: 46 runs, 26 success, 20 skipped, 0 failed, 0 running (docs-only diff, path filters skip the code lanes); mergeable_state clean. check-governed-merges --pr 30986: Tier H (docs/adr/** ×2), 8 changed lines.
5. Accepted deviations: amendment clause kept after the ruled sentence (see 2). Neighbour PRs docs(adr): amend ADR-0142 D5 — an organization authors its own assignment rules, evaluated before the declared ones; answer merge question 11 #30828, ci(release): start the next snapshot preflight on a daily schedule, stopping at the maintainer approval (ADR-0125 amendment) #30862, docs(adr): ADR-0146 amended in place — an upgrade plan may report a .custom override that hides a vendor change, as advice only #30899 and feat(spec,security): a controlled_by_parent object may declare an optional master #30940 touch other docs/adr records; none touches README.md. The later lander re-runs pnpm adr:index after merging main.
6. Landing: Tier H: pm:awaiting-merge + review requested from the approvers. ⛔ No ready, queue or auto-merge before an APPROVED review from a GOVERNED_APPROVERS account; after it, the seat lands.