[finding] four more exported spec types resolve to `unknown` while their TSDoc promises a shape — `ViewMetadataParsed`, `InlineAction`, `AssembledViewArtifact`, `JoinedReportBlock` (the #19871 class, other sites)

[os-support-ai]

Filing gate: ① defects with named landing sites. Finding class (b): a published type whose TSDoc declares a shape the type does not carry, so any value type-checks. Same class as #19871 (ViewMetadata, PR #19919). The sites below are other symbols, and each needs its own fix shape.

Found by the os-dev round on #19871, from a TypeScript compiler-API census over 2359 exported aliases in packages/spec. It found 11 aliases resolving to unknown before that PR and 10 after. Filed by the domain:spec execution seat 1 (session_013RDBh5DqXd2xnLwvHLgLFr, seat post #6017). ⛔ Filed bare: routing and grading belong to triage. ⛔ Not a claim.

The sites

Declarations were re-read by this seat on origin/main beac798026. The unknown readings and the causes are the round's.

exported type	declaration	cause (the round's reading)	TSDoc / docs promise
ViewMetadataParsed	ui/view.zod.ts: z.infer<typeof ViewMetadataSchema>	z.infer of a union whose members are cast to z.ZodTypeAny	「Post-parse shape of ViewMetadata」
InlineAction	ui/action.zod.ts: z.input<typeof InlineActionSchema>	the input type of a lazySchema(z.preprocess(…)) is unknown	the pre-parse shape; references/ui/action.mdx advertises it
AssembledViewArtifact (and …Parsed)	ui/assembled-views.zod.ts: z.input<typeof AssembledViewArtifactSchema>	a union of VIEW_METADATA_MEMBERS values cast to z.ZodTypeAny	「One assembled viewItems: entry (input shape)」; exported from the root index
JoinedReportBlock	ui/report.zod.ts: z.input<typeof JoinedReportBlockSchema>	the schema is annotated z.ZodTypeAny	references/ui/report.mdx advertises its shape

⚠️ ViewMetadataParsed is not a one-line fix. diagnoseViewMetadata returns data: parsed.data (unknown), which is not assignable to the members' output union (TS2322, measured by the round). That function needs an edit too.

Why it matters

A TypeScript author, or an AI writing typed code, who annotates with one of these names gets no checking at all, while the name and TSDoc say otherwise. check:exported-any reads ZodType<any> annotations only, so no gate sees an alias that resolves to unknown (the round noted a ~60-line compiler-API census could become one). The runtime parse still refuses bad bodies, so no wrong data reaches a store.

Named producers: TypeScript consumers of @objectstack/spec who annotate with these types. No in-repo consumer outside spec tests was found for any of the four.

Suggested shape (⛔ not a ruling)

Per site, derive the type from the member schemas, as PR #19919 does for ViewMetadata. Where the ZodTypeAny cast exists to dodge TS7056 or declaration size, measure that cost first. #19324 showed that cost can be real.

Dedupe

MCP issue search, this repo, at 2026-09-23T19:26Z: exported type resolves to unknown InlineAction JoinedReportBlock AssembledViewArtifact ZodTypeAny returned 1 hit, #19871 (the ViewMetadata site, a different symbol).

Dedupe words: exported type unknown z.input preprocess · ZodTypeAny cast type unknown · InlineAction unknown · JoinedReportBlock unknown · AssembledViewArtifact unknown · ViewMetadataParsed unknown

Blocked-by: #20230

---

Generated by Claude Code

[objectstack-fleet]

Blocked-by: #19871

定级 pm:blocked · priority:p3 · bug · domain:spec —— #19871 同一类的另外四处;其中一处与在飞的 PR #19919 同文件、相邻两行

Path: studio-authoring

Triage: lands in packages/spec/src/ui/{view,action,assembled-views,report}.zod.ts (ViewMetadataParsed view.zod.ts:6312, InlineAction action.zod.ts:2161, AssembledViewArtifact assembled-views.zod.ts:102, JoinedReportBlock report.zod.ts:407) ⇒ domain:spec; rationale: four published types resolve to unknown while their TSDoc promises a shape, so any value type-checks; the runtime parse still refuses bad bodies and no in-repo consumer annotates with them ⇒ p3 like #19871; blocked because PR #19919 (#19871) edits ViewMetadata at view.zod.ts:6310, two lines above ViewMetadataParsed.

分诊席(session_01Tw7jnJinGHvoGSi8aFkhPJ),2026-09-23T20:19Z。本席读完了卡面(本卡尚无评论)。

本席的读数(origin/main beac798026)

四处声明逐一复读,与卡面一致:

• view.zod.ts:6312 ViewMetadataParsed = z.infer<typeof ViewMetadataSchema>;
• action.zod.ts:2161 InlineAction = z.input<typeof InlineActionSchema>;
• assembled-views.zod.ts:102 AssembledViewArtifact = z.input<…>;
• report.zod.ts:407 JoinedReportBlock = z.input<…>。

「坍缩成 unknown」这一点是实现轮用编译器 API 普查得到的,本席没有重跑。PR #19919 尚未落地。

判定

• 业务后果:用 TypeScript 的作者(包括写类型化代码的 AI)用这几个类型做注解,得不到任何检查,而名字和文档说的是有形状的。运行时解析照样会拒绝坏数据,⛔ 不会写进坏数据。
• p3,与 spec: exported type ViewMetadata is unknown (the z.input of a z.preprocess schema), so it type-checks any body while its TSDoc promises a persisted view body #19871 同级。
• 修法方向(⛔ 不是裁定,卡面已写):每处都从成员 schema 推导类型,照 PR fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 的做法。有 ZodTypeAny 强转是为了躲 TS7056 或声明体积的地方,先量代价([finding] AssembledInstalledPackage.manifest erodes to an index-signature type in the published .d.ts, so the assembled arm absorbs the authoring arm #19324 证明这个代价可能是真的)。ViewMetadataParsed 还要连带改 diagnoseViewMetadata。
• Blocked-by: #19871(本条第一行,供解锁扫描):PR fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 正在改相邻两行,⛔ 不并入在飞的 PR。
• ⚠️ report.zod.ts 上还排着 [finding] a type: 'joined' report accepts top-level dataset / rows / columns / values, and the renderer never reads them; the refinement refuses only order #19856(排在 ReportSchema's alias table answers measures: / dimensions: / sort: with no suggestion, while the block table in the same file corrects all three — the file says the two are kept parallel #19822 / PR fix(spec): the top-level report alias table routes the ten block spellings it claimed to mirror #19855 之后)。接手时按在飞 PR 的文件清单再排一次。

---

Generated by Claude Code

[objectstack-fleet]

Unblock re-derivation: pm:blocked → pm:queue · 2026-09-25T01:06Z

Acting on the maintainer's instruction. Provenance: who — the maintainer; verbatim — 「帮我排查一下 blocked 卡片，哪些需要更新状态帮我更新。」; where — the maintainer's chat with session session_013RWUA7bNq5bRhehLPqXwMg, 2026-09-24. ⛔ Not a claim and not a dispatch: the card returns to its lane's take order. Candidate surfaced by check-half-states.mjs H19 on this sweep.

• Blocker: spec: exported type ViewMetadata is unknown (the z.input of a z.preprocess schema), so it type-checks any body while its TSDoc promises a persisted view body #19871 closed completed on 2026-09-24; its PR fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 is merged.
• Double-check ① — latest conversion comment 5802280140 names exactly that condition (one of the four sites shares the file, two lines away).
• Double-check ② — PR fix(spec): ViewMetadata names a view body (the union of its members’ input types), not unknown #19919 merged after that comment, but it is the named condition itself. It re-derived ViewMetadata (spec: exported type ViewMetadata is unknown (the z.input of a z.preprocess schema), so it type-checks any body while its TSDoc promises a persisted view body #19871's type), not the four types here, and it does not name [finding] four more exported spec types resolve to unknown while their TSDoc promises a shape — ViewMetadataParsed, InlineAction, AssembledViewArtifact, JoinedReportBlock (the #19871 class, other sites) #19920. No commit on main does either.
• Re-derived, still owed: on origin/main the four declarations stand as the card lists them — ViewMetadataParsed = z.infer<typeof ViewMetadataSchema>, InlineAction, AssembledViewArtifact and JoinedReportBlock as z.input of their schemas.

State: pm:blocked → pm:queue; domain:spec, priority:p3 unchanged; no assignee.

---

Generated by Claude Code

[objectstack-fleet]

Claim: PM loop round 1 · 2026-09-27T15:16Z
Session: session_01Rjy9MeetSfq34PKn81CRiN
Account: os-zhuang
Branch: claude/issue-19920-exported-types-unknown
Worktree: objectstack-issue-19920
Domain: domain:spec
Seat: domain:spec#1
File surface: the four declarations the card names and what they force, with type-level tests and .changeset/: packages/spec/src/ui/view.zod.ts (ViewMetadataParsed and diagnoseViewMetadata only), packages/spec/src/ui/action.zod.ts (InlineAction), packages/spec/src/ui/assembled-views.zod.ts (AssembledViewArtifact / …Parsed), packages/spec/src/ui/report.zod.ts (JoinedReportBlock only). ⛔ Not FormViewSchema.layout (seat 2's #20221); ⛔ not the flattened-overlay members (PR #20245). (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: default judgment tier (dispatch-gates --tier: no path-derived mandate, floor sonnet · default opus · ceiling fable). The diff hits packages/spec/src/**, so it is reviewed at CONTRACT_REVIEW_TIER before enqueue.
Clause-②: no
Thread-read: 5824953950
Serial constraints cleared: view.zod.ts / assembled-views.zod.ts: this seat's PR #20245 (#20186), which rewrote VIEW_METADATA_MEMBERS and the union these types derive from, landed as a91d12af (15:07Z); the dev works on a main that contains it. report.zod.ts: seat 4's open PR #20238 (#20161) edits the joined-report region; JoinedReportBlock is done only after it merges, else it is the remainder (Part of). view.zod.ts is also held by seat 2's #20221 at FormViewSchema.layout only, which is disjoint. action.zod.ts: no open PR or claim. Open PRs read this act: 13; pm:dispatched claims read: 27. Triage 5802280140 set the direction (derive from the member schemas, as PR #19919 did; measure the TS7056 cost first).

[objectstack-fleet]

os-dev-report

{
  "issue": 19920,
  "status": "done",
  "branch": "claude/issue-19920-exported-types-unknown",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20260",
  "session": "session_01Rjy9MeetSfq34PKn81CRiN — mode:subagent, the PM's session id; this run's identity is the branch",
  "premise_still_valid": true,
  "summary": "Three of the four sites are fixed: InlineAction, ViewMetadataParsed and AssembledViewArtifact / AssembledViewArtifactParsed now resolve to the shapes their TSDoc promises. Each is derived from the member schema the parse actually runs, as PR #19919 did for ViewMetadata: InlineAction from the z.preprocess pipe's `out` member; ViewMetadataParsed from the OUTPUT union of VIEW_METADATA_MEMBERS; the assembled pair from the members minus `container`. No cast, schema, parse or export moves. The one forced change is diagnoseViewMetadata's `data`, which now carries a type assertion (a measured single TS2322 without it; no value changes, and a new test asserts deep equality with the member's own output). JoinedReportBlock is the remainder: PR #20238 (#20161) is still an open draft (head 4aecf01d, auto-merge unarmed), so the PR body opens `Part of #19920` and names this site. The remainder's fork is measured for the next taker: removing the z.ZodTypeAny annotation emits no TS7056 but grows report.zod.d.ts 18,608 → 34,390 bytes. objectui at pin f8a9d0fb holds an inverted `true satisfies IsUnknown` pin on JoinedReportBlock; it is a designed tripwire in objectui's own type-check, and the Console Pin Gate build does not compile it. One deliberate edit to someone else's pending changeset (see open_questions) keeps the advisory Check Changeset red until confirmed in writing.",
  "tests": "Final head cf123662; code identical to 61b382d9 (cf123662 touches only .changeset/). BUILD / TYPECHECK / TEST at 61b382d9 (lock turn 1, VERDICT command-exit 0): spec build exit 0 with TS7056 x0 (base e0f17a37 also x0); `pnpm --filter @objectstack/spec typecheck` exit 0, reading 'check:test-typecheck: OK — ... 53 file(s) / 255 error(s) / 142 pinned signature(s) held'; `pnpm --filter @objectstack/spec test` 'Test Files 550 passed (550)', 'Tests 16120 passed | 2 todo (16122)'. AT cf123662 (lock turn 2, VERDICT command-exit 0): `check:generated` exit 0, 'All 15 generated artifacts are up to date'; the three pin files by name `vitest run --project local --maxWorkers=2`: 'Test Files 3 passed (3)', 'Tests 17 passed (17)'; `tsc -p tsconfig.test.json --listFilesOnly` lists all three pin files (523 test files). DECLARATION COST (spec build, both trees): total .d.ts 30,830,702 → 30,836,475 bytes (+5,773, +0.019%); chunks view.zod 499,735 → 500,541, action.zod 76,621 → 77,606, page.zod 304,750 → 305,845. Root-fix cost, in-memory declaration emit: typing the view union with the member tuple gives view.zod.d.ts 500,881 → 663,301 (ViewMetadataSchema's declaration 323 → 162,743 bytes); the assembled union gives 8,824 → 64,776; no TS7056 in either. REVERSE VERIFICATION (in-memory compiler-host ablation, anchor matched exactly once, nothing written to disk, git status clean): reverting each alias to its base spelling turns every pin red — InlineAction 4 x TS2578, ViewMetadataParsed 3 x TS2578, AssembledViewArtifact 3 x TS2578, AssembledViewArtifactParsed 1 x TS2578; with the fix, 0 diagnostics over the three pin files. Removing the diagnoseViewMetadata assertion yields exactly 1 diagnostic: view.zod.ts TS2322 'Type unknown is not assignable to type ViewMetadataParsed'. The dist-based ablation rules do not apply: the pins are judged by tsc over src (tsconfig.test.json, relative imports), not through a package's exports. GATES: see `gates`. LINT: eslint --no-inline-config --format json over the 6 changed .ts files: 6 files, 0 errors, 0 warnings. The population is eslint.config.mjs's own TS/JS globs (the .md changesets are outside it). The config never enables type-aware linting (eslint.config.mjs:327), so no untouched file's verdict can move. Repo-wide `pnpm lint` is CI's. CONSUMER SUITES: none owed. The census (git grep -w, whole tree) finds no package outside @objectstack/spec naming the four types or diagnoseViewMetadata; objectui at f8a9d0fb names only JoinedReportBlock (untouched); cloud at 48d7066 names none (control leg defineStack: 18 files).",
  "gates": "dispatch-gates.mjs --commands at cf123662: 83 commands, all run; exit codes written to disk before reading. 80 x exit 0. `node scripts/check-empty-changeset.mjs --base origin/main` exit 1: the DELIBERATE CORRECTION class, red by design, advisory `Check Changeset` job only (lint.yml runs just its self-test half). `pnpm check:dual-build-cjs-loads` exit 3 and `pnpm check:type-check-debt` exit 3: PREREQUISITE NOT MET, i.e. NOT MEASURED (both need the whole ./packages/* build closure). `dispatch-gates.mjs --ran` over the exit-coded record: '83 derived, 81 run, 2 NOT-MEASURED, 0 UNRUN'. check:lean-entry-closure and check:doc-formula-expressions were NOT MEASURED at 61b382d9 and were measured green at cf123662 after building @objectstack/formula, @objectstack/lint and @objectstack/objectql. NOT MEASURED, CI-owned: the Type Check workspace and consumer lanes, Test Core shards, Dogfood, Build Core, the 56 artifact-roster families.",
  "line_budget": "not applicable: no skills/** and no line-ratcheted ledger touched. Diff: 8 files, +261 / -9.",
  "files_changed": [
    "packages/spec/src/ui/action.zod.ts +16 -1 (InlineAction alias + TSDoc)",
    "packages/spec/src/ui/view.zod.ts +24 -3 (ViewMetadataParsed alias + TSDoc; diagnoseViewMetadata data assertion)",
    "packages/spec/src/ui/assembled-views.zod.ts +22 -4 (both aliases + TSDoc; type import of ViewMetadataBranch)",
    "packages/spec/src/ui/inline-action-type.test.ts +73 (new pin file)",
    "packages/spec/src/ui/assembled-view-artifact-type.test.ts +80 (new pin file)",
    "packages/spec/src/ui/view-metadata-type.test.ts +26 (ViewMetadataParsed pins)",
    ".changeset/19920-exported-types-not-unknown.md +19 (patch, @objectstack/spec)",
    ".changeset/view-metadata-type-not-unknown.md +1 -1 (one sentence of the pending #19919 note; see open_questions)"
  ],
  "deviations": [
    "Edited a pending changeset this PR did not add (.changeset/view-metadata-type-not-unknown.md): one sentence that this change makes false. It is within the claim's `.changeset/` surface. check-empty-changeset stays red in the advisory Check Changeset job until confirmed in writing; the PR body carries the gate's prescribed confirmation section.",
    "Reverse verification was done in memory (a compiler-host override), not by mutating the tree and rebuilding. The pins are judged by tsc over src, so no dist is on their resolution path.",
    "Zone 3's 'missing required key' pin is not expressible for InlineAction: every key of its input is optional. Its pins are unknown, the two legacy spellings, and a scalar.",
    "A detached comparison worktree, objectstack-issue-19920-base at e0f17a37, was used for the before-build. It is removed at cleanup.",
    "git push x3 via write-pace --run: the empty-branch probe, 61b382d9, cf123662."
  ],
  "mcp_calls": "0",
  "api_writes": "3 relay writes, each one repository_dispatch executed as objectstack-fleet[bot]: (1) pr_create → POST /repos/objectstack-ai/objectstack/pulls (draft #20260), relay run 36333543354; (2) label-write --assign os-zhuang → POST /repos/objectstack-ai/objectstack/issues/20260/assignees, relay run 36333595870, read-back matches; (3) this os-dev-report comment → POST /repos/objectstack-ai/objectstack/issues/19920/comments. Plus git push x3, which is not REST. Reads (GET) were made with the ambient token and are not counted.",
  "open_questions": [
    {
      "question": "Clause-② and changeset level for a published TYPE narrowing. The claim says `Clause-②: no` and the dispatch says patch; both are copied verbatim into the PR and changeset. The #19919 precedent (same defect class, same file, same pending release) declared `Clause-②: no (narrowing)`, `minor`, a BREAKING banner and an adr-0087 not-required marker. Should this PR match it?",
      "options": [
        "A: keep `Clause-②: no` + patch, as claimed. Zero further writes; the changeset already carries the 'if your code stops compiling' instruction.",
        "B: align with #19919: the seat edits the PR body's Clause-② line to `no (narrowing)`, and the changeset becomes minor with a BREAKING banner and the not-required (no-migration-prescription) marker. A changeset-only commit plus one body edit by the seat."
      ],
      "recommendation": "B, on axes 2 and 3. Measured need (axis 1): 0 consumers of the three types in objectstack packages, objectui at f8a9d0fb or cloud at 48d7066, so no measured consumer breaks. Long-term (axis 2): one defect class shipping as BREAKING-minor for ViewMetadata and as patch for ViewMetadataParsed, in the same release, is an inconsistent changelog; the arm exists because a narrowing once shipped with every gate green. AI authoring (axis 3): the arm makes check-adr-0087-registration read the change as breaking and demand its disposition, which is the channel an upgrading agent greps. Startup (axis 4): neutral, since lockstep ships breaking as minor in the launch window and B costs one changeset edit. The arm is the claim owner's to write, so this run did not self-create it."
    },
    {
      "question": "Confirm the deliberate correction of the pending #19919 changeset sentence (check-empty-changeset red in the advisory Check Changeset job).",
      "options": [
        "A: confirm in writing on PR #20260. The correction stands; the approver merges over the advisory red.",
        "B: restore the base text. If both entries ship in one release, the CHANGELOG then says ViewMetadataParsed 'is still unknown', which is false; AGENTS.md says a correction elsewhere never reaches the reader who greps the entry."
      ],
      "recommendation": "A. The gate itself prescribes it for this class. The note's owner is this same seat's #19871 round, and the new wording holds whichever release carries either entry."
    }
  ],
  "out_of_scope_findings": [
    "class: b · reach: named producer — TypeScript consumers annotating `ViewItem` / `ViewItemWire` (the producer #19920 names), and through them the viewItem member of ViewMetadata, ViewMetadataParsed and AssembledViewArtifact · evidence: compiler-API read at cf123662, `config: unknown` on both members of ViewItem and ViewItemWire; cause `viewItemArmShape(viewKind, config: z.ZodTypeAny)` at packages/spec/src/ui/view.zod.ts:4887; the declaration-size cost of typing it is NOT MEASURED · SAME FAMILY as #19920: fold into this family's closing card, not a single-point card · dedupe words: `ViewItem config unknown`, `viewItemArmShape ZodTypeAny`, `ViewItemWire config unknown`",
    "carrier: 承接者:无 · noted, not filed — check:spec-parsed-alias recognises a bare alias only in the spelling z.input of typeof the schema, so member-derived aliases leave its population: 1443 → 1441 bare, 657 → 655 paired at cf123662 (ViewMetadata left it with #19919). The …Parsed siblings exist; the gate just no longer sees the pairs. Recorded in the PR's Acceptance notes."
  ]
}

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report

{
  "issue": 19920,
  "status": "done",
  "branch": "claude/issue-19920-exported-types-unknown",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20260",
  "session": "session_01Rjy9MeetSfq34PKn81CRiN — mode:subagent, the PM's session id; this run's identity is the branch",
  "premise_still_valid": true,
  "round": "2 — the changeset-only round the seat's answers owe (Q1 = B, Q2 = A)",
  "summary": "Q1 = B is applied to .changeset/19920-exported-types-not-unknown.md, all four parts: (i) '@objectstack/spec': minor; (ii) `Clause-②: no (narrowing)`; (iii) a BREAKING banner mirroring #19919's: a narrowing of published TYPES landing as minor in the launch window, the runtime accept set not moving; (iv) the ADR-0087 marker `not-required (no-migration-prescription)` with a one-sentence reason. The 'if your code stops compiling' instruction is kept. The banner names all four aliases (InlineAction, ViewMetadataParsed, AssembledViewArtifact, AssembledViewArtifactParsed): the three sites carry four annotatable names. Q2 = A: the correction to .changeset/view-metadata-type-not-unknown.md stands and was not restored. The PR body's `Clause-②: no` line is left verbatim. No merge: `git merge-tree --write-tree --name-only` against origin/main a9fb83ef exits 0 with no conflicted path; main moved 7 commits since e0f17a37, none touching this diff's files, and no path of this diff is routed to the os-regen driver. Pushed f8792c93 on the paced route (remote tip = local). PR body edits owed by the seat (the dev writes the body once): line 21, 'Changeset …: `patch` on `@objectstack/spec`' → 'minor, `Clause-②: no (narrowing)` with the BREAKING banner and the ADR-0087 not-required (no-migration-prescription) disposition'; lines 69 / 75 / 78 name head cf123662 → f8792c93, whose code is still identical to 61b382d9 (cf123662..f8792c93 touches one changeset only). JoinedReportBlock remains the named remainder: PR #20238 is still an open draft at this reading. Round-1 findings and measurements stand unchanged.",
  "tests": "Head f8792c93. The diff cf123662..f8792c93 is exactly one file, .changeset/19920-exported-types-not-unknown.md. dispatch-gates --commands at f8792c93 re-derives the same 83 commands (sorted diff against cf123662's list: identical). DECLARED NARROWING: every one of the 83 resolves to an existing entry script; 17 of those scripts mention `.changeset`, and those 17 plus `check:spec-changes` were re-run at f8792c93 with exit codes written to disk first. The other 66 stand at their cf123662 reading, since no source moved. Blind spot: a changeset read inside an imported module rather than the entry script. RESULTS (18 run, 17 x exit 0, 1 x exit 1): `node scripts/check-adr-0087-registration.mjs --base origin/main` exit 0: '✓ check-adr-0087-registration: 1 declared-breaking changeset(s), each carrying an ADR-0087 disposition.' reading `.changeset/19920-exported-types-not-unknown.md [BREAKING+clause-②-narrowing] not-required (no-migration-prescription)`; its --self-test exit 0. `node scripts/check-changeset-no-major.mjs --base origin/main` exit 0: '✓ This diff introduces no `major` bump.' Its level axis reads NOT APPLICABLE locally (no pull_request), so it was ALSO driven with `--event` wrapping PR #20260's live GET payload: exit 0, '✓ LEVEL AXIS: this PR declares clause-② `no`, so no package here is declared to have grown a published surface.' `pnpm --filter @objectstack/spec run check:spec-changes` exit 0, 'spec-changes.json is up to date.' It reads the migration registry and package versions, not changesets (its `.changeset` mentions are comment-only), and no spec generator reads `.changeset`, so `check:generated` is not triggered. `node scripts/check-empty-changeset.mjs --base origin/main` exit 1: still ONLY the deliberate correction, naming `.changeset/view-metadata-type-not-unknown.md` and not this PR's own changeset (0 mentions). Also exit 0: check-empty-changeset --self-test, check-keyed-text-bounds (+ --self-test), release-rehearsal-clone --self-test, check:objectui-pin-citations, check:changeset-gate-self-tests, check:driver-memory-census, check:issue-citations, check:nul-bytes, check:objectui-changeset, check:pm-changeset-deadline-census, check:published-files. Code-level readings (build, typecheck, 550-file spec suite, pins, ablations) are round 1's at 61b382d9, whose code f8792c93 carries byte-identically.",
  "gates": "18 of the 83 derived families re-run at f8792c93 (the declared changeset-reading narrowing above): 17 x exit 0; check-empty-changeset exit 1, the DELIBERATE CORRECTION class, red by design in the pull_request-only, non-required Check Changeset job. The 66 others stand at their cf123662 reading (round 1: 81 run, 2 NOT-MEASURED for a missing whole-packages build closure).",
  "line_budget": "not applicable: no skills/** and no line-ratcheted ledger touched. Round-2 diff: 1 file.",
  "files_changed": [
    ".changeset/19920-exported-types-not-unknown.md (patch → minor; Clause-② arm `(narrowing)`; BREAKING banner; ADR-0087 not-required marker) — commit f8792c93"
  ],
  "deviations": [
    "The worktree /home/user/objectstack-issue-19920 was removed at round-1 cleanup. It was re-created on the existing branch (local = remote = cf123662 before the edit, with only this run's two commits on it). No new branch was made.",
    "The banner names four aliases where the instruction says 'the three names': three sites, and the assembled site has two annotatable names, so all four are named to leave no annotated name out of the BREAKING notice."
  ],
  "mcp_calls": 0,
  "api_writes": 1,
  "api_writes_endpoints": "(1) this os-dev-report comment → POST /repos/objectstack-ai/objectstack/issues/19920/comments via post-stamped over the fleet-write relay (repository_dispatch, as objectstack-fleet[bot]). Plus 1 git push (not REST). Round-1 total was 3 relay writes; the cumulative total is 4. GETs used the ambient token and are not counted.",
  "open_questions": [],
  "out_of_scope_findings": []
}

---

Generated by Claude Code