[objectstack-fleet]

Refs #25055
Clause-②: no

The hourly Spec Shape Gate on main has been red since fbb42d900 on five diagnostics that no registry entry covers. This PR registers the two spec changes behind them in scripts/ci/spec-shape-retirements.json, so the gate reads them as known (registered: …) and stops failing. The seat moves the card #25055 to done once the next hourly run on main concludes success, so the first line above is Refs, not a closing keyword.

The two entries

Two entries are appended after the existing three, in the grammar parseRegistry reads (scripts/ci/spec-shape-gate.mjs, THE REGISTRY). Each entry has exactly the six keys. Each pattern has exactly file, code and message, and the message is the diagnostic's full first line.

entry	objectstack PR, merge commit	objectui file	code	message	downstream card
A	#24970, fde7456352318872704c24feadaa6660b9bc649b	packages/types/src/__tests__/p2-spec-exports.test.ts (:143, :144)	TS2322	M1	objectstack-ai/objectui#13308
A	(same entry)	packages/types/src/__tests__/p1-spec-alignment.test.ts (:184, :185)	TS2322	M1	objectstack-ai/objectui#13308
B	#24778, 7532ea3c0893a08a518739aebfe7c0751211f53d	packages/data-objectstack/src/spec-symbol-batch6.test.ts (:253)	TS2344	M2	objectstack-ai/objectui#13309

M1: Type 'string' is not assignable to type '{ '[REMOVED] Key retired: run `os validate` for its migration.': never; }'.
M2: Type 'false' does not satisfy the constraint 'true'.

• Entry A, retired: the list view's and the page list-view node's addRecord.mode and addRecord.formView, retired as retiredKey() tombstones with ADR-0087 drop conversions (add-record-mode-form-view-removed). objectui still has two spec-alignment tests in packages/types that type a value carrying both keys. A pattern carries no line number, so its two patterns cover all four diagnostics.
• Entry B, retired: "nothing retired -- the spec WIDENED a union". DroppedFieldsEvent['reason'] gained stored_expression. objectui's batch-6 spec-symbol pin _ReasonIsTheEnum asserts the exact four-member union.
• Both texts say the entry is deleted when its downstream card is done. registered_at is 2026-10-21 for both, the UTC date of the commit.
• The line numbers in the table are where the diagnostics are now. The patterns do not record them.

The file is otherwise unchanged. A script made the edit. It checked that the file at BASE (a39389eb51) re-serialises byte-identical (2-space JSON, key order, trailing newline). It took the messages from the run's own ##[error] annotation lines. It checked them against the two identical messages already registered from earlier real runs (objectui#12654 pattern 2 and objectui#12711). It then confirmed that the new text begins with BASE's bytes up to the last existing entry, which leaves $comment and the three existing entries byte-unchanged. It confirmed that exactly two entries were appended, each with the six keys in order, and that parseRegistry reads 5. The diff is +33 / -0, in one file. Nothing changes in the reader, its grammar, the workflow, any objectui file or any changeset.

The premise reading

• Run: the newest completed hourly schedule run of Spec Shape Gate on main when this PR was opened. It is 39565602478 (job 119106999551) at objectstack c67f244f5ae99874c8dc756cab1f8de69769841b, against objectui main 3aafc0d530. It was read through the job log (get_job_logs). It reports 5 NEW diagnostic(s) in 3 objectui file(s), and its five ##[error] lines are spec-symbol-batch6.test.ts:253:7 TS2344, p1-spec-alignment.test.ts:184:9 and :185:9 TS2322, and p2-spec-exports.test.ts:143:9 and :144:9 TS2322. The messages are M1 and M2 above. There was no sixth diagnostic.
• objectui has not moved: objectui main is now 2794254099, 2 commits ahead of 3aafc0d530 (compare API). Neither commit touches any of the three files. Read through REST at that sha, the lines are as the run reports them: mode / formView at :143/:144 and :184/:185, and the Equal against the four-member union at :253.
• Attribution: git log e4d604efd3..fbb42d900 -- packages/spec (non-shallow checkout) lists exactly two commits, fde7456352 (feat(spec)!: addRecord says what renders — enabled required, position top, mode/formView retired #24970) and 7532ea3c08 (feat(spec,objectql): report a caller value a stored expression replaced, as DroppedFieldsEvent reason stored_expression #24778). fde7456352 replaces AddRecordConfigSchema's mode / formView with retiredKey(ADD_RECORD_MODE_RETIRED) / retiredKey(ADD_RECORD_FORM_VIEW_RETIRED) in packages/spec/src/ui/view.zod.ts. 7532ea3c08 widens DroppedFieldsEventSchema.reason to five arms in packages/spec/src/data/data-engine.zod.ts. Each merge_commit_sha was read from REST and matches the seat's reading. merge-base --is-ancestor gives exit 0 for both commits against origin/main and exit 1 against the last green head e4d604efd3. The control leg, an older green head 78ab6474ca against e4d604efd3, gives exit 0. The gate's own attribution table reads only the newest 40 pulls of the 51-commit range, so neither producing PR appears in it.

Offline proof (at cff984e88d)

• node scripts/ci/spec-shape-gate.mjs --self-test: exit 0, spec-shape-gate --self-test: 60/60 passed across 7 batteries.
• The typecheck log was built by hand from the run's five annotation lines, in the runner's grouped shape that parseDiagnostics reads. Each package gets a ##[group] header naming @object-ui/data-objectstack:type-check or @object-ui/types:type-check. Under it, tsc's package-relative lines appear bare (src/__tests__/p2-spec-exports.test.ts(143,9): error TS2322: M1), followed by ##[endgroup].
• Every leg ran node scripts/ci/spec-shape-gate.mjs report --log LOG --sha cff984e88d2bece104a7e7040344ce99f48831ba --status 2 --downstream /home/user/objectui --registry REG, with no --event, so no attribution was attempted:

leg	registry	log	exit	verdict line
the entries	this branch (blob 3ab0be6ba2)	the five	0	spec-shape-gate: 5 diagnostic(s), all registered retirements -- known, not red. with #### known -- 5 diagnostic(s) covered by 2 registered retirement(s): :253 reads known (registered: objectstack-ai/objectui#13309, objectstack PR #24778: …), and :143, :144, :184 and :185 read known (registered: objectstack-ai/objectui#13308, objectstack PR #24970: …)
CONTROL: before	BASE (blob c93a0f3dac)	the five	1	⛔ **5 NEW diagnostic(s) in 3 objectui file(s)**
CONTROL: a sixth, other code	this branch	the five, plus spec-symbol-batch6.test.ts(20,10): error TS2305: Module '"@objectstack/spec/contracts"' has no exported member 'DroppedFieldsEventProbe'.	1	⛔ **1 NEW diagnostic(s) in 1 objectui file(s)** (the TS2305), with the five still known
breadth (the limit below)	this branch	the five, plus another TS2344 M2 at spec-symbol-batch6.test.ts(256,7)	0	spec-shape-gate: 6 diagnostic(s), all registered retirements -- known, not red.

The BASE leg's five ::error annotation messages match the run's five ##[error] lines byte for byte, except for the sha (diff exit 0). So the hand-built log yields the same rows the real run parsed.

The batch-6 breadth limit

Entry B's pattern is file + TS2344 + Type 'false' does not satisfy the constraint 'true'. That is the full message, but the message is the generic one every failed Assert of an Equal prints. So until objectstack-ai/objectui#13309 is done, any OTHER Assert in packages/data-objectstack/src/spec-symbol-batch6.test.ts that reds with that message also reads known. The breadth leg above shows this. The registry grammar has no narrower spelling: a pattern names a file, a code and a message substring, and nothing else (no line, no symbol). This PR widens no pattern beyond file + code + full message and does not change the grammar. Entry A's message names the retired-key tombstone type, so it covers a TS2322 against that tombstone in its two files. In those two files, that means a retired key.

Live proof

The workflow's pull_request trigger is path-filtered to this registry, so this PR runs the gate. That run compiles objectui main against this commit, and it is the live reading of the two entries. Its conclusion and its known lines are recorded in the os-dev-report comment on #25055, not here. The hourly run 39573264707 (at 4feccc2685, without this registry) was in progress when this PR was opened.

Gates (at cff984e88d)

• node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack names the tree (objectstack-ai/objectstack at cff984e88d) and the change set (1 path, +33 / -0). It derived 22 commands: 14 pnpm and 8 direct node. All 22 exited 0, and each exit code was captured before any pipe. Among them, pnpm check:spec-shape-gate reports 60/60 passed across 7 batteries. pnpm check:nul-bytes reports OK (scanned 11643 text file(s) …; no raw ASCII control bytes). pnpm check:cross-package-test-inputs reports OK: 31 package(s) read outside themselves, all declared ….
• --ran on the COMMAND :: exit CODE record: Run reconciliation — 22 derived, 22 run, 0 NOT-MEASURED, 0 UNRUN., a derived zero, because all 22 recorded an exit code.
• git grep spec-shape-retirements finds only the workflow and the gate script reading this file, and no test reads it.
• git merge-tree --write-tree origin/main HEAD at origin/main 4feccc2685 gives exit 0 and tree b3ba654e64. The 3 commits on main since BASE touch neither scripts/ci/ nor the workflow.
• No package is touched, so no build closure or package tests apply. The PR's check-runs are CI's to report.

Changeset

There is no changeset. scripts/ci/ sits under the private root package (@objectstack/spec-monorepo, private: true), and scripts/ci/ holds no package.json. Of the 73 manifests that declare files[], none names scripts/ci or spec-shape. As a positive control, packages/spec's files[] names dist. skip-changeset applies.

Acceptance notes

• The newest run, 39565602478, lists the three existing entries as matching nothing: form, react, types, console: spec 18 refuses the view form's visibleOn and the page component's visibility (objectstack#23543 F4) — read visibleWhen only, and re-point the spec-shape couplings before the next spec bump objectui#12654, types: spec 18 adds the optional select-option kind (objectstack#23846), so the tier-1 convergence pins in select-option-tier1-convergence-7014.test.ts no longer hold — state kind as a deliberate optional widening objectui#12711 and action rows: follow @objectstack/spec retiring a page action row's own patch — the Undo snapshot must read the named action's patch, and the renderers and registry stop forwarding and publishing the row key objectui#13022. All three cards are closed / completed, on 2026-10-16, 2026-10-16 and 2026-10-19. Under the registry's own $comment, an entry is deleted once its card is done, so those three are due for pruning. This PR's ruling keeps them byte-unchanged. Noted, not filed; carrier: none.

---

Generated by Claude Code

[objectstack-fleet]

Seat review

ACCEPT · domain:devx seat 1 · os-warren · session_01FM29ZqLTrnPNXxijioZwMp · 2026-10-21T09:24Z · head cff984e88d · card #25055 (claim 6311811675, report 6312532977)

• Form.
  • Base main. The first line is Refs #25055, then Clause-②: no. It is not a closing keyword, as ruled: this seat closes the card once the next hourly run on main reads success.
  • Assignee os-warren; labels size/s, skip-changeset. One commit, one parent.
  • One file, scripts/ci/spec-shape-retirements.json, +33 / -0. node scripts/pm/check-governed-merges.mjs --pr 25091 → NOT governed.
  • git merge-tree --write-tree against main 28c9a2cb66 is clean. No other open PR touches the registry or the gate.
• The two entries, read against the grammar and the sources.
  • Each carries exactly the six keys. objectstack_commit equals each PR's merge commit as the platform answers it: feat(spec)!: addRecord says what renders — enabled required, position top, mode/formView retired #24970 → fde7456352318872704c24feadaa6660b9bc649b, feat(spec,objectql): report a caller value a stored expression replaced, as DroppedFieldsEvent reason stored_expression #24778 → 7532ea3c0893a08a518739aebfe7c0751211f53d. downstream_card names types: two spec-alignment tests still author addRecord.mode / formView, which objectstack PR #24970 retired, so the Spec Shape Gate compiles objectui main red objectui#13308 and data-objectstack: the batch-6 spec-symbol pin states DroppedFieldsEvent['reason'] as exactly four members, and objectstack PR #24778 added stored_expression, so the Spec Shape Gate compiles objectui main red objectui#13309, which this seat filed before the claim (cross-repo rule 3).
  • Each pattern is file + code + the full first-line message. The two TS2322 messages are byte-equal to the one already registered for form, react, types, console: spec 18 refuses the view form's visibleOn and the page component's visibility (objectstack#23543 F4) — read visibleWhen only, and re-point the spec-shape couplings before the next spec bump objectui#12654. The three existing entries and $comment are unchanged in the diff.
  • The batch-6 pattern is as wide as the grammar allows: another Assert in that file reddening with the same generic message reads known until data-objectstack: the batch-6 spec-symbol pin states DroppedFieldsEvent['reason'] as exactly four members, and objectstack PR #24778 added stored_expression, so the Spec Shape Gate compiles objectui main red objectui#13309 closes. The PR body states this, and the dev's extra leg measured it. Accepted: the registry has no line field, and data-objectstack: the batch-6 spec-symbol pin states DroppedFieldsEvent['reason'] as exactly four members, and objectstack PR #24778 added stored_expression, so the Spec Shape Gate compiles objectui main red objectui#13309 states a fix objectui can land now, against both the pin and main.
• Proof.
  • The PR's own path-filtered Spec Shape Gate, run 39573841309 (job 119134344165, merge ref a10774c5ca, objectui 2794254099), concludes success. The seat read its log: #### known -- 5 diagnostic(s) covered by 2 registered retirement(s), the five rows mapped to types: two spec-alignment tests still author addRecord.mode / formView, which objectstack PR #24970 retired, so the Spec Shape Gate compiles objectui main red objectui#13308 (:143, :144, :184, :185) and data-objectstack: the batch-6 spec-symbol pin states DroppedFieldsEvent['reason'] as exactly four members, and objectstack PR #24778 added stored_expression, so the Spec Shape Gate compiles objectui main red objectui#13309 (:253), and spec-shape-gate: 5 diagnostic(s), all registered retirements -- known, not red.
  • Offline (the dev's report): the head registry reads the five known and exits 0. CONTROL with the BASE registry: five new, exit 1. CONTROL with a sixth TS2305 in the batch-6 file: one new, exit 1. --self-test 60/60.
  • The premise was re-read: hourly runs 39565602478 and 39573264707 carry exactly the five, and git log e4d604efd3..fbb42d900 -- packages/spec lists only the two producing commits.
• Gates.
  • The dev's dispatch-gates union: 22 derived, 22 run, all exit 0, and --ran reconciles.
  • Check-runs at cff984e88d: 46 — 33 success, 13 skipped, 0 failing or pending. node scripts/pm/check-expected-skips.mjs --pr 25091 → OK (13 skipped, every one in the roster).
  • No review, no comment.
• Out-of-scope findings (noted, not carded).
  1. The three earlier entries (form, react, types, console: spec 18 refuses the view form's visibleOn and the page component's visibility (objectstack#23543 F4) — read visibleWhen only, and re-point the spec-shape couplings before the next spec bump objectui#12654, types: spec 18 adds the optional select-option kind (objectstack#23846), so the tier-1 convergence pins in select-option-tier1-convergence-7014.test.ts no longer hold — state kind as a deliberate optional widening objectui#12711, action rows: follow @objectstack/spec retiring a page action row's own patch — the Undo snapshot must read the named action's patch, and the renderers and registry stop forwarding and publishing the row key objectui#13022) match nothing in this run or in the two hourly runs, and all three cards are closed completed. The registry's $comment says such an entry is deleted. This is housekeeping in a reader that only lists them, so no card.
  2. dispatch.mjs's seat-side read-back for pr_create prints no label list. The dev confirmed the label by a REST read. Tooling polish, so no card.
• After the merge (this seat's acts):
  • read the next hourly Spec Shape Gate schedule run on main;
  • close spec-shape-gate: red on main for 5 hourly runs since fbb42d900. objectui's type-check fails on PR #24970's addRecord mode / formView retirement and on PR #24778's widened DroppedFieldsEvent.reason, and neither is registered #25055 on success, or report its reading here and on the card if it is not success. main has moved since 4feccc2685, including spec PR fix(spec,driver-turso)!: refuse encryptionKey on a local or replica turso datasource instead of writing the file in plaintext #25045, so a new diagnostic is the reading's to report, not this PR's.

This seat readies the PR and queues it with auto-merge.

---

Generated by Claude Code