feat(spec): Operation Message Catalog gains refusal-situation keys for approvals recall and sharing write-denial

[os-zhuang]

Fixes #12493

Clause-②: yes; dispatched at review tier.

The spec-side half of the shape-A ruling on the approvals-recall card (triage 2026-08-26). Two measured hardcoded-English FORBIDDEN refusals gain user-situation keys in the Operation Message Catalog, in all four platform locales, so their emitters can later call renderOperationMessage instead of throwing raw English. Scope held: keys, catalog text, catalog-convention tests, changeset — zero edits outside packages/spec (plus the changeset file). Emitter call-site conversion stays with the consumer cards: #11993 remains open (its services half converts the approvals emitter after this lands) and #12260 remains open (probable second consumer, repair-pending; its consumer side is not implemented here).

New keys

• record_write_denied — the user can see this record, but changing or deleting it is beyond their access. Measured emitter (recorded, not edited): packages/plugins/plugin-sharing/src/sharing-plugin.ts:1120, the by-id write gate throwing the FORBIDDEN insufficient-privileges string with the object API name and row id interpolated in; op there is update or delete, so the key covers both write verbs — the user's situation and remedy are the same, and the refused verb stays a developer fact on developerMessage/details. Deliberately NOT record_access_denied restated: that sentence ("You do not have access to this record") would be false on a row the read path already admitted — the user is typically looking at the record the write gate refuses.
• approval_recall_not_submitter — the user asked to recall an approval request someone else submitted. Measured emitter (recorded, not edited): packages/plugins/plugin-approvals/src/approval-service.ts:2706. The copy names who CAN act — the submitter, or an administrator (the code path carries a privileged-admin override, so sending the user to an administrator is accurate).

Both sentences take no placeholders, re-derived per site per the family rule (the only nameable facts at both throw sites are object API names and opaque ids, which must not reach a toast). Both render through renderOperationMessage with the identical override/locale/fallback ladder as the record_access_denied family and are addressable for deployment overrides as errors.record_write_denied / errors.approval_recall_not_submitter. Module header updated: members list, situation table (two rows), and per-site placeholder derivations.

Tests

New describe block in packages/spec/src/system/operation-message.test.ts mirroring the existing family battery: caller-locale rendering (zh-CN + en pinned by value), base-language regional match (ja), en fallback for an uncarried locale, developer-vocabulary absence in every locale (family list plus FORBIDDEN and the insufficient-privileges phrase), distinctness from every sibling key per locale, placeholder-free catalog entries asserted on the entry (not the rendering), deployment-override wins, throwing-i18n-service containment. The pre-existing "every locale defines every key en defines" completeness test now also covers the new keys.

Verification (all through scripts/pm/os-verify-lock.sh; exit codes captured before any pipe; verdict lines quoted from the gates' own output)

Union re-run at the final head 8ee1941 (after merging origin/main forward — see landing-watch below):

• pnpm --filter @objectstack/spec build — "check-dts-emitted: @objectstack/spec - 34/34 declared declaration file(s) present."
• pnpm --filter @objectstack/spec test — "Test Files 435 passed (435) / Tests 11533 passed (11533)" at head 8ee1941.
• pnpm --filter @objectstack/spec typecheck — clean, including "check:test-typecheck: OK — @objectstack/spec's test layer compiles under packages/spec/tsconfig.test.json" (the new test file is inside the checked test program).
• pnpm --filter @objectstack/spec check:generated — "All 14 generated artifacts are up to date." Zero baseline row deltas: the catalog ships as constants, not schemas, so no generated artifact moves.
• pnpm --filter @objectstack/spec check:authorable-surface — green; "authorable-defaults/ verified against upstream e2d1f80 — 1255 default(s) unchanged". The baseRev-trails note is the documented informational line, not an error.
• Gate derivation: node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack at head (answer taken from this checkout of objectstack at 999bf8b, re-confirmed after merge). All derived path-matched and convention-triggered families ran locally and are green, including check:doc-formula-expressions (after building its prerequisite packages), check:dev-prereqs (after the full package-workspace build), check:engine-double-contract ("OK — 689 pinned, 134 in the DEBT ledger, 3 exempt"), check:where-matcher ("303 matcher(s) discovered, 303 answer... correctly or refuse it loudly"), check:query-options-erasure ("ratchet holds... none new"), check:cross-package-test-inputs, and check:type-check-debt ("--re-measure: OK — 31 ledger entr(ies) re-measured, 1687 raw tsc error(s) total, none above its recorded number"). Ratchet families re-run at the merged head 8ee1941. Left to CI: the whole-repo eslint run (pnpm lint) and the workflow families whose population is unreachable from these paths — CI runs the full farm on every PR regardless.
• pnpm check:nul-bytes — green.

Reverse verification

Direction predicted, then observed: turn red. With the catalog file reverted to the base revision (mutation proved on disk: 0 grep hits for the new key in the catalog, 6 in the kept test file) and the new tests kept, the targeted suite went RED — 7 failed / 27 passed, all seven failures in the new block (the override-wins case stays green by design: its translate stub answers regardless of the catalog). Restore proved by hash: git checkout HEAD on the path, git diff HEAD empty, disk blob hash equal to the HEAD blob hash; re-run 34/34 green. No build step is part of this loop — the test imports the module by relative source path, so no dist is on the resolution path.

Changeset

.changeset/operation-message-refusal-keys.md, minor on @objectstack/spec, following the landed published-surface precedent (the FLOW_INPUT_SCHEMA_INVALID ledger-row changeset): states plainly that this release ships catalog keys only, the emitters still throw raw English until the consumer halves land, and what operators see change when they do.

Landing-watch

origin/main moved mid-flight by two CLI-test commits (e2d1f80, 1ca763b — disjoint from packages/spec); merged forward (merge, not rebase) and the union plus ratchet families re-ran at the merged head 8ee1941 as reported above. Also disjoint from the in-flight #12497 surface (security/permission.zod.ts + registries) — no overlap with this diff.

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

3 anchor(s) derived from 1 changed package(s); no hand-written page names any of them, so this run has nothing to list — not a clean bill of health. This check sees only pages that NAME a derived anchor: one that documents this change in prose, or enumerates it in an authoring dialect, names none and stays invisible to it on every run.

What this run could not see

• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 126 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 0043c92240ffe5b53df43753f9ab1edd2446b474 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 1d6e10f5190d6a507fa12c6da332fab22cac1736 — the merge of head 8ee19415799865a2d0acc259818196057ba7cc21 into base 0043c92240ffe5b53df43753f9ab1edd2446b474, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 1d6e10f5190d6a507fa12c6da332fab22cac1736 && git checkout 1d6e10f5190d6a507fa12c6da332fab22cac1736
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 0043c92240ffe5b53df43753f9ab1edd2446b474 8ee19415799865a2d0acc259818196057ba7cc21 && git checkout -B drift-repro 0043c92240ffe5b53df43753f9ab1edd2446b474 && git merge --no-ff 8ee19415799865a2d0acc259818196057ba7cc21

node scripts/docs-audit/affected-docs.mjs --json 0043c92240ffe5b53df43753f9ab1edd2446b474

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.