fix(platform-objects): re-translate the object leaves that contradict their current en source

[objectstack-fleet]

Fixes #20539

Clause-②: no

What changed

A translated leaf in packages/platform-objects/src/apps/translations/{ja-JP,es-ES,zh-CN}.objects.generated.ts that a translator wrote by hand keeps its value when its en source changes later. Some of those leaves still said what the old source said. This PR measures that set first, then re-translates the leaves whose meaning now contradicts the current en: 19 leaves (ja-JP 4, es-ES 11, zh-CN 4). The leaves whose en only gained detail or was reworded are listed below and left as written.

• Values only. No key is added or dropped, no en file is edited, and no provenance companion or echo-decision ledger changes (measured below; the tooling requires none).
• .changeset/20539-stale-authored-object-leaves.md: @objectstack/platform-objects patch.
• No new gate, per triage. No test pins a translation string.

The measurement (taken on this branch at base cd901d7a5, before any edit)

Instrument

A diff of en since each authored leaf's last edit, as the card names it, run over git history:

1. Population. Every string leaf of each locale's objects bundle at the ref, minus two classes that are not authored: a leaf byte-equal to the current en leaf (an echo, owned by the echo ledgers), and a leaf with a row in the locale's *.source-hashes.generated.ts (a recorded byte copy, owned by check:i18n-stale-fill). What is left is the authored set.
2. Last edit. Walk every first-parent main commit that touched the locale bundle, oldest first, parse the object literal at each, and record for each path the commit where its value last changed. Values are compared parsed, so a reformat is not an edit. For a leaf that landed through a merge commit, en is also read at the branch side.
3. Verdict. Compare en's value for the same path at that commit with en today. If they differ, en moved after the leaf was last edited, and the leaf is a candidate. Meaning is then judged by hand for each candidate.

It reads the en bundle history rather than the object source files. Today en is the source, byte for byte (pnpm check:i18n green). Before 8af76aebf the en bundle could lag its source, and that lag only ever adds candidates; it never hides one. The bundle diff is therefore at least as sensitive as a source diff, and it survives object files moving between packages. Full history is required: the checkout was shallow (499 commits) and was unshallowed to 15092 before the run.

The same logic, condensed, reproduces the full instrument's candidate sets exactly at cd901d7a5 and 3fa342501:

// node instrument.mjs REF  (run from the repo root; needs full, unshallowed history)
import { execFileSync as x } from 'node:child_process';
const REF = process.argv[2] || 'HEAD', D = 'packages/platform-objects/src/apps/translations/';
const git = (...a) => x('git', a, { encoding: 'utf8', maxBuffer: 268435456 });
const leaves = (o, pre = '', m = new Map()) => { for (const [k, v] of Object.entries(o)) { const p = pre ? pre + '.' + k : k; if (typeof v === 'string') m.set(p, v); else if (v && typeof v === 'object') leaves(v, p, m); } return m; };
const at = (rev, f) => { const s = git('show', rev + ':' + D + f); return leaves(new Function('return (' + s.slice(s.indexOf('= {', s.indexOf('export const')) + 2, s.lastIndexOf('};') + 1) + ');')()); };
const enNow = at(REF, 'en.objects.generated.ts');
for (const L of ['ja-JP', 'es-ES', 'zh-CN']) {
  const f = L + '.objects.generated.ts', last = new Map();
  let prev = new Map();
  for (const c of git('log', '--first-parent', '--format=%H', REF, '--', D + f).trim().split('\n').reverse()) {
    const cur = at(c, f);
    for (const [p, v] of cur) if (prev.get(p) !== v) last.set(p, c);
    prev = cur;
  }
  const recorded = git('show', REF + ':' + D + L + '.source-hashes.generated.ts');
  let authored = 0; const cands = [];
  for (const [p, v] of at(REF, f)) {
    if (v === enNow.get(p) || recorded.includes('"objects.' + p + '"')) continue; // echo / recorded byte copy
    authored++;
    const c = last.get(p), enThen = at(c, 'en.objects.generated.ts').get(p);
    if (enThen !== enNow.get(p)) cands.push(p + ' @' + c.slice(0, 9));
  }
  console.log(L, 'authored=' + authored, 'candidates=' + cands.length);
  for (const k of cands) console.log('  ' + k);
}

Two widening checks ran beside it:

• Penultimate edit. For each non-candidate, the full instrument also compared en at the leaf's previous edit. This catches a sweep that touched a leaf mechanically after en had moved. It flagged 12 / 8 / 12 leaves (ja-JP / es-ES / zh-CN). All of them were re-translated in the same commit that moved en, so none is stale.
• Retired-term scan. The since-last-edit instrument cannot see a leaf that was already wrong at its last edit. For the two renames the candidates surfaced (ADR-0057 department to business unit, and v5.0 project to environment), every leaf that carries the locale's word for the retired term (部門 / 部门 / departamento case-insensitive; プロジェクト / 项目 / proyecto) was compared with the en leaf at the same path. A leaf is flagged when en does not carry the English term there.

Known-positive control

sys_two_factor.fields.backup_codes.help, line 1095 of all four bundles. en has read "Backup recovery codes, encrypted at rest (a single opaque ciphertext, not readable JSON)" since 266654d82. The three translations were last edited at 7bb92056e, when en read "JSON-serialized backup recovery codes". The instrument flags it in all three locales, zh-CN included. The retired-term scan's positive is sys_metadata.fields.environment_id.label, which both instruments flag.

Population and radius

• Covered: the objects bundles of @objectstack/platform-objects for ja-JP, es-ES and zh-CN. There are 1529 leaves per locale. Echoes of the current en number 43 / 54 / 42, recorded byte copies 0 beyond those, and authored leaves 1486 / 1475 / 1487. History covered: every first-parent commit since the bundles were created at 6bacbced2 (62 / 62 / 63 commits touching each locale file).
• Not covered: a leaf that was wrong at its last edit for any reason other than the two renames above; the platform-objects metadata-forms bundles; the other eight packages' bundle sets. Raw candidate counts for those are in the acceptance notes.

Counts per locale

	ja-JP	es-ES	zh-CN
authored leaves	1486	1475	1487
candidates (en moved since last edit), before	17	21	17
of which contradicting, before	4	9	4
retired-term scan, wrong at authoring, before	0	2	0
re-translated here	4	11	4
candidates after (at 3fa342501)	13	12	13
contradicting after	0	0	0
retired-term scan after	0	0	0

The after-candidate sets are exactly the before sets minus the re-translated leaves: 0 added, and none of the new values equals its en leaf. The 13 / 12 / 13 that remain are the list below.

The judgment rule

A leaf contradicts the current en when a reader who acts on it would believe something about the current field or object that en now says is false. Three cases:

• en now denies what the leaf asserts;
• the object was re-modelled, so the leaf describes a different thing;
• the leaf names the record's entity by a term that a rename retired from it.

A leaf does not contradict when en only added detail, narrowed or reworded without retracting what the leaf says, or changed punctuation.

Re-translated (19)

path	locales	why it contradicted
sys_two_factor.fields.backup_codes.help	ja-JP, es-ES, zh-CN	Said JSON-serialized, where en says "a single opaque ciphertext, not readable JSON" (266654d82).
sys_notification.description	ja-JP, es-ES, zh-CN	Said "per-user notification inbox entries". ADR-0030 re-modelled the object as one event row per emit(), carrying no recipient; the object header says it "was previously a per-user inbox".
sys_job_run.description	ja-JP, es-ES, zh-CN	Said "audit trail". ca19ee81f changed the source to "history" because the object is job-run history and not the audit trail, which is sys_audit_log (the binding ruling that commit cites).
sys_metadata.fields.environment_id.label	ja-JP, es-ES, zh-CN	Said "Project (deprecated)" for the environment_id column. The v5.0 rename project to environment has no aliases.
sys_business_unit.fields.manager_user_id.help, .effective_from.help, .effective_to.help, sys_business_unit_member.description, sys_business_unit_member.fields.is_primary.help	es-ES	Named the business unit a departamento, where en says business unit (ADR-0057 D2; department is now one kind of business unit). 61ed5c7c9 renamed these in en, ja-JP and zh-CN and missed es-ES.
sys_business_unit.fields.manager_user_id.label, sys_business_unit.fields.id.label	es-ES	Same rename: "Responsable del departamento" and "ID de departamento" against en "Business Unit Head" and "Business Unit ID". These were wrong when authored (e16f2a830), so the since-last-edit instrument cannot see them; the retired-term scan found them.

New values: ja-JP 保存時に暗号化されたバックアップ回復コード（単一の不透明な暗号文であり、読み取り可能な JSON ではありません） · 通知イベント — emit() 1 回につき 1 行（ADR-0030 レイヤー 2 の入口） · バックグラウンドジョブの実行履歴 · 環境（非推奨）. zh-CN 备用恢复码，静态存储时加密（单个不透明密文，而非可读的 JSON） · 通知事件——每次 emit() 产生一行（ADR-0030 第 2 层入口） · 后台任务执行历史 · 环境（已废弃）. es-ES uses unidad de negocio, the word the bundle already uses for the object and its parent_business_unit_id / business_unit_id labels; the rest is in the diff.

Stale but not contradicting (listed, left as written)

path	locales	what en did
sys_team.fields.organization_id.help	all three	Added "Null in single-tenant; auto-stamped in multi-tenant."
sys_business_unit.fields.organization_id.help	all three	Added the same sentence.
sys_business_unit.description	all three	Reworded and extended (canonical BU tree, data-partition role, kind examples). "Hierarchical org tree node" is still true.
sys_business_unit_member.fields.is_primary.help	ja-JP, zh-CN	Nothing to fix: the leaf already says business units. It was translated ahead of en at 61ed5c7c9.
sys_oauth_application.fields.client_secret.help	all three	Added the enforcement citation and "shown once". The leaf's "hashed at rest" is still true.
sys_notification.label / .pluralLabel	all three	Narrowed "Notification" to "Notification Event". The less specific word is not false, and nothing else in these bundles carries it: sys_inbox_message is "Inbox Message", in its own package.
sys_notification.fields.actor_id.help	all three	Reworded "the notification" to "the event".
sys_email.fields.headers_json.help	all three	Added "Never returned on the generic data path…". The leaf does not claim that it is.
sys_email.fields.error.help	all three	Widened from transport errors to transport or pre-delivery validation errors. The leaf is narrower but not contradicted. This is the closest call.
sys_metadata_history.description	all three	Reframed as a durable event log of overlay changes. "Version history" is still true, and the source never retracted the audit wording: 59261af32 calls the row the audit row.
sys_metadata_history.fields.metadata.help	all three	Added "(null for deletes)".
sys_metadata_history.fields.change_note.help	all three	Added a trailing period.

sys_notification.description, the second example the card names, was measured stale and contradicting in all three locales. It is re-translated above.

Dispatch hypotheses, measured

• H1 held. The control is live at f1e921ab8e and at base cd901d7a5, at line 1095 of all four files, and zh-CN is in the set.
• H2 falsified. Neither companion records an authored object leaf. *.source-hashes.generated.ts records only byte copies: 43 / 54 / 42 rows, each an echo of the current en. *.source-hashes.ts is the hand table for apps / dashboards / pages only: 130 rows each, 0 under objects. So "recorded hash differs" names nothing here, and the instrument had to be git history.
• H3 held: values only. Measured that no companion changes. collectFilledFromHashes records a leaf only when its value equals the current source or a recorded digest. At 3fa342501, node scripts/check-i18n-bundles.mjs --write --filter=platform-objects rewrote the bundle set byte-identical (git status empty). Control: with ja-JP sys_job_run.description set to the en string (via scripts/ablation-replace.mjs), pnpm check:i18n goes red with platform-objects DRIFTED (1). Restored after: blob equals HEAD, and git diff HEAD is empty.
• H4 falsified for this set. No echo-decision ledger row, and no test file under packages/platform-objects, names any of the 19 paths (git grep exit 1). The old values appear nowhere outside the bundles (exit 1). Both controls hit: a path pinned in the es-ES ledger, and Cc in the ja-JP ledger. No pin moved.

Verification (all at 3fa342501)

• node scripts/pm/dispatch-gates.mjs --commands (no paths) derived 55 commands, and all 55 exit 0. --ran printed "55 derived famil(ies) accounted for — 55 run, 0 NOT-MEASURED". pnpm check:dual-build-cjs-loads first refused with exit 3 (nine unrelated packages had no dist/). Those were built and the gate re-ran: exit 0.
• Named by the dispatch: pnpm check:i18n ("OK (9 package(s) — all bundles in sync…)"), pnpm check:i18n-stale-fill ("OK (10 bundle set(s) — no new stale fills, 0 baselined)"), pnpm check:nul-bytes, pnpm check:published-files and pnpm check:cross-package-test-inputs all exit 0.
• pnpm --filter @objectstack/platform-objects test: 58 files, 942 tests passed. pnpm --filter @objectstack/platform-objects typecheck: exit 0.
• Lint, narrowed to the three edited bundles: eslint --no-inline-config --format json read 3 files, 0 errors, 0 warnings. ESLint#isPathIgnored answers false for all three, read from the repo's own config. The config never enables type-aware linting (no parserOptions.project), so a change to string values in these files cannot move the verdict on any other file. The full pnpm lint run is left to CI.

Acceptance notes

• Beyond the radius (raw instrument counts at cd901d7a5; meaning judged only where noted):

  • platform-objects metadata-forms: 9 / 9 / 6;
  • plugin-approvals: 17 / 17 / 16. All of these are en options moving from the machine value to a title (pending to Pending), with no change of meaning;
  • plugin-audit: 3 / 3 / 3;
  • plugin-security: 7 / 5 / 7;
  • plugin-webhooks: 3 / 3 / 3;
  • plugin-sharing and the three services: 0.

  Read by hand and reported to the seat as one family finding: plugin-webhooks sys_webhook.fields.definition_json.help in all three locales says the JSON carries the full headers and auth config, where en says credentials are NOT stored there. plugin-audit sys_activity.fields.environment_id.label still says Project. The metadata-forms permission section still promises custom context variables, and the agent section still promises tools; en dropped both.

• bu-rename-consistency.test.ts spells its es-ES term as the case-sensitive /Departamento/. That is why the two lowercase es-ES labels re-translated here (field labels, which the test's header says it asserts) passed it. This PR fixes the values. The test's blind spot remains, and a case-insensitive /departamento/i would close it. That edit is outside this card's file surface and is not made here.

• The history walk needed full history. The shared object store was unshallowed (git fetch --unshallow origin main), which every worktree on this clone now sees.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/platform-objects, touching 6 documentable anchor(s).

25 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json c6b37cd08d0eb622d4f59b79effdc954f86dad9f.

⛔ 9 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• the SDK route bridge reached 54 of 206 client-bound route-ledger rows — the other 152 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 152: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 55 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 97 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.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 3 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 c6b37cd08d0eb622d4f59b79effdc954f86dad9f → packageMentionDocs.

Which tree this was computed on

This run read content/docs from ee85ff77d35feb3a6577fb640d7c0a64ea80eeb3 — the merge of head 3fa342501bc9fb46dfa19c738451027fbb820b61 into base c6b37cd08d0eb622d4f59b79effdc954f86dad9f, 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 ee85ff77d35feb3a6577fb640d7c0a64ea80eeb3 && git checkout ee85ff77d35feb3a6577fb640d7c0a64ea80eeb3
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin c6b37cd08d0eb622d4f59b79effdc954f86dad9f 3fa342501bc9fb46dfa19c738451027fbb820b61 && git checkout -B drift-repro c6b37cd08d0eb622d4f59b79effdc954f86dad9f && git merge --no-ff 3fa342501bc9fb46dfa19c738451027fbb820b61

node scripts/docs-audit/affected-docs.mjs --json c6b37cd08d0eb622d4f59b79effdc954f86dad9f

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

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs c6b37cd08d0eb622d4f59b79effdc954f86dad9f → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[objectstack-fleet]

Contract review

Served-tier: CONTRACT_REVIEW_TIER
Head-sha: 3fa342501bc9fb46dfa19c738451027fbb820b61
Local-runs: none

Inputs: card #20539 (body and its 3 comments: triage 5882208447, Claim 5890841958, os-dev-report 5891725468); PR #20652 body and file list (4 files, +46 / -19); the net diff of the branch against origin/main (merge-base cd901d7a5, origin/main at c6b37cd08 when read); the check-runs on the head. Read-only: nothing built, run or re-run; the en values below are git show origin/main: reads.

Check-runs on the head at 2026-09-29T14:02Z: 32 runs — 18 success, 3 skipped (Console Pin Gate, Build Docs, Packed-tarball smoke), 11 in_progress (Lint and Repo Gates, Test Core 1-6 of 6, Type Check workspace, Type Check consumer gates, Temporal Conformance, the Dogfood Regression Gate aggregate), 0 failure. Recorded as read, not polled.

① Derived judgments

1. Public surface: string VALUES only, inside the three exported translation constants (jaJPObjects, esESObjects, zhCNObjects) of @objectstack/platform-objects. No key added or dropped, no type change, no en edit, no *.source-hashes* companion or echo-decision ledger change, no test change. Accept-set: unchanged — nothing an author can write moves. Right.
2. Each of the 19 re-translated values, judged against the CURRENT en value of the same key on origin/main:
   • sys_two_factor.fields.backup_codes.help (ja-JP, es-ES, zh-CN): en reads "Backup recovery codes, encrypted at rest (a single opaque ciphertext, not readable JSON)". All three now say exactly that — encrypted at rest, a single opaque ciphertext, not readable JSON. Right.
   • sys_notification.description (all three): en reads "Notification events — one row per emit() (ADR-0030 Layer 2 ingress)". All three say that. Right.
   • sys_job_run.description (all three): en reads "Background job execution history" (ca19ee81f replaced "audit trail" with "history", verified in that commit's diff). All three now say history, none says audit. Right.
   • sys_metadata.fields.environment_id.label (all three): en reads "Environment (deprecated)". All three now say Environment (deprecated); each locale's sibling help already said deprecated. Right.
   • es-ES sys_business_unit.fields.manager_user_id.label and .help, .effective_from.help, .effective_to.help, .id.label, sys_business_unit_member.description, sys_business_unit_member.fields.is_primary.help (7): en says business unit in each; es-ES now says "unidad de negocio", the term the bundle already uses for the object label and its business_unit_id labels. Right.
     Nothing is added beyond what en says, and nothing contradicting is left in the same objects.
3. Retired-term sweep of the three HEAD bundles (departamento and proyecto case-insensitive; 部門, プロジェクト; 部门, 项目) against en at the same path: the only carriers are sys_business_unit.description (kind examples) and sys_business_unit.fields.kind.options.department, and en carries "department" at both. 0 mismatches — the dev's "0 after" count holds.
4. The 13 / 12 / 13 candidates left as written: each read against en. All are cases where en added detail or reworded (the two organization_id helps, client_secret, headers_json, sys_metadata_history.*, actor_id, sys_notification.label / .pluralLabel). None asserts what en now denies. The two closest calls: sys_email.fields.error.help (the locales are narrower — transport error only — where en adds pre-delivery validation) and sys_metadata_history.description (locales still say audit trail; unlike sys_job_run, no source commit retired that word — 59261af32 keeps the audit framing). Both are stale-not-contradicting under the card's ruling. Right; a judgment call, recorded.
5. No companion or ledger change: consistent with the tooling as the dev measured it — the source-hashes companions record byte copies of en only, and a hand translation that differs from en is not a recorded fill. The head's own check:i18n and check:i18n-stale-fill verdicts belong to Lint and Repo Gates, in_progress at read.
6. The dev's acceptance note on bu-rename-consistency.test.ts: verified on origin/main line 22 — es-ES is spelled as the case-sensitive /Departamento/, so the two lowercase labels passed a test whose header says it asserts every field label. True.
7. The source commits the PR body cites (ca19ee81f, 266654d82, 61ed5c7c9, e16f2a830, 7bb92056e, 59261af32) exist in history and their subjects say what the PR body says of them.

② Semver level

.changeset/20539-stale-authored-object-leaves.md declares @objectstack/platform-objects patch. The package publishes (no private, publishConfig.access: public, currently 17.5.0), and a bug fix in a released package takes patch — right; not skip-changeset, and no minor is owed because nothing an author can write changes. Clause-②: no, in the PR body and in the changeset body, with no arm — right for a values-only translation diff. Check Changeset is success on the head.

Each factual sentence of the changeset: (a) an authored leaf is kept as written when its en source moves — true (the bundle header, quoted on the card). (b) on a ja-JP, es-ES or zh-CN console the sys_two_factor record page described backup_codes as JSON-serialized where en says one opaque ciphertext, not readable JSON — true (the old values in the diff; en on origin/main). (c) nineteen leaves — 4 + 11 + 4 = 19 value changes in the diff, true. (d) the four keys re-translated in all three locales and what each now says — true. (e) "es-ES only: seven sys_business_unit / sys_business_unit_member labels and help texts" — the count and the substance are true; the kind is imprecise: 2 labels, 4 help texts and 1 description (sys_business_unit_member.description). A nit — the headline line already names descriptions — not a blocker. (f) leaves whose en only gained detail or was reworded are unchanged — true. (g) values only, no key added or removed, no provenance table changes — true on the file list. The title's (#20539) follows the package's own CHANGELOG convention.

③ Boundary flags

open_questions: none.

1. Two es-ES labels found by a retired-term scan outside the card's named instrument — ANSWERED, in scope. The ruling is "re-translate the leaves whose meaning now contradicts en"; the since-last-edit diff was the suggested instrument, not the boundary of the set. Both labels contradict en today, sit in the same file, are values only, and are inside the Claim's file surface. Accepted.
2. Stale-but-not-contradicting leaves left as written — ANSWERED, this is what the ruling asks; the 13 / 12 / 13 are listed in the PR body with what en did. Accepted (①.4).
3. No companion or ledger changed — ANSWERED. The Claim allowed those files only where the regenerator requires it, and the dev's measurement says it requires nothing here. Accepted; the at-head gate family is in Lint and Repo Gates, in_progress at read.
4. Shared clone unshallowed — ANSWERED, a local-environment effect on the dev's clone with no file in the diff. Noted; no repo effect.
5. Out-of-scope finding a (plugin-webhooks sys_webhook.fields.definition_json.help contradicting en on a security field in all three locales; plugin-audit sys_activity.fields.environment_id.label still saying Project; the metadata-forms permission and agent section descriptions promising what en dropped) — ESCALATED to the seat: file the one family card the dev proposed. Not for this PR; outside its file surface.
6. Out-of-scope finding b (bu-rename-consistency.test.ts case-sensitive es-ES regex, "carrier: none") — ESCALATED to the seat as a decision: the guard gap is dormant now that the values are fixed; a one-token case-insensitive spelling would close it. Fold into the family card or leave it. Not a blocker for this head.
7. The other listed deviations (a check:dual-build-cjs-loads re-run after building nine packages, the model-free commit trailer, measurement scripts kept in the scratchpad) are process notes with nothing in the diff.

Governed surface: none touched — the file list is three translation bundles and one changeset; Governed Surface Queue Guard is success.

Implemented-by: claude/issue-20539-stale-authored-leaves
Reviewed-by: session_01DEvba2nBuD4tWzfq8r8NFY

VERDICT: PASS

Landing note for the owning seat: eleven check-run families were still in_progress when this record was read; landing waits on their green, per the Post-Task Checklist.

---

Generated by Claude Code