fix(gates): answer PREREQUISITE NOT MET with the frame's exit code in the last three i18n gates

[claude]

Fixes #14008
Fixes #13825

Two cards, one PR, because triage bound them: they change the same line of
packages/cli/scripts/check-app-nav-i18n.mjs — #14008 the exit code that banner
announces, #13825 the banner's wording — and landing them apart would only make
them conflict (#14008 triage, R+81 and R+83: 「与 #13825 同笔落地」).

What moved

scripts/import-prerequisite.mjs is the frame 45 gates inherit, and since #13983
it answers PREREQUISITE NOT MET with EXIT_PREREQUISITE_NOT_MET = 3 — the code
five sibling sites already meant by those two words. Three gates still answered
them with 1. They no longer do:

site	before	after
scripts/check-i18n-coverage.mjs — CLI-not-built refusal	EXIT=1	EXIT=3
scripts/check-i18n-coverage.mjs — COULD NOT MEASURE (partial round)	process.exit(1)	process.exit(EXIT_PREREQUISITE_NOT_MET)
scripts/check-i18n-coverage.mjs — POPULATION EMPTY	process.exit(1)	process.exit(EXIT_PREREQUISITE_NOT_MET)
scripts/check-i18n-bundles.mjs — prerequisite refusal	EXIT=1	EXIT=3
packages/cli/scripts/check-app-nav-i18n.mjs — prerequisite refusal	EXIT=1	EXIT=3

Three of the five were exercised live in an unbuilt worktree, exit code captured
BEFORE any pipe, both before and after the edit. The other two sit behind the
CLI-build probe, which fires first, so they are unreachable without a build and
are reported here as source-level readings only — stated rather than glossed.

⛔ No new literal 3. All three files import EXIT_PREREQUISITE_NOT_MET and
EXIT_FINDINGS from scripts/import-prerequisite.mjs and print them into the
advisory, so the number a reader sees and the number the process returns cannot
drift apart. scripts/cli-build-prerequisite.mjs stayed untouched: it did not
need to re-export the constant, because importing the frame directly works from
both roots — packages/lint/scripts/* already reach repo-root gate infrastructure
that way, which is the precedent the packages/cli import follows.

⛔ Only the refusals moved. Each of these paths already printed "nothing was
measured / nothing was compared". Every real-verdict exit and every self-test exit
in the three files is still 1: check-i18n-coverage.mjs keeps 3 of its 6, and
check-i18n-bundles.mjs 4 of its 5, and check-app-nav-i18n.mjs 3 of its 4.

The doc comments that argued for the old number ("inventing a second failure code
would be a new contract nobody asked for") were the strongest remaining statement
of a position #13983 already settled, so they are replaced rather than left to
contradict the code beside them.

The #13825 half — the wording, and its machine-checkable acceptance

The old sentence was true but prescribed echo "EXIT=$?" without saying WHERE, so
a reader doing the natural thing — ... | tail -4; echo "EXIT=$?" — read tail's
status, which is the exact false green the prescription exists to prevent. Per
triage (「抄现成的,⛔ 不要另写一版」) the file adopts #13429's converged paragraph
verbatim apart from this gate's own command. ⛔ No eleventh wording, and ⛔ nothing
appended to the old sentence: the old sentence is deleted.

Acceptance is the census, not the reading:

• git grep -l "no pipe shape repairs it" — 7 files before, 8 after, the new
  one being packages/cli/scripts/check-app-nav-i18n.mjs. That grep now really
  does equal the whole repo; before this PR every "all unified" judgement made
  with it was short by one.
• git grep -n "the PIPE's status" -- 'scripts/**' 'packages/**' — 1 hit before,
  0 after (exit 1 from grep, captured before any pipe).

⚠️ One thing worth knowing for the next editor: the new doc comment on that gate
first explained the defect by quoting the old sentence, which silently put the
census-negative string back and would have left the acceptance criterion failing
while the code was correct. It is reworded, and the comment now says out loud that
the string must not be reintroduced even as a quotation.

Governed rider

.claude/skills/pm-dispatch/references/lanes/services.md:22 read, verbatim:

check:i18n 以「PREREQUISITE NOT MET — workspace CLI 未 build」退 1 不是漂移。

That sentence becomes false the moment these gates move, so it moves in the same
stroke — the minimal factual edit triage specified, 退 1 → 退 3, Chinese text kept
Chinese, nothing else in that file, one character of diff. It is the only place in
the repo keying on the old number (git grep -n "退 1" returns exactly this line).

Because the diff carries a .claude/** edit this is a governed surface PR
(AGENTS.md Prime Directive 14): opened as a draft, never flipped ready, and
auto-merge is never armed by this seat. The PM requests the human reviewers.

Changeset: skip-changeset, and the reading behind it

packages/cli's package.json declares "files": ["dist", "README.md", "CHANGELOG.md"]. scripts/ is not in it, so
packages/cli/scripts/check-app-nav-i18n.mjs is an unshipped dev gate rather than
a shipped file whose advisory text reaches consumers — the npm tarball is
unchanged by this PR. The other three paths are repo-root scripts/** and
.claude/**, neither of which publishes from any released package. ⇒ the diff
publishes nothing, which is exactly what the label is for (AGENTS.md: the label is
for a diff that publishes nothing, never a substitute for a fix's patch
changeset). Had scripts/ been shipped, this would have taken a patch changeset
instead.

Verification

All 38 gate families this card owes, derived AFTER the edits from the real change
set (node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack
— 36 by path + 2 by change kind; the --repo assertion checked against this
checkout's origin), run on the final commit fcdbc6d2, over a workspace
built through the shared verify lock (turbo run build, 67/67 then 70/70 tasks,
both exit 0). Every exit code was captured BEFORE any pipe — redirect to a file,
read $?, then look at the file.

37 of 38 green (exit 0). The two gates this PR edits are among them, and they
answer their real questions rather than refusing:

pnpm check:i18n              EXIT=0   check-i18n-bundles: OK (9 package(s) — all bundles in sync…)
pnpm check:i18n-coverage     EXIT=0   check-i18n-coverage: OK (13 config(s), 691 baselined…, none new)

check:i18n-coverage's own --self-test is part of that pass and it exercises the
missing-CLI-build classifier directly ("the missing-CLI-build, i18n-rule and
per-config-failure classifiers all go red, stay distinct…"), so the refusal path
this PR moved is covered by the gate's own pins, not only by the live runs above.

Also green, and named because they are the families that read what this diff
touched: check:self-test-wired, check:pm-skill-ratchet, check:pm-skill-id-lint,
check:pm-governed-merges, check:skill-frame-sync, check:doc-authoring,
check:published-files, check:undeclared-dep-imports, check:entry-guard,
check:parse-guard, check:watch-hint-literal, check:ratchet-remedy-authority,
check:pnpm-filter-targets, check:cross-package-test-inputs,
check:dual-build-cjs-loads, check:pm-dispatch-gates,
node scripts/pm/bare-root-worklist.mjs --self-test,
pnpm --filter @objectstack/lint run check:doc-formula-expressions. Plus
pnpm check:nul-bytes (8070 files, clean) and a direct control-byte sweep of the
four changed files (grep -naP over the C0 range, zero hits).

1 of 38 NOT MEASURED, declared rather than counted green:
node scripts/check-test-completeness.mjs returns 3 — PREREQUISITE NOT MET — this gate grades a saved 'turbo run test' log, and no log was named. It does not
run tests and cannot produce one; CI tees the log and passes the path on every
invocation. That is a standing property of running it outside CI, identical for
every card, and it is NOT a red and NOT a pass.

pnpm lint (whole-repo ESLint) is CI's run and is not claimed here.

🤖 Generated with Claude Code

https://claude.ai/code/session_01LAwHpn4uVuf4N1geBcD5i3

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

• 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 — 0 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 2263ca4d679026335f559184c0bed4e76d35a242 → packageMentionDocs.

[claude]

PM status (domain:devx execution seat, session session_01LAwHpn4uVuf4N1geBcD5i3): reviewed and ACCEPTED on #14008 (checklist there) and #13825. The one-character change to .claude/skills/pm-dispatch/references/lanes/services.md makes this a governed-surface PR, so it stays a draft and is not armed by this seat; review requested from os-zhuang and hotlong for the human merge. CI on fcdbc6d22 had no red at review time (3 checks still running).

---

Generated by Claude Code

[claude]

PM note on the red (domain:devx execution seat, session session_01LAwHpn4uVuf4N1geBcD5i3): Test Core (1/6) on fcdbc6d22 — packages/cli pnpm run test exit 1 after 14 min (job 100537281700; annotations carry only the exit code, the log download is refused from this container). Ruled not this PR's: the diff changes a refusal exit code + banner text in packages/cli/scripts/check-app-nav-i18n.mjs and two root gates, and no test file anywhere references any of the three gates (git grep over *.test.* / *.spec.* on origin/main: none; packages/cli/test and src: none) — the cli vitest suite cannot see this change, while the shard/package/duration match the #14648 signature that has hit five other PRs on this lane tonight post-fix. Single permitted non-own re-trigger, spent now: update-branch (real merge of current origin/main). Same shard red again ⇒ parked under the #14648 reading; any other red is real and goes back to the dev. The PR stays a governed draft either way (human merge).

---

Generated by Claude Code