Wires the .agents/scripts/audit-translations.mjs script into the
existing pnpm run verify chain as the first gate. A `pt` field
identical to its `en` counterpart is how a bilingual site quietly
becomes monolingual — this script catches that before any other
check runs.
Added an allowlist mechanism: the script now reads a sibling
`.agents/scripts/audit-translations.allowlist.json` file. Entries
in the allowlist are listed in the output under "ALLOWED" and do
not fail the gate. Currently 18 entries: numeric card labels
("01"–"06") in chapters/landing.json and chapters/summary.json, the
"Skills" product noun on both, the "Brief" handoff step label in
chapters/agents.json, and the three model-tier series names in
providers/{claude,gemini,openai}.json.
Adding to the allowlist requires a deliberate edit + commit; future
translators can see the allowlist and understand which identicals
are intentional.
The verify chain order is now:
1. audit-translations.mjs — fail-fast on translation regressions
2. verify.mjs — content + interaction contracts
3. audit-ui.mjs — responsive / no-external-dep audit
4. check-tokens.mjs — design-token enforcement
Ownership notes:
- package.json is owned by the astro-architect agent per
.agents/rules/git-worktrees.md. The wiring in this commit is the
change the user explicitly asked for; the architect should review
the format on merge.
- scripts/verify.mjs is owned by the verification-engineer agent
per the same table. The pre-existing assertion that `package.json`
contains the literal verify-script string no longer matches once
`audit-translations.mjs &&` is prepended. This commit updates the
assertion from a strict `.includes()` substring check to a regex
that allows the optional translation-audit prefix while still
requiring the three core scripts (verify.mjs, audit-ui.mjs,
check-tokens.mjs) to run in order. The regex still rejects any
chain that drops one of them.
Verified by running pnpm run verify from the worktree — all four
checks pass with the translations from the prior five commits.
The repository had two skill directories. `skills/` held the four written
for this project; `.agents/skills/` held the ones the agent context refers
to. Nothing said which an agent should read, and `.agents/ORCHESTRATOR.md`
only ever pointed at the second.
Move the first four into `.agents/skills/` so there is one location, and
add the vendored packages this chapter work used: `animation-vocabulary`
and `improve-animations` (emilkowalski/skills), `teach` (mattpocock/skills),
plus a local `translation` skill and `audit-translations.mjs` for the EN/PT
pairs.
`skills-lock.json` pins the vendored three by source and content hash, so a
later re-vendor is a diff rather than a guess. `.claude/skills/` is
symlinks into `.agents/skills/`, which is what makes them loadable here
without a second copy on disk.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Marcos Paulo