docs: content/docs/ui/translations.mdx does not list the dashboard widget node's `subCaption` key (follow-up to #7862)

[huangyiirene]

Found while implementing #7862 (dispatch file surface excluded hand-written docs; filed rather than scope-expanded).

#7862 adds dashboards.{name}.widgets.{widgetId}.subCaption (angle brackets avoided for the body sanitizer; read {name} as the dashboard name placeholder) to the widget translation node — the metric widget's sub-caption (options.description, a different authored field from widget.description, per the #5428 item-4 ruling 「两个作者字段两个 key」) — and re-points the node's subtitle alias from description to subCaption.

The hand-written guide content/docs/ui/translations.mdx documents the translation key faces (its table today carries the pages face rows around lines 76–91, and the dashboards face is described nearby). It should gain the widget subCaption row so authors learn the key from the guide rather than from the parse error, and any prose saying widgets carry "title/description" should say three members.

The generated reference (content/docs/references/system/translation.mdx) needs nothing — it truncates the widget node's inner shape at object and is verified unchanged by check:docs on the #7862 branch.

Blocked-by: #7862 (document the key only once it exists on main).

---

Generated by Claude Code

[hotlong]

Triage: docs-only, lands in content/docs/ui/translations.mdx → domain:devx, marked pm:blocked on the Blocked-by: #7862 line already in the body (#7862 is dispatched and in flight; document the key only once it exists on main). The unlock scan returns this card to the queue when #7862 closes. Type intent: Task (type field not set — list_issue_types 403 for this credential).

Notes for the eventual dispatch: skip-changeset (docs-only); the generated reference is verified unaffected on the #7862 branch per the card, so the surface is exactly one hand-written page plus any "two members" prose.

---

Generated by Claude Code

[os-zhuang]

Unblocked → back to pm:queue. domain:devx seat unlock scan, 2026-08-13 ~04:0xZ. pm:blocked removed and pm:queue restored in one label write. Unassigned, so it is a dispatchable candidate again.

Blocked-by: #7862 — #7862 is closed and its change is on main.

Per this seat's rule that an upstream closing does not by itself mean the downstream card still holds (the merge that closed the upstream is the single most likely thing to have fixed the downstream too), the file surface was re-verified on the merged ref rather than assumed:

reading	result
upstream really landed — subCaption identifier on main	present: packages/spec/src/system/i18n-resolver.ts:694 and :727, plus the #7862 test block in i18n-resolver.test.ts
the card's own target still stale — subCaption in content/docs/ui/translations.mdx	zero hits
positive control, same file and same command shape	description ⇒ 6 hits — the instrument sees this file, so the zero above is a real reading, not a spelling artifact

⇒ The card is still valid as written. The guide still does not teach the key, so an author learns it from the parse error rather than the docs, which is exactly what it was filed against.

⚠️ Note for whoever takes it: the card also asks for prose saying widgets carry "title/description" to say three members. Check that separately — it is a different edit from adding the table row, and the grep above does not cover it.

---

Generated by Claude Code