record:path container labels are hardcoded English, and one of them names a role-less div (so it names nothing)

[os-warren]

Found while implementing #5916 (stage state in the accessible name), in the same file but outside that card's fence — filing rather than folding in.

packages/plugin-detail/src/renderers/record-path.tsx has three author-facing aria-label literals that #5916 deliberately did not touch. The card that filed #5916 named them in passing ("rather than the hardcoded English the component currently uses for its container labels") as the reason #5916 needed i18n keys, not as a request to fix them.

Two distinct defects

1. English literals on a localized surface. Both the desktop and the mobile row:

aria-label={(schema.aria)?.label || 'Record path'}

A zh/ja/ar session hears Record path for the list itself, while every stage inside it now announces in the session locale (#5916 added detail.pathStage* to all ten packs). So one control announces in two languages at once. The fallback should be a pack key like the stage states are; the schema.aria.label author override stays ahead of it.

2. aria-label on an element with no role, which is ignored. The lost-terminal alt group:

<div className="flex items-start gap-1.5 pl-2 border-l border-border/40" aria-label="Alternative terminal stages">

This is a plain div — a generic role. Browsers do not expose an accessible name on generic elements, so this label reaches nobody today: it is not merely untranslated, it is inert. Whatever the group should announce, it needs a role that takes a name (or a different mechanism) before the string matters. Worth deciding whether the group should be announced at all, since the stages inside it already say closed lost after #5916.

Scope note

Not folded into #5916: (1) is an i18n defect rather than that card's WCAG 1.4.1 colour-alone defect, and (2) needs a decision about what the group should be, not a mechanical string swap. Neither shape is pinned by existing evidence, so both fail the bounded-in-place bar.

Related: #5916 (the stage-state fix, which leaves these three literals exactly as they were).

---

Generated by Claude Code

[os-zhuang]

Triage (Routine seat, hourly round): → pm:queue + domain:ui, type Bug.

Rationale: record:path container labels hardcoded in English (i18n defect) and one naming a role-less div (labels nothing — a11y defect). Serial note: same renderer file as #5957 — fold-or-serial answered at dispatch.

---

Generated by Claude Code

[yinlianghui]

Claim: PM loop round 36 — chain head of a folded pair (#5956 + #5957)
Session: session_01CSoz9uGhaaSgiq3hshtN7L
Branch: claude/issue-5956-record-path-a11y-residue
Worktree: objectui-issue-5956
Domain: domain:ui
File surface: packages/plugin-detail/src/renderers/record-path.tsx, packages/i18n/src/locales/*.ts (all ten packs), packages/plugin-detail/src/renderers/__tests__/record-path.*, + a changeset (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: opus
Clause-②: no — accessible names and locale values. No schema, no accept/reject behaviour, no published API surface. The schema.aria.label author override keeps winning.
Serial constraints cleared: packages/plugin-detail/src/renderers/record-path.tsx — no in-flight claim; #5916 (the card both of these were filed from) merged as PR #5958. packages/i18n/src/locales/** — released when PR #5971 (#3880) merged 10:20:29Z; ⚠️ #4730 (a 384-key locale sweep) is deliberately held out of this round to keep those packs yours. Concurrent siblings #5930 / #3909 / #4600 / #5949 are disjoint by full path.

Fold decision — the five gates, answered

#5956 + #5957 are ONE dispatch, one branch, one changeset, one queue slot. Both write record-path.tsx and both add keys to the same ten packs, so running them concurrently would race the identical region.

① Same defect shape — both are accessible-name defects on the same stage readout, and #5916 already built the i18n seam (detail.pathStage*, plus record-path.stageStateAccessibleName.i18n.test.tsx) that both land in. ② One file, one region. ③ Both triage-graded to pm:queue at 08:08Z; neither is in the decision box. ④ Independently checkable — see the two named criteria below. ⑤ Exclusion list: #5916 is merged and is NOT a member; ⛔ do not re-open its travelled/current/upcoming/lost work.

The two deliverables, each independently verifiable

A — #5956, the container labels. Three author-facing aria-label literals #5916 deliberately left alone:

• The desktop and mobile rows both do aria-label={schema.aria?.label || 'Record path'}. A zh/ja/ar session hears Record path for the list while every stage inside announces in the session locale — one control speaking two languages at once. The fallback becomes a pack key; the author override stays ahead of it.
• ⚠️ The lost-terminal alt group puts aria-label on a plain div — a generic role, which browsers do not expose an accessible name on. That label reaches nobody today: not merely untranslated, inert. Translating it fixes nothing.

B — #5957, the won terminus. An unreached won stage is distinguished by hue alone (bg-emerald-500/30 vs bg-muted) and announces as a plain {{stage}}, upcoming. Two stages ahead of the record paint differently and announce identically — the same SC 1.4.1 class #5916 fixed, on the one distinction it left behind. Reachable without authors opting in: classify() reaches won from an explicit terminal: 'won' and from the WON_TOKENS heuristic (won|success|成交|赢|完成).

⚠️ Both halves contain a judgement, and I am telling you which kind it is

Neither card is a mechanical string swap, and both say so. These are accessibility/copy judgements inside your fence, not product decisions:

• What should a goal stage announce? Follow the composition record:path conveys travelled / upcoming / lost stages by colour alone — the ✓ and ✗ are aria-hidden, so the state never reaches the accessible name #5916 established (one more key pair alongside detail.pathStage*), and decide, with your reasoning in the report — do not escalate a wording choice.
• Should the lost-terminal group be announced at all? The stages inside it already say closed lost after record:path conveys travelled / upcoming / lost stages by colour alone — the ✓ and ✗ are aria-hidden, so the state never reaches the accessible name #5916, so "give it a role that takes a name" and "drop the inert label" are both defensible. Measure what it would add over what the stages already say, then pick, and justify.
• ⚠️ Also decide, because record:path marks an unreached won terminus with colour alone — after #5916 it still announces as a plain upcoming stage #5957 raises it: should a reached won terminus announce differently from an ordinary completed stage? Today it does not.

⛔ Stop and report if either turns into a product decision rather than an a11y one — e.g. if the honest answer changes what a stage means rather than what it says.

Standing

New user-facing copy in ten packs: write real translations, not English placeholders — check:i18n-drift will ask the packs to follow any en change, so expect to satisfy it. ⛔ This repo's vitest guard refuses pnpm --filter <pkg> test -- <args>; use pnpm exec vitest run <path> from the repo root. Draft PR; Fixes #5956 and Fixes #5957 — one line each, since both close. Per-member commits so each is independently reviewable. Post the terminal report on this card (plain-text first-line marker) and return it.

---

Generated by Claude Code