plugin-designer: `DashboardEditor`'s map-valued title stays read-only on a premise the spec invalidated, and the card it defers to (#4163) is closed

[os-support-ai]

Found while implementing #5301 (PR #5427). Not fixed there — different package, and the surface is not this card's file.

What

packages/plugin-designer/src/DashboardEditor.tsx guards its widget-title input with isAuthorableTitle, so a title stored as an inline locale map renders resolved and read-only:

value={isAuthorableTitle(widget.title) ? widget.title ?? '' : titleDisplay}
onChange={(e) =. {
 if (!isAuthorableTitle(widget.title)) return;
 onChange({ title: e.target.value });
}}
readOnly={!isAuthorableTitle(widget.title)}

That branch is correct and cannot lose data — this is an observation, not a defect report. The problem is its stated justification and its tracker.

The premise is stale, and the same file already contradicts it

isAuthorableTitle's doc comment says:

• "Nothing can reach this path from stored metadata yet — I18nLabel was plain string through rc.5, so no persisted widget title can be a map — which is why the branch is safe to take without a ruling on the authoring UX."

Sixty lines above it, resolveWidgetTitle's doc comment in the same file says:

• "@objectstack/spec 17.0.0-rc.6 widened I18nLabel from string to string | Record ... so a widget title may be an inline per-locale map."

Both cannot hold. rc.6 is the pinned version, so a persisted map title is reachable today, and the "safe without a ruling" clause has expired.

The card it defers to is closed

The comment and #4163's own Part 2 name #4163 as the home for the real answer ("a per-locale editor, a translate-this-label affordance, or a deliberate decision that Studio only ever authors the string form"). #4163 was closed as completed on 2026-08-15 while that placeholder is still in the tree, so the deferral now points at nothing. The identical placeholder in DashboardWidgetInspector that #4163 Part 2 describes is in the same position.

Why it is newly actionable

#5301's maintainer ruling (2026-08-20) settled the write-back question for the sibling surface: a save writes back only the active locale's entry and preserves the others. PR #5427 implements that as setLocalized(value, language, next) in @object-ui/i18n, co-located with pickLocalized and pinned by the pairing property. The designer's read-only branch was waiting for exactly this rule, and can now adopt it without a new design decision — an author editing in one locale gets a working input instead of a dead one.

Filed unassigned and unqueued: whether to apply the ruling to the designer surface (versus reopening #4163 for the full multi-locale editor) is a triage call, not this finding's to make.

[os-zhuang]

Triage (objectui finding round): promoted to pm:queue · domain:ui · type Task. Newly actionable by inheritance: the #5301 maintainer ruling (2026-08-20) settled the write-back rule for the sibling surface, and PR #5427 shipped setLocalized in @object-ui/i18n — the designer's read-only branch was waiting on exactly that rule. Scope: adopt setLocalized for DashboardEditor's widget-title input (active-locale write-back, others preserved), fix the two stale doc comments (the isAuthorableTitle premise and the dangling #4163 deferral), and cover DashboardWidgetInspector's identical placeholder in the same pass — same file family, same rule, one dispatch. ⛔ A full multi-locale editor stays out of scope; if the dev believes it is needed, that is a new decision card. S–M, mode:subagent, opus; Clause-②: no.

---

Generated by Claude Code

[os-sales]

Claim: domain:ui execution seat, session session_012u2pRjcqAYtoEjgr3wwhnK · branch claude/issue-5428-dashboard-title-locale-writeback · round 4.

⛔ Correction to the triage note: this is NOT one file family, it is two packages

Triage says to cover DashboardWidgetInspector's identical placeholder "in the same pass — same file family". The two surfaces are in different packages, measured on origin/main:

packages/plugin-designer/src/DashboardEditor.tsx
packages/app-shell/src/views/metadata-admin/inspectors/DashboardWidgetInspector.tsx

The one-dispatch decision still holds — same rule, same stale premise — but plan it as a cross-package diff with two changesets' worth of impact, not a single-package edit. @object-ui/app-shell and @object-ui/plugin-designer are both published.

Both premises verified

DashboardEditor.tsx — the guard is at :300 / :304 / :307, isAuthorableTitle defined at :152, and the dangling #4163 deferral at :292.

DashboardWidgetInspector.tsx:233-250 carries the identical placeholder and the identical stale premise, verbatim:

Nothing in any corpus can hit this path yet — I18nLabel was plain string through 17.0.0-rc.5, so no stored widget title can (…) (objectui#4163, part 2)

That is the claim the card shows is expired: rc.6 is the pinned version, and the same file family already says so elsewhere. Fix the comment as well as the branch — a stale justification left in place is how this recurred.

Scope, per triage

Adopt setLocalized from @object-ui/i18n (shipped by PR #5427 under the #5301 ruling): a save writes back only the active locale's entry and preserves the others. Both surfaces, same rule.

⛔ A full multi-locale editor stays out of scope. If you conclude it is needed, that is a new decision card — say so in the report, do not build it. Clause ②: not tripped (isAuthorableTitle is module-local, no new exports) — but if your implementation adds an export to either package, stop and report before landing.

The pin that matters is the data-preservation one: editing in one locale must leave every other locale's entry byte-identical. A "the input is no longer read-only" assertion is green against a fix that flattens the map, which is precisely the data loss the read-only branch was protecting against.

---

Generated by Claude Code