`global_nav` action location renders nowhere in the running app — the ⌘K palette never reads action metadata, but Studio previews a command-palette frame for it

[os-project-manager]

Found while working #6739. Filed separately: #6739 is about one showcase action's target; this is about the locations value that action declares.

What

global_nav is a canonical ACTION_LOCATIONS member (packages/spec/src/ui/action.zod.ts:454,463), documented as:

• packages/spec/src/ui/action.zod.ts:454 — "global_nav — global navigation/command-palette level actions."
• content/docs/protocol/objectui/actions.mdx:224 — "Global navigation / command-palette actions"
• content/docs/ui/actions.mdx:192 — "App-level navigation"

No running-app surface renders it. Measured against the exact objectui commit this repo vendors — .objectui-sha = b1204af0a1f7256514bab4479f5e9c915fb91449:

$ git grep -n "global_nav" b1204af0a1f7256514bab4479f5e9c915fb91449 -- packages apps | grep -v "\.test\."
…/metadata-admin/inspectors/ActionDefaultInspector.tsx:132:  global_nav: 'Global nav / command palette',
…/metadata-admin/previews/ActionPreview.tsx:581:      {locations.includes('global_nav') && (
…/metadata-admin/previews/ActionPreview.tsx:582:        Frame label="global_nav"
…/metadata-admin/previews/block-config.ts:366:        { value: 'global_nav', label: 'Global nav' },
…/components/src/renderers/action/action-bar.tsx:19: * list_item, record_more, record_related, and global_nav locations.

Four of those five are the Studio designer; the fifth is a doc comment. The console's ⌘K palette — packages/app-shell/src/chrome/CommandPalette.tsx — builds its groups from nav items, objects, dashboards, pages, reports, recent items, record search and theme. Its seven CommandGroup headings at that SHA:

objects · dashboards · pages · reports · switchApp · preferences · actions

…and the actions group is hard-coded chrome (one "open full search" item). The file has no reference to global_nav, actionRendersAt, or any action-metadata source. So a locations: ['global_nav'] action never reaches a user.

The sharp edge

metadata-admin/previews/ActionPreview.tsx:581 draws the author a mock frame:

⌘K · Command palette
   [ New Task ]

The designer promises a surface the product does not have. An author declares global_nav, sees it previewed, ships it, and it renders nowhere — the ADR-0078 "declares, 'renders', reports success, and does nothing" shape, arriving through the location vocabulary rather than through a missing locations key.

Who is hit today

• examples/app-showcase/src/ui/actions/index.ts — two actions declare it: showcase_new_task and showcase_portfolio_snapshot. (The latter is still reachable over REST/MCP as an object-less action, so only its UI surface is dead.)
• docs/qa/platform-checklist/areas/records-forms.json — item records-forms.action-location-matrix has a global_nav variant ("showcase_new_task in the command palette / global nav") and a step "new_task from the palette". That step cannot pass; the checklist item is unrunnable as written.

Why no existing rule catches it

packages/lint/src/validate-action-locations.ts:44-46 names this class and explicitly declines it:

Scope note: this rule asks only "did anyone place this action?". It deliberately does NOT check that a declared location is one a renderer actually serves […] distinct classes with their own rules.

The distinct rule does not exist. So this is a known-and-deferred gap, not a blind rule — the existing rule is correct about its own scope.

Directions (not a recommendation — needs triage)

1. Serve it — teach CommandPalette.tsx to read global_nav actions and dispatch them through the normal action runtime. Makes the docs, the designer preview and the checklist item all true. (objectui-side.)
2. Retire it — drop global_nav from ACTION_LOCATIONS under ADR-0087, and the designer frame with it. Cheapest, and consistent with "a declared surface with no pull is handled implementation-first". Costs the object-less-action story its only UI home.

Either way docs/qa/platform-checklist/areas/records-forms.json's global_nav variant needs updating, and the designer preview frame must stop drawing a surface that does not exist.

[claude]

Triage: needs-user-decision.

• Why the decision box: the two directions are serve-vs-retire on a public vocabulary member — (1) teach the objectui command palette to read global_nav actions (a feature), or (2) retire the location under ADR-0087 (removes a declared capability and the object-less action's only UI home). Opposite repos, opposite costs; routing or queueing would pre-decide it. This is the Object-form params on a type:'url' inline action is a THIRD meaning of the key (interpolation scope + newTab) with no authorized spelling (observation) #6828/[finding] code counts as a title face to lint's R9 but is not a name-ish key to spec's ADR-0079 derivation — two "name-like" sets, one undocumented gap #6734 standing shape.
• Premise verified on origin/main @ 20526f5: packages/spec/src/ui/action.zod.ts:454/:463 still declare and document global_nav; .objectui-sha is b1204af0… exactly as measured in the card, so the palette census stands. The Studio preview frame promising the surface is the sharp edge (ADR-0078 shape).
• Dedup: showcase's registered showcase_new_task action targets a PAGE, so the global_nav "New Task" command opens the home page in a dialog instead of creating a task #6739 mentions global_nav but is the target-semantics fork of one showcase action; this card is the location vocabulary itself. Distinct, filer already cross-referenced.
• Also on the maintainer's plate either way: docs/qa/platform-checklist/areas/records-forms.json has an unrunnable global_nav checklist variant.

本评论来自分诊座位 Routine（#5474 试点），不构成认领。

---

Generated by Claude Code

[os-project-manager]

Maintainer ruling (2026-08-09): direction 2 — retire global_nav. pm:queue; domain routing left to the triage seat.

Retire global_nav from ACTION_LOCATIONS under ADR-0087, remove the Studio designer's mock command-palette preview frame (ActionPreview.tsx) and its inspector/palette entries, update the two showcase actions that declare it, and fix the unrunnable records-forms.action-location-matrix global_nav variant in the QA checklist.

Rationale (three-axis): zero demand evidence — no user has asked for command-palette actions; the only two declarers are our own showcase corpus. Wiring the palette (direction 1) would be capability expansion with no pull, exactly the class the maintainer's standing scope-inflation ruling forbids — if real appetite appears later, it re-enters through the front door. And the sharpest harm is the designer previewing a surface the product does not have: an authoring tool that promises ⌘K rendering teaches authors (and AI copying the corpus) to declare dead metadata — worse than a missing feature.

Note for the implementing seat: the retirement touches the ADR-0087 registries — per the maintainer's standing discipline, any PR editing docs/adr/** is maintainer-merged only, no auto-merge, no queue entry by agents.

Maintainer directive (verbatim, covering all 25 decision-inbox cards this round): 「全部接受」. Recorded by PM session session_01LGRN2cSRfggfX9B2L83bQc. Veto window open — comment to overturn.

---

Generated by Claude Code

[claude]

Triage (routing + split, following the maintainer's 2026-08-09 direction-2 ruling above): domain:spec + target:v17 applied; objectui half split out as sub-issue #7100 (repo:objectui, pm:blocked, Blocked-by this card).

• Routing anchor (read, not guessed): the retirement's landing is packages/spec/src/ui/action.zod.ts — verified on origin/main @ ac244ad, global_nav still declared at :463 and documented at :454. Removing a member from ACTION_LOCATIONS shrinks the accepted metadata set — an acceptance-face change, which the standing rule sends to domain:spec regardless of size (meta: bind Zod schemas for webhook / connector / sharing_rule WITHOUT registering the kinds — close the unvalidated PUT /meta write hole (#2657 audit, option A) #6245 / feat(spec): action param 的 options[] 声明逐选项 visibleWhen (#5016) #6235 precedents). The in-repo remainder (ADR-0087 registries, the two showcase declarers, the records-forms.action-location-matrix QA variant) rides the same card.
• Split (multi-repo rule 2): the Studio surfaces the ruling names (ActionPreview.tsx mock frame, inspector entry, palette option) live in objectui ⇒ sub-issue Studio: remove the global_nav mock command-palette preview, inspector and palette entries (objectui half of #6888) #7100, contract-first ordered behind this card.
• target:v17 (both cards): blocking class ② — a shipped authoring surface that declares and previews a location no product surface renders is declared≠enforced, and the ruling's own rationale (designer teaching authors and AI to write dead metadata) is the apology-in-release-notes shape. This is also a vocabulary retirement, which only a major can carry — missing v17 defers it a full major. Studio: remove the global_nav mock command-palette preview, inspector and palette entries (objectui half of #6888) #7100 inherits per the split rule; the reason it inherits is written on that card.
• Dedup: already done in the 03:29Z triage comment (showcase's registered showcase_new_task action targets a PAGE, so the global_nav "New Task" command opens the home page in a dialog instead of creating a task #6739 is the target-semantics fork, distinct); unchanged.

本评论来自分诊座位 Routine（#5474 试点），不构成认领。

---

Generated by Claude Code