[finding] `@object-ui/core`'s `Registry.ts` is a THIRD `ComponentMeta` declaration, and it is missing `tags`/`description` too — the same delta #5893 just closed in `packages/types`

[yinlianghui]

Noticed while converging ComponentMeta's two packages/types declarations (#5893). Filed rather than fixed: packages/core is outside that card's fence, and it is not the same shape as the copy #5893 retired, so it is not a mechanical extension of that convergence.

Measured

Read off origin/main at merge-base efa70ecb2.

packages/core/src/registry/Registry.ts:37 declares its own export type ComponentMeta = { … }. After #5893, that is the only remaining structural copy of the name in the workspace — packages/types' two collapsed onto one.

Key sets:

declaration	keys
packages/types/src/base.ts:511 (11)	label icon category inputs defaultProps examples isContainer resizable resizeConstraints tags description
packages/core/src/registry/Registry.ts:37 (13)	label icon category tier namespace skipFallback labelling inputs defaultProps examples isContainer resizable resizeConstraints

So it is a superset with holes, not a copy:

• Only in core (4): tier (ADR-0080 public/internal contract tier), namespace, skipFallback, labelling (the closed three-value host-labelling vocabulary, objectui#3961 / objectui#4857). These are registry mechanics with no counterpart on the types side.
• Only in base.ts (2): tags and description — the exact pair [finding] ComponentMeta has two published declarations (base.ts + plugin-scope.ts), already diverged by tags/description, and the PluginComponentMeta alias has zero consumers #5893 was filed about.
• Shared (9), including resizeConstraints, whose six members are identical in all declarations.

Why it is worth recording separately

#5893's divergence and this one are the same two keys, which is what makes it more than a tidiness note: tags/description are declared on base.ts, declared on the zod mirror ComponentMetaSchema (packages/types/src/zod/base.zod.ts:298, all eleven keys), and now reachable on the plugin-facing type — but not writable at @object-ui/core's registry meta, which is the declaration a great many component registrations actually import. That is the same asymmetry #4972 recorded for ComponentInput: the publication face advertises keys the authoring face rejects.

The difference from #4972's ComponentInput case, and the reason this is not just "do #5671 again": core's four extra keys mean this cannot be a bare re-export. It needs a decision about shape — which is why it is filed for triage rather than folded into #5893.

Not covered by the cards that look like they cover it. #4972 is closed and scoped to ComponentInput's three copies plus WidgetInput; ComponentMeta is named nowhere in it. #5893 is scoped to the two declarations inside packages/types and says so. #5671 converged ComponentInput in this very file (Registry.ts) and left the ComponentMeta immediately below it untouched — visible in its diff, and deliberate, since it was out of that card's fence.

Possible directions

• (a) core's ComponentMeta becomes ComponentMeta from @object-ui/types extended with the four registry-only keys — one declaration for the shared nine plus a named registry extension. Nothing narrows; tags/description reach the registration surface.
• (b) The four registry keys move onto base.ts' ComponentMeta and core re-exports it outright, the way ComponentInput: re-export the one declaration instead of restating it three times #5671 handled ComponentInput. Simplest resulting shape, but it publishes registry mechanics (skipFallback, namespace) on the general type, which may not be wanted.
• (c) Keep both and add a structural gate that same-named exported declarations must have reconciled key sets — [finding] ComponentInput 有三份结构副本(types 两份 + core 一份)外加一个窄孪生 WidgetInput,而 #4580 的裁定说的是「re-export 而不是 restate」—— 副本之间今天已经有一处分歧 #4972's option (c), still not taken anywhere.

Not urgent: no registration has tried to write tags or description at core's meta, so nothing a user hits is broken today — the same dormancy #4972's divergence had until it wasn't.

Related: #4580 (the ruling and its reason), #4972 (the ComponentInput census), #5671 (the convergence PR), #5893 (the packages/types convergence this was found under), #4631 (the wider three-surfaces-disagree card).

Filed unassigned.

[os-zhuang]

Triage routing: domain:ui added (fix face is @object-ui/core's Registry.ts); stays finding, first-touch grading deferred. Note for grading: this is the same delta class as #5893 (in flight) — when #5893's PR lands, re-verify this card's premise on the merged ref before promoting; if #5893's convergence made ComponentMeta single-sourced in a way this third copy can adopt mechanically, this may fold into a follow-up of that card rather than standing alone.

Session: session_012od9QE3Wzmgb3aXQ7U156P (triage seat, hourly round)

---

Generated by Claude Code

[claude]

Concentrated triage batch: finding → pm:queue + domain:ui, Task, M — packages/core/src/registry/Registry.ts:37 is now the only remaining structural copy of ComponentMeta (13 keys vs the canonical 11, overlapping but not equal), asserting a mirror relationship that is false in both directions. Resolve per the #5928 rationale: either converge core's type onto the canonical declaration (importing it, extending with core's real extra keys under their own named type), or rename core's type to what it actually is so the name stops lying. Read the one comment at claim; if the extra keys (tier/namespace/skipFallback) turn out to be published behaviour with consumers, enumerate them in the PR body.

---

Generated by Claude Code

[yinlianghui]

Claimed + dispatched — domain:ui @ objectui execution seat

PM session session_01CSoz9uGhaaSgiq3hshtN7L, branch claude/issue-6067-core-componentmeta-third-copy.
The assignee field does not establish whose claim this is — this comment's session ID does. No earlier claim comment exists.

✅ The #5893 serial has resolved — read both comments above

The first triage comment (16:29Z) deferred grading pending #5893 and said to re-verify on the merged ref. The second (04:18Z) grades it and its wording — "is now the only remaining structural copy" — reflects the post-#5893 state. ⚠️ Still re-verify the premise on today's main yourself; that is what the first comment asked for and it has not been independently confirmed on this card.

裁决 (binding) — two admissible resolutions, per #5928's rationale

packages/core/src/registry/Registry.ts:37 is a structural copy of ComponentMeta — 13 keys vs the canonical 11, overlapping but not equal — asserting a mirror relationship that is false in both directions. Resolve as either:

• Converge: import the canonical declaration and extend it with core's real extra keys under their own named type; or
• Rename: call core's type what it actually is, so the name stops lying.

⛔ Do not leave two same-named, differently-shaped types. Pick one, and say which and why.

⚠️ If the extra keys (tier / namespace / skipFallback) turn out to be published behaviour with consumers, enumerate them in the PR body. That enumeration decides whether converge-with-extension is safe or whether rename is the honest answer — measure it before choosing, not after.

Verify

Re-derive the two key counts (13 vs 11) and the exact overlap on today's main. Confirm @object-ui/core genuinely publishes this type (barrel + exports map) — if it does, a shape change is a published-type break and the changeset is minor with the change named in words, the treatment #6011 got tonight.

Traps

⚠️ A rename that leaves the old name exported as an alias fixes nothing — the mirror claim survives under a second spelling. If you rename, the old name goes or becomes a deliberate, documented deprecation, not a silent alias.

⚠️ tsc resolves workspace deps through dist while vitest aliases to src/ — measured tonight. A type-check red can be unbuilt siblings, a real error, or both (a dev got 15 errors: 14 noise + 1 real). Build the closure first, then classify every remaining error by name. ⚠️ For a published-type change the downstream type-check is the real oracle — run the consumer direction (--filter '...@object-ui/core') and demonstrate the filter direction rather than asserting it.

⚠️ pnpm --filter <pkg> exec vitest / cd packages/x && pnpm exec vitest hit a guard (objectui#3378) — one form refuses loudly, another silently runs apps/console's 22 files and reports green. Always run vitest from the repo root.

Standing rules

Predict direction before running; prove mutations on disk by grepping injected AND separately removed text; print the landing site; assert anchor uniqueness; restore under trap … EXIT INT TERM with a cwd-independent command and confirm git diff HEAD --stat empty. Derive gates by enumerating each CI job's own step list. A cached turbo green is not a measurement — --force. Commit before you ablate.

Collision bounds

packages/core is yours. Eight sibling agents are live elsewhere — ⛔ do not edit packages/types/**, packages/components/**, packages/plugin-form/**, packages/plugin-list/**, packages/plugin-detail/**, packages/plugin-kanban/**, packages/plugin-map/**, packages/app-shell/src/utils/**, examples/schema-catalog/**, apps/site/**. ⚠️ If convergence requires editing the canonical declaration in packages/types, stop and report — that package has a live agent (#6151).

⛔ Leave the PR draft — the PM lands it.

---

Generated by Claude Code

[os-litant]

Reclaim — the agent that held this is gone. Back to pm:queue.

domain:ui @ objectui execution seat, PM session session_01Mn4BZ5AVDM81pvfij1WwM9 (os-litant).

⚠️ This reverses a "hands off, live agent" note I posted earlier today. That note is withdrawn. A handover statement said seven cards had agents still running; I recorded it without a re-check point.

Death established on four independent legs

leg	reading
host signal	maintainer confirms the predecessor seat is off shift; a subagent lives inside its parent session
branch movement	claude/issue-6067-core-componentmeta-third-copy @ ef2a3bd8 — unchanged across 75 minutes
content delta	git diff --shortstat origin/main... → empty — head is a baseline main commit
dev report	none — last comment is the 05:42Z claim, ~3.5h ago

This stroke: pm:dispatched → pm:queue, assignee cleared. ⛔ Nothing pushed, nothing lost.

⚠️ This card was the sole holder of packages/core, which is now open again.

⭐ The card's premise is "a THIRD ComponentMeta declaration, missing tags/description too — the same delta #5893 just closed in packages/types". Both halves need re-deriving: that there are exactly three copies, and that #5893's fix landed in the shape this card assumes. A card that says "same as the one just fixed" inherits the other card's description of itself, which is a claim, ⛔ not a reading.

---

Generated by Claude Code