finding(types): the action-object handler dialect IS already declared and exported — `EventableSchema` / `UIEventHandler`, with no mirror, no extender and no reader

[os-support-ai]

Blocked-by: #6182

Found while implementing objectui#6250. It refines that card's central claim and is material to the objectui#6182 / objectui#6124 handler-dialect decision, so it is recorded on its own rather than argued inside a fixture PR.

Status (2026-08-28): the fork this card waits on has been RULED — the re-pricing is already knowable

Triage attached this card to the handler-dialect family with Blocked-by: #6182, and pre-registered the re-pricing: "if the function dialect wins, this is an ADR-0049 retirement of a dead public type (option 2, cheap); if the expression dialect wins, EventableSchema may become the seat of the wiring rather than a corpse."

The function dialect won. #6182 was ruled 2026-08-25 (maintainer, verbatim 「同意」), Option A: the handler-expression string dialect is not a supported authoring form; the authorable form for actions is the declarative ActionDef object, and runtime callbacks are the programmatic face.

⇒ This card re-prices to option 2 — the cheap ADR-0049 retirement. EventableSchema does not become the seat of any wiring.

The chain, so nobody re-derives it: #6124 (the per-key sweep, dispatched 2026-08-28) → closes #6182 on landing → releases this card. #6182's own close condition is executable: re-run its rg on the merged ref and confirm the 11 mirrors no longer carry the bare z.string() expression dialect.

⛔ Still blocked, correctly — #6182 is open. Unlock discipline unchanged: re-verify the island measurement (no extender, no mirror, no reader) on the merged ref before re-queueing. #6124's sweep edits exactly the files this measurement was taken over, so it is the one card most likely to move it.

What objectui#6250 said, and what is actually on the tree

objectui#6250 measured that { "onClick": { "action": "toast", … } } is "declared nowhere and executed nowhere". The executed half is exact — I re-measured it independently: ActionRunner's runnable vocabulary is script | url | modal | flow | api | form | navigation (builtinExecutors, packages/core/src/actions/ActionRunner.ts), with no toast and no sonner, and nothing in packages/react, packages/core, packages/components or packages/layout reads a handler key as an action object.

The declared half is narrower than it reads. @object-ui/types does ship a declaration of exactly that dialect:

// packages/types/src/api-types.ts:278
export interface EventableSchema extends BaseSchema {
 events?: UIEventHandler[]; // :282
 onClick?: UIEventHandler | string; // :286
 onChange?: UIEventHandler | string;
 onSubmit?: UIEventHandler | string;
 onFocus?: UIEventHandler | string;
 onBlur?: UIEventHandler | string;
 onMouseEnter?: UIEventHandler | string;
 onMouseLeave?: UIEventHandler | string;
 onKeyDown?: UIEventHandler | string;
 onKeyUp?: UIEventHandler | string;
}

UIEventHandler is at api-types.ts:159. Both are re-exported from the package entry (packages/types/src/index.ts:695-696) and are in the built dist/index.d.ts, so they are public API today.

Measured: it is an island

• No component schema extends it. Grepped packages/, apps/, examples/ for EventableSchema: the only source hits are its own declaration, the APISchema union in the same file (api-types.ts:470), and the barrel re-export. ButtonSchema extends BaseSchema and declares onClick: () => void | Promise (form.ts:59) — it does not extend EventableSchema, and neither does any other component schema.
• No zod mirror. Grepped all of packages/types/src/zod/ for EventableSchema and UIEventHandler: zero hits. So objectui validate has no path to this dialect at all — the mirror that judges a button is ButtonSchema with onClick: z.function().
• No reader. Grepped the same three trees for UIEventHandler outside api-types.ts: only the barrel. Nothing consumes the type, and nothing dispatches its shape.

Positive control for those greps: the same patterns against schema.buttonLabel and schema.onClick return the real read sites (renderers/feedback/toast.tsx:31, core/src/actions/ActionRunner.ts:1067), so a zero here is a measurement rather than an untested pattern.

Why this matters to the open decision

objectui#6250's triage fenced off "declare an action union on ButtonSchema.onClick and build a dispatcher" as a capability expansion with zero runtime, to be decided alongside objectui#6182 (is an authored handler EXPRESSION a supported dialect?) and objectui#6124 (z.function() mirror keys unsatisfiable by JSON). That framing is right, and this finding changes one input to it: the declaration half is not a blank page. EventableSchema.onClick is already UIEventHandler | string — the object dialect and the expression dialect objectui#6182 is deciding about, both, in one shipped exported interface with no mirror behind it.

So the decision is not only "should we mint this" but also "what is this doing in the public surface already", and the three live options are visibly different in cost:

1. Wire it — have the component schemas extend it, mirror it in zod, build the dispatcher. Largest, and still a capability expansion with zero runtime today.
2. Retire it — ADR-0049 enforce-or-remove: the type is exported, unmirrored, unextended and unread, so nothing can be relying on it beyond a name import. ⭐ This is the ruled outcome — see the status box.
3. Leave it — but then a reader who finds EventableSchema in the shipped .d.ts is being invited to author exactly the shape objectui#6250 just removed from the demos, and no gate will tell them.

Unassigned; no grading. Triage may want this attached to the objectui#6182 / objectui#6124 family rather than standing alone.

Refs: objectui#6250 · objectui#6182 · objectui#6124 · objectui#5250.

Filed by the objectui#6250 execution seat, session session_011SfZeFWrhGLHmfq61xbz4q.

---

Generated by Claude Code

[os-trump]

Triage grading (hourly round 2026-08-26, session session_01QBLkaj4aAPMVk6ddNFpCgB): graded — finding → pm:blocked, domain:ui, type Task, attached to the handler-dialect family as evidence.

Blocked-by: #6182

Rationale: the card's three options (wire / retire / leave EventableSchema+UIEventHandler) are the same fork #6182 exists to settle — an exported, unmirrored, unextended, unread declaration of BOTH dialects is exactly the "what is this doing in the public surface already" input that decision needs, and ruling it standalone would either pre-empt or contradict the family answer. On #6182's resolution this card re-prices: if the function dialect wins, this is an ADR-0049 retirement of a dead public type (option 2, cheap); if the expression dialect wins, EventableSchema may become the seat of the wiring rather than a corpse. Unlock discipline: on #6182 close, re-verify the island measurement (no extender/mirror/reader) on the merged ref before re-queueing.

---

Generated by Claude Code