docs(skills): record ComponentInput's five ADR-0049 tombstones in the plugin-development guide - #7647

Part of #7636 — this implements part 1 only, the half the domain:ui PM ruling dispatched. Part 2 (the gate question) is measured and filed as its own card, #7646, per that same ruling.

What was wrong

skills/objectui/guides/plugin-development.md documented five members of ComponentInput as ordinary writable optionals:

inputType?: string;    // Widget hint for the designer control
min?: number; max?: number; step?: number;   // numeric bounds
placeholder?: string;

All five are ADR-0049 retirement tombstones on the real type. The guide therefore taught a write that is a tsc error at the authoring site and a named parse refusal at runtime. This is the published skills tree — the guide is copied into users' repositories — so the blast radius is outside this repo, which is exactly where the tombstones' named refusal was bought to help.

The fence now records the tombstones as ?: never and points at description as the published remedy, in the real declaration's own wording.

⚠️ The green gate is NOT the evidence — read this part

pnpm check:skill-examples is green on this fence before and after, and that is the point. The fence declares its own local type ComponentInput alias instead of importing the published one, so the gate compiles a private copy and cannot compare it to the real declaration.

Evidence 1 — proven against the real declarations, with lit controls

Two instruments, both run against the built package, both agreeing:

tsc, importing the real ComponentInput and attempting each write. Run twice, once resolved to the published packages/types/dist/base.d.ts and once to packages/types/src/base.ts — identical results:

k_inputType.ts(5,3):   error TS2322: Type 'string' is not assignable to type 'undefined'.
k_min.ts(5,3):         error TS2322: Type 'number' is not assignable to type 'undefined'.
k_max.ts(5,3):         error TS2322: Type 'number' is not assignable to type 'undefined'.
k_step.ts(5,3):        error TS2322: Type 'number' is not assignable to type 'undefined'.
k_placeholder.ts(5,3): error TS2322: Type 'string' is not assignable to type 'undefined'.

Zod, ComponentInputSchema.safeParse per key — all five refused with code=invalid_type and the issue path naming the key, each carrying its own remedy string.

Lit control (both instruments): 2 of 2 genuinely-writable keys read as writable. advanced and description compile clean and parse clean. A control reading zero would have meant a dark instrument and the readings would have been discarded.

An earlier attempt at this instrument WAS dark — it resolved no module at all and reported all seven probes as failures. It was rebuilt with an explicit declaration path, and the --listFiles output was checked to confirm which declaration was actually read.

Evidence 2 — the two-arm control on the gate itself

Both arms had the mutation confirmed on disk by anchored grep counts, and were restored with git checkout HEAD -- PATH, verified by git diff HEAD empty and blob hash identical to the HEAD blob.

Arm B is the finding: the gate judged the fence (it is inside the 13 of 13, not skipped) and passed a snippet wrong about every member it names. That falsification is what justifies #7646.

⚠️ A missed prediction, reported as a miss. The first arm B left the corrected inputType?: never in place and added inputType?: string. Predicted green; it went red — TS2300: Duplicate identifier plus TS2717. That is the gate correctly catching an internally invalid snippet, and it says nothing about drift. The arm was rebuilt as an internally consistent lie before it measured the intended thing.

Gates

All run on this PR's head — re-derived for the actual diff, not just the dispatched list.

check:skill-examples and check:doc-snippets each first returned exit 2, PRECONDITION NOT MET. That is NOT MEASURED — not a pass and not a red. Both were re-run after building what each names via its own --build-filter, and the verdicts above are from those re-runs.

Changeset: none owed, citing the gate's own verdict line — No source or published contract of a released package changed in this range, so no changeset is owed (1 file changed, 0 of them published source).

Line-count ratchet

Docs-only, net +7 lines, same number on both readings:

The +7 buys the tombstone block: what ?: never means, that it is a tsc error and a named Zod refusal, why the keys were never published (the serializer's six forwarded keys), and the description remedy. It also carries the one distinction a reader can get wrong in a costly direction — that BaseSchema.placeholder, the node-level prop, is a different key and is unaffected. The old text was 4 lines of writable-optional declarations; nothing else was rewrapped or re-flowed to buy those lines.

Clause-② determination

Clause-②: no — my own determination, and it agrees with the PM's. The diff is one markdown file in the published skills tree. No schema key moves, no export moves, no accept/reject behaviour moves: the ?: never tombstones and the retirementTombstone() refusals already exist and already behave this way. This PR only stops the documentation from contradicting them.

Landing route

skills/** is a governed surface — agent drafts, human merges. Opened as a draft, and it is not enqueued and should not be. Parking for a human is the expected outcome here, not a failure.

Out of scope, filed not fixed

• finding(scripts,docs): 26 doc fences re-declare a published type instead of importing it, so check:skill-examples and check:doc-snippets judge a private copy — a deliberately wrong fence passes green #7646 — 26 doc fences repo-wide re-declare a published type instead of importing it; 3 of the 13 os:check-marked fences do, and 18 more sit under content/docs/ where check:doc-snippets is opt-out rather than opt-in. Distinct from check-doc-snippet-types 的扫描面不含 skills/ —— agent 复制进用户仓的代码样例,没有任何 typechecker 编译过 #5465, where the scanner does not reach the file at all; here it reaches it and the local re-declaration defeats it.

🤖 Generated with Claude Code

https://claude.ai/code/session_01KbJQ1y1J12nZxYzFWhP8Q3

Generated by Claude Code