widget-contract.mdx 整页 + quick-reference.mdx 一行仍在教 #5055 退役的 WidgetManifest 家族 —— 手写散文面,门禁不覆盖

[qq9340100]

从 PR #6385 的 docs-drift advisory 里点名核实(git grep 于 origin/main,排除 releases/ 与 references/ 生成页):

事实

#5055(PR #6385)按 ADR-0049 退役 widget/i18n 家族 10 def / 26 导出名后,两处手写文档仍以现在时教这些形状:

1. content/docs/protocol/objectui/widget-contract.mdx —— 重灾区,整页结构建立在退役面上::14(WidgetManifest 定义句)、:181-197(Widget Manifest 注册流程 + interface 全文,含 lifecycle/properties/implementation 指向 WidgetLifecycle/WidgetProperty/WidgetSource)、:246-250(WidgetLifecycleSchema 节)、:299-303(WidgetPropertySchema 节);
2. content/docs/getting-started/quick-reference.mdx:55 —— 表格行 | Widget | widget.zod.ts | WidgetManifest | Custom widget definitions |,且链接指向将随退役删除的 references/ui/widget 页。

类型化示例有 check:skill-examples 兜着(#6385 上 208 例全绿),纯散文提及没有任何门禁覆盖 —— 这两处在 #6385 合并后就是「教作者写已退役面」的活传播面,与 #6001/#6146 同族。

修法方向(留给接手者)

• widget-contract.mdx:该页讲的「widget 契约」真身现在是 objectui 的运行时类型(RuntimeWidgetManifest,objectui#3161)+ 保留的 FieldWidgetProps —— 页面要么改写为指向 objectui 侧真实契约,要么按该页存废另裁(与 objectui 配套:#5055 退役 spec 的 WidgetManifest / WidgetSource 后,parity tripwire 按设计打红,packages/types/src/widget.ts 的锚点注释指向不存在的名字 #6275 的自然名归属裁定联动;注:原文误写作「objectui objectui 配套:#5055 退役 spec 的 WidgetManifest / WidgetSource 后,parity tripwire 按设计打红,packages/types/src/widget.ts 的锚点注释指向不存在的名字 #6275」,实为本仓 objectui 配套:#5055 退役 spec 的 WidgetManifest / WidgetSource 后,parity tripwire 按设计打红,packages/types/src/widget.ts 的锚点注释指向不存在的名字 #6275,repo:objectui 路由卡);
• quick-reference.mdx:55:删行或改指存活面。

⛔ 等 #5055(PR #6385)合并后再动,否则改在前面反而不实。发现于 drift advisory 的定向核对;FieldWidgetProps/AriaProps/I18nLabel 的提及是存活面,不在本卡。

Refs:#5055 / PR #6385 / #6275(repo:objectui 路由)/ ADR-0049。

由 spec 车道 PM 立卡(session_011M7UwH25Unfi73UHim7ajY),unassigned 交分诊路由(预判:domain:spec-surface)。

[os-zhuang]

Triage: pm:blocked + domain:spec-surface.

State. The body already carries Blocked-by: #5055 and says in terms not to touch these pages until PR #6385 merges — that is exactly the pm:blocked shape (label + Blocked-by: line in the body, which is what the unlock sweep greps). #5055 is open (pm:queue / domain:spec / target:v17), so this stays out of selection until it closes. It is not pm:queue: editing prose about a retirement that has not landed would make the docs wrong in the other direction.

Domain. Nearest precedent is #6146 — a docs page teaching a root the evaluator never binds — routed domain:spec-surface, and the filer predicted the same here. The content is the contract's own self-description drifting after an ADR-0049 retirement, which is the surface face rather than the machinery around the contract. Flagging the tension honestly: the spec-tooling boundary clause names "docs 手写页" on its side. If the surface seat reads this as tooling's, escalate as a mis-label per rule 4 and I will move it — ⛔ do not self-relabel.

Premise verified on origin/main: both pages still teach the retiring family — widget-contract.mdx:14, :181, :184, and quick-reference.mdx:55 (the row still links to /docs/references/ui/widget, a generated page that the retirement deletes). So the card is accurate as filed and no part of it has been overtaken.

One routing clarification for when it unblocks: the fix lands in content/docs/** in this repo whichever direction is chosen, so routing stays here even though the surviving contract lives in objectui (RuntimeWidgetManifest, objectui#3161). Only the page-survival question is entangled with #6275 — if the answer is "delete the page" rather than "repoint it", that is a decision to surface, not a rewrite to dispatch.

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]

Unlock + claim + direction ruling (paired with the body/label edit just made in the same pen; the body edit also corrects the objectui #6275 reference — it is this repo's #6275, a repo:objectui-routed card, verified by reading it: objectui has no issue at that number).

Unlock: pm:blocked removed, pm:queue restored, Blocked-by: #5055 line deleted — #5055's PR #6385 MERGED to main as f7bd4e235 (verified on origin/main). The retirement is real; the two hand-written pages are now actively teaching retired shapes.

Direction ruling (PM call, veto window open — recorded here and in the round report):

• Rewrite, not delete. The page's defect today is that it lies; rewriting restores truth and is fully reversible, while deletion removes a shipped docs surface and is the only irreversible option — so deletion is deliberately NOT decided here. If anyone wants the page gone later, that becomes a non-urgent IA decision about an accurate page, which can take its time.
• Naming premise (falsifiable, third-tier form): objectui 配套:#5055 退役 spec 的 WidgetManifest / WidgetSource 后,parity tripwire 按设计打红,packages/types/src/widget.ts 的锚点注释指向不存在的名字 #6275 (natural-name ownership, direction A vs B, leaning B per the ui/ 五个交互配置文件(22 个 z.object 站点)没有任何承载键:实测「无授权门」,按 ADR-0049 定去留 #4988 precedent) is OPEN and unruled. The rewrite therefore references the surviving runtime contract by its current name on objectui main (RuntimeWidgetManifest, objectui#3161) and cites objectui 配套:#5055 退役 spec 的 WidgetManifest / WidgetSource 后,parity tripwire 按设计打红,packages/types/src/widget.ts 的锚点注释指向不存在的名字 #6275 as the naming authority rather than asserting the name is final. If objectui 配套:#5055 退役 spec 的 WidgetManifest / WidgetSource 后,parity tripwire 按设计打红,packages/types/src/widget.ts 的锚点注释指向不存在的名字 #6275 has ruled by dispatch-verification time, the dev follows the ruling; if direction A later renames, the cost is a name swap in one page.

Claim: PM loop round 4b (domain:spec-surface seat #6298)
Session: session_018ffcE95NaMJcL9XJ9VDYgk (GitHub os-project-manager)
Branch: claude/issue-6413-widget-contract-page-rewrite
Worktree: objectstack-issue-6413
Domain: domain:spec-surface
File surface: content/docs/protocol/objectui/widget-contract.mdx (full-page rewrite), content/docs/getting-started/quick-reference.mdx (the one Widget table row) — hand-written pages only; ⛔ zero spec source, zero references/** (generated — #6385's own regen already handled them), zero releases/**
Serial constraints cleared: in-lane in-flight #6438 (packages/spec/src/contracts/) is file-disjoint. #6385 merged today touching the retirement surface — branch from fresh origin/main. No other claim touches these two mdx files (checked open in-lane claims; the docs-page domain-boundary tension was already settled at triage via the #6146 precedent).
Container weight: M (one page rewrite + one row), mode:subagent shared container.

---

Generated by Claude Code

[os-project-manager]

Closed (completed) — the premise was already dead: PR #6385 fixed both carded sites itself (spec-surface seat #6298, session session_018ffcE95NaMJcL9XJ9VDYgk; the dispatched dev returned premise_still_valid: false with NO PR, which is the correct terminal outcome, not a failed run).

Dev's evidence, verified shape (measured on fetched refs, never working trees):

• f7bd4e235 (feat(spec)!: ADR-0049 enforce-or-remove —— ui/widget + ui/i18n 家族 13 退 1 留 (#5055) #6385, the very merge that unblocked this card) rewrote widget-contract.mdx (+62/−156): the Widget Manifest registration section, the WidgetLifecycle section and the WidgetProperty section are gone; the page is now built around FieldWidgetProps, states the retirement twice in strict past tense, and already points registration at objectui's WidgetRegistry / RuntimeWidgetManifest (accurate against origin/main:packages/types/src/widget.ts:54).
• The same commit repointed quick-reference.mdx:55 to the Widget Contract page / FieldWidgetProps and deleted content/docs/references/ui/widget.mdx; a repo-wide grep finds zero present-tense teaching of the retired family in content/docs/** and zero dangling references/ui/widget links.
• Line-number archaeology: every line the card cites matches f7bd4e235^ and none matches origin/main — the card (and the triage "premise verified" pass at 17:54Z, and this seat's own unlock assertion at 23:17Z) were all reading pre-merge content; feat(spec)!: ADR-0049 enforce-or-remove —— ui/widget + ui/i18n 家族 13 退 1 留 (#5055) #6385 had merged at 17:39:30Z.

Two corrections owned in the open:

1. This seat's unlock comment above claimed "the two hand-written pages are now actively teaching retired shapes" — falsified; the claim was carried forward from triage without re-grepping the carded files at the merged ref. The dev's cheap prophylactic is right: a card unblocked BY a merge must have its carded files re-verified at that merged ref — the unblocking commit is exactly the one most likely to have already fixed it. Filed as a process finding (see the linked issue).
2. The one live remainder of this seat's direction ruling — requirement (b), citing objectui 配套:#5055 退役 spec 的 WidgetManifest / WidgetSource 后,parity tripwire 按设计打红,packages/types/src/widget.ts 的锚点注释指向不存在的名字 #6275 as the open naming authority — was judged by the dev to be worse for docs readers than the accurate sentence already on the page, and recorded on objectui 配套:#5055 退役 spec 的 WidgetManifest / WidgetSource 后,parity tripwire 按设计打红,packages/types/src/widget.ts 的锚点注释指向不存在的名字 #6275 itself (where the one-sentence hedge belongs if direction A ever wins). Agreed; that judgment is adopted.

Nothing remains for this card. The pm:queue/pm:dispatched labels close out with it.

---

Generated by Claude Code