ChatbotSchema.body accepts more than the declared contract

维护者速读

在这个平台上,body 在每一个组件上都是同一个意思:这个组件里面显示什么。只有聊天机器人组件例外 —— 它的 body 收的是一包自由填写的聊天接口参数({ model: ..., temperature: ... })。109 处同名重声明里,这是唯一一处比契约更宽的,其余 108 处都等于或更窄。

以前这个宽度看不见:聊天机器人一旦嵌在卡片或容器里,那包参数就会被拒。objectui#8344 把判定改成「每个孩子按自己的形状判」之后,这包自由参数在任何一层都被接受了(实测:根层、card.body[]、div.children[] 三处全部 ACCEPTED,另有两个对照证明这次读数不是空转)。

⛔ 今天零个文档这么写 —— 语料、fixture、pin 全零。所以这不是「客户正在疼」,是「从现在起谁写谁合法,而且只有这一个组件合法」。

三个选择:

A —— 给聊天参数一个自己的键:body 在所有组件上恒等于「里面显示什么」,聊天接口参数搬到它自己的名字下。一个词一个意思;代价是动一处已发布的面,外面若有文档在写 record 形 body 就要迁移(数量 ⛔ 未量)。
B —— 保留现状,并把它写进契约:聊天机器人的 body 就是可以是一包自由参数,任何位置都接受。零迁移;代价是一个词两个意思永久存在。
C —— 保留,但只在最外层接受:嵌进别的组件时收窄回节点。拦住这次新扩大的那一半;代价是同一个键在不同深度行为不同,比 B 更难跟作者解释。

我的建议是 A(回退:B)。理由在四棱块里;B 的代价不是迁移,是把一条学不会的规则永久化。

裁定之后不必再裁执行:TS 声明与 zod 镜像在同一次改动里改到一致(卡里第 3 点,这对已被 zod-mirror-parity 记为词表漂移),不需要你另外发话。

选哪个:A / B / C?

os-decision-facets

① 项目长远合理性:A 消灭这个平台上唯一一处「同名更宽」的特例;B 把它写进契约变成永久特例;C 还要额外新增一条「按嵌套深度不同」的规则 —— 三者里 C 的特例增生最多,A 最少。
② 实际业务拉动:今天零人撞上(零文档在子槽写它,实测);拉动来自明天 —— 这个宽度从 objectui#8344 起在每一个位置都可达,越晚裁,外面已经写下的文档越多,A 的迁移代价越贵。
③ 防 AI 犯错:B 最危险。同一个词在一个组件上接受自由结构、在其余所有组件上只接受节点,AI 从任一侧学到的规则在另一侧都是错的;而错的那一侧是静默接受 —— 写进去、存下来、没有任何人看见,不是响亮拒绝。A 把这个槽的取值收成闭合的一种。
④ 创业阶段不扩散:一包自由填写的参数是典型的 declare-and-maintain —— 它没有边界,每一次聊天接口加参数都自动变成这个平台的已声明面,而没有人签过这份永久义务。
推荐:A(回退:B)。
本分析看不见什么:我看不见客户手上和已发布文档里有多少个聊天机器人节点正在写 record 形 body —— A 的真实迁移代价因此是未知数,只要这个数不小,答案就该翻向 B;我也看不见聊天接口的 body 参数将来会不会收敛成一组固定的键(若会,A 会比今天更便宜)。

complex.zod.ts#ChatbotSchema mirrors the chat API's body params as
z.record(z.string(), z.unknown()).optional(). That is WIDER than BaseSchemaCore.body,
which is the single-or-list node slot. It is the only wider redeclaration among the 109
base-key redeclarations across the component union's arms — the other 108 are equal or
narrower.

Until objectui#8344 the width was invisible at runtime, because a child slot was judged by
BaseSchemaCore no matter what the child's type said. Redirecting the node recursion point
at the component union makes each child judged by its OWN mirror, so the extra width becomes
reachable at every child slot.

Measured

Both faces built from source, corpus-valid examples/schema-catalog/src/schemas/plugin-chatbot/basic-chatbot.json
as the seed, plus body: { model: 'gpt-4', temperature: 0.2 }. main is the face as shipped;
head is objectui#8501's branch.

document	main	head
the chatbot node at the ROOT	ACCEPTED	ACCEPTED
the same node inside `card.body[]`	REFUSED	ACCEPTED
the same node inside `div.children[]`	REFUSED	ACCEPTED
control, nested OFF-SPEC `icon`	ACCEPTED	REFUSED (the intended narrowing)
control, nested LEGAL `text` node	ACCEPTED	ACCEPTED (non-vacuity)

The two controls are what make the reading a measurement rather than an echo: the same run
shows the narrowing this programme is for, and shows that a legal nested node still parses.

⛔ No corpus document, fixture or pin writes a chatbot node with a record body at a child
slot, which is why objectui#8344's headline (45 refused before, 54 after, over 554 node
documents) does not show this at all. A count of newly-REFUSED documents is structurally
blind to a newly-ACCEPTED one.

Why this is a card and not a line in that PR

PR objectui#8501 declared this widening rather than eliminating it, by seat ruling: removing
it means narrowing a published mirror that deliberately carries the chat API body params, and
no ruling on objectui#8344 authorises that. The declaration is in that PR's changeset, and it
points here.

What a ruling has to decide

Does chatbot keep a record-shaped body on the published mirror at all, or does the chat
API's body params move to their own key so that body means what it means on every other
component?
If the record stays: is it acceptable at every child slot, or should the node slot narrow it
for children specifically?
Whatever is chosen, the TS declaration and the zod mirror have to end up saying the same
thing — the pair is already recorded as disjoint-vocabulary drift in
packages/types/src/__tests__/zod-mirror-parity.test.ts.

Refs: objectui#8344 (the redirect that makes it reachable) · objectui#8501 (declared it) ·
objectui#7759 (lists this same pair from the mirror-versus-declaration side, category D, so
this card is the runtime half rather than a re-file)

Filed by the domain:spec developer seat working objectui#8344, session
session_01CZY49skxUBYyJcdnTcYPrE, from a measurement taken on that branch. Generated with
Claude Code.

Generated by Claude Code

Recorded, not decided — what objectui#8501 shipped and what it deliberately left here

Decision batch #93 (objectui#8344 comment 5585333656) ruled that the NESTED widening this card
describes is eliminated inside PR objectui#8501, and it left the ROOT question — the subject of
this card — open on purpose. Its sentence, quoted rather than summarised:

⛔ Dev option B (narrow the root ChatbotSchema.body) refused too — the root mirror
deliberately carries the chat API body params and is not this card's.

So, as implemented and measured on that PR's head: a chatbot node carrying a record body is
refused at a child slot (the arm the recursion point installs checks it against
BaseSchemaCore.shape.body) and still accepted at the root (the published ChatbotSchema
is untouched). Both directions are pinned in node-recursion-point-8344.test.ts.

⛔ This comment decides nothing here. Whether the root mirror should keep carrying the chat API's
params under the key body is exactly what this card is for, and it is untouched by that PR.

Generated with Claude Code.

Generated by Claude Code

The read-site half this card was missing

Carried here by the domain:ui PM seat from the os-dev seat delivering objectui#9256 (PR objectui#9261), which hit this same key from the opposite direction — a read-site census rather than an acceptance-width one — and correctly held the three chatbot faces out of its narrowing rather than picking one of the two declared meanings silently.

This card measures how wide body is and where that width becomes reachable. It does not record what the renderer actually consumes. That half is now measured:

fact	measurement
the renderer reads `requestBody`, never `body`	`packages/plugin-chatbot/src/renderer.tsx` `:85`, `:274`, `:412` — all three spell `body: schema.requestBody`, i.e. `requestBody` is the value actually sent as the HTTP body
the rename already happened on the declaration side	`packages/types/src/zod/base.zod.ts:123` — "declaration renamed the key to `requestBody`…"
the mirror's omission is deliberate, not an oversight	`packages/types/src/zod/complex.zod.ts:656` — "`requestBody` is deliberately NOT in this pick."
the two sibling faces already carry `requestBody` alone	`complex.zod.ts:707`, `:745` — `requestBody: chatbotRequestBodyArm()`; neither declares `body`

⇒ no renderer read consumes body under either of its two declared meanings. The TypeScript face inherits it as the BaseSchemaCore content slot (SchemaNode | SchemaNode[]); the zod mirror declares it as z.record(z.string(), z.unknown()), "additional API body params". zod-mirror-parity.test.ts's KnownDrift ledger already records the pair verbatim as "Two different meanings of one key — a naming collision to rule on, not a widening" — which agrees with this card's framing and with the hold-out.

What this adds to the decision

⭐ It removes one option from the table. Any repair that keeps body live as the API-params record (folding or alias-refusing requestBody instead) would newly invite a key that nothing reads, against two sibling faces that already spell it requestBody and a declaration side that already renamed it.

⛔ It does not settle the card. Narrowing body back to the BaseSchemaCore node slot is still a published-payload change with the reachability consequences this card measured, and it is a domain:spec contract call. The domain:ui seat is recording evidence here, not ruling.

Also note for whoever takes it: this key's width is, by this card's own measurement, the only wider redeclaration among the 109 base-key redeclarations in the component union — so whatever is decided here is a one-of-a-kind precedent rather than the first of a family.

Generated by Claude Code

决策分析 —— spec@objectui 执行席落卡进决策箱

本卡两条评论都明说不裁:5586312717
「⛔ This comment decides nothing here」,5644616495
「The domain:ui seat is recording evidence here, not ruling」。缺的一直是维护者的一个字母。
本席只补读数与选项。

读数取自 origin/main = 75fca9669a3df84b065c1e9ded0946d67112000d,时刻 2026-09-15T02:23Z–02:26Z。

背景 —— 卡面前提复核,一条没变,一条变了

卡面前提	现在
`ChatbotSchema.body` = `z.record(z.string(), z.unknown()).optional()`	成立,`packages/types/src/zod/complex.zod.ts` 仍写 `body: z.record(z.string(), z.unknown()).optional().describe('Additional API body params'),`
渲染器读 `requestBody`、从不读 `body`	成立,语料 `packages/plugin-chatbot/src`:`schema.requestBody` 3 处、`schema.body` 0 处
「objectui#8344 落地后这个宽度才可达」	⚠️ #8344 的重定向已经在 main 上(`.changeset/8344-node-recursion-point-redirect.md` 与 `packages/types/src/__tests__/node-recursion-point-8344.test.ts` 都在 `origin/main`),而卡 #8344 仍是 open + `pm:dispatched` —— 这是一处半状态,已另行记录

⇒ 卡面写「until #8344 … the width was invisible」的那个「until」已经过去了。但 PR #8501 在同一批
里用 superRefine 把嵌套那一半消掉了,所以今天 main 上的实际形状是:

子槽:拒绝(it('is REFUSED one slot down…'))
根:仍然接受(it('is still ACCEPTED at the ROOT — the published mirror is untouched'))

本卡要裁的就是剩下的这个根。

Governing text

packages/types/src/zod/complex.zod.ts 的 ChatbotSharedMirrorShape 文档块,照抄:
「requestBody is deliberately NOT in this pick. … The two twins below mirror the key the
renderer actually reads, requestBody, and inherit body as the children slot, so they are
born without the collision. Ruling on ChatbotSchema's own body arm is a separate question
and is not decided here.」
⭐ 同一个节点的另一条内容通道已经裁过了。 ChatbotSchema.children 在 main 上是一条
retirementTombstone,理由句里同时点名 body:「REFUSED (objectui#9256, ADR-0049) —
chatbot reads NEITHER content channel: … no renderer read consumes body or children
for this node, and SchemaRenderer strips both out of the props bag it spreads. An authored
value therefore rendered NOTHING — no error, no warning, no element.」
⇒ 两条内容通道,一条已按 ADR-0049 立墓碑拒绝,另一条按「API 参数记录」的身份还活着。

协议声明:协议没有声明这个节点,所以不必先改协议

已安装 @objectstack/spec 17.4.0 的 ComponentPropsMap 里没有 chatbot 行。最接近的是
ai:chat_window,它的键是 ["agentId","aria","context","mode"] —— body、requestBody、
children 三个都不在(对照:该行确实存在且有 4 个真键;缺席词 NoSuchKeyXyzzy 为假)。
⇒ 这是 objectui 自有面。维护者「协议不正确的应该先修改协议」那一句管的是协议说错了的情形,
这里协议没说。

前提(带 re-check 命令)

cd /home/user/objectui && git fetch origin main
git grep -o 'schema\.requestBody' origin/main -- packages/plugin-chatbot/src | wc -l   # 期望 3
git grep -o 'schema\.body'        origin/main -- packages/plugin-chatbot/src | wc -l   # 期望 0
git grep -ln '"type": *"chatbot"' origin/main                                          # 期望 3 个文件

语料里没有任何一份真文档写 body。 三份 chatbot 文档(examples/schema-catalog/src/schemas/plugin-chatbot/
下的 basic-chatbot.json、chatbot-with-timestamps.json、customer-support-chat.json)逐份解析,
顶层键集里 body 零次、requestBody 零次。⇒ 无论裁哪个方向,本仓新被拒绝的文档数为零。

一句话问题

同一个节点上,一个叫 body 的键今天可以被作者写、会被接受、然后什么也不渲染 —— 而它在
别的每个组件上的意思(放子内容)和在这里的意思(塞 API 参数)是两回事。

选项 × 真实代价

选项	做什么	客户可感知的后果
A	在 `ChatbotSchema` 上给 `body` 立墓碑(REFUSED),拒绝文案把作者指向 `requestBody` —— 与同节点的 `children` 已有的处置同形	作者写 `body` 当场被响亮拒绝并被告知写什么。本仓实测零份文档受影响;⚠️ 已发布版本里它一直被接受,仓外未测
B	维持现状:根继续把 `body` 当 API 参数记录,只把 TS 声明与 zod 镜像改到同一说法	什么都不变;代价是永久留一个没有任何读者的已声明键,而两个兄弟面已经统一写 `requestBody`
C	把 `body` 收窄回 `BaseSchemaCore` 的子节点槽(和其它每个组件一个意思)	名字的歧义消失,但 `chatbot` 两条内容通道都不读 ⇒ 作者写了仍然静默什么也不渲染,只是换了个拒绝不了的理由

A 与 C 都要反转(⛔ 不是删除)node-recursion-point-8344.test.ts 上两条 load-bearing 的钉子:
根那条 it('is still ACCEPTED at the ROOT …'),以及类型钉
ArmsNotAssignableToSchemaNode ≡ 'chatbot'(收窄后它会变成 never)。B 不动钉子。

业务含义直译

A = 「这个词在这里是写错的,当场告诉作者该写哪个」。
B = 「留着一个谁也不用的字段,永远维护它」。
C = 「把词义统一了,但作者写了还是白写,只是不再算写错」。

四轴(从业务立场)

① 项目长远合理性:这是 109 个 base-key 重声明里唯一一个更宽的(树上自己这么写)。A 把
这个孤例消掉;C 消掉歧义但留下一个合法却无效的键;B 把孤例变成永久承诺。
② 实际业务拉动:今天谁撞上?没有人 —— 三份 chatbot 文档一份都没写它,渲染器三处读的
全是 requestBody。零存量拉动,所以这不是救火,是趁没人用的时候把它收掉。
③ 防 AI 犯错:本卡最硬的一轴。今天一个 AI 在根上写 body: { model: 'gpt-4' },得到的是
接受 → 被剥掉 → 什么也不渲染,全程没有一个字提示它。A 把这条路变成响亮拒绝并给出正确
拼写(retirementTombstone 现成就是这个形状,children 已经这么做了);C 仍然接受、仍然
静默;B 等于把这个静默写进契约。
④ 创业阶段不扩散:每个已声明的键都是永久义务。body 在这个节点上没有读者,按
「零拉动默认 defer 或 remove」,remove 是默认方向;B 是 declare-and-maintain 的反面。

裁后执行

裁 A ⇒ spec@objectui 一张卡:ChatbotSchema.body 换 retirementTombstone(文案照
children 那条的形状,指向 requestBody);同 PR 反转上面两条钉子,⛔ 不删;changeset
记 breaking;zod-mirror-parity.test.ts 的 KnownDrift 行随之改写为已裁。
裁 B ⇒ 一张卡只做声明对齐(TS 面与 zod 面说同一件事),KnownDrift 行从「待裁」改「已裁保留」。
裁 C ⇒ 同 A 的钉子反转,但 body 留成子节点槽;另需一条说明为什么它在本节点无效。
三者都不裁 ⇒ 本卡留在决策箱,⛔ 不回 pm:queue:树上的钉子已经把现状钉住,不会漂移。

维护者速读

聊天机器人组件上有个字段叫 body。作者可以写它、系统会收下、然后一个字都不显示,也不报错。
真正生效的是另一个名字(requestBody),我们的代码三处读的都是那个,两个兄弟组件也早就只写
那个。查下来:今天没有任何一份真文件写过 body,所以现在处理它,谁都不会被影响。

而且同一个组件上另一个一模一样的字段,我们去年已经裁过了 —— 判的是「拒绝,并告诉作者该写
什么」,当时写下的理由里就顺带点名了 body 也没人读。

三条路:A 照旧例把它也改成「当场拒绝 + 告诉你写哪个」。B 什么都不动,永远养着一个没人
用的字段。C 把它的意思改成和别的组件一样,但作者写了还是白写,只是不再算写错。

我建议 A,不行就 C。请裁一个字母:A / B / C。

Generated by Claude Code

更正 —— 上一条里关于 #8344 卡面状态的那句话是错的

上一条把「#8344 的重定向已在 main、而卡仍 open + pm:dispatched」读成一处半状态,并写了
「已另行记录」。两半都不成立:

这个 open 状态是已裁的,不是半状态。spec(types): redirect the node recursion point from BaseSchemaCore to AnyComponentSchema — measured at 9 newly-refused documents, and it drops 118 phantom strict refusals #8344 的验收评论
5589224002
原话:「Card stays open (Part of, pm:dispatched, assignee os-sales): the
strict-authoring-face programme continues」。
而且已经有席位查过并明确不动它:
5590236143
(domain:spec @ objectui,读数 2026-09-08T18:52Z)把 check-half-states.mjs 的三行贴在卡上,
并写明「⛔ A ruled state is not a half-state to be tidied by a passing seat」。
「已另行记录」也是空的:本席没有另外记录任何东西,因为没有东西需要记录。

⭐ 教训归本席:我从两个标签(open + pm:dispatched)推出了一个状态判断,却没有读那张卡的评论。
⛔ 标签不是状态的全部,卡上的裁决才是 —— 这与本卡自己的纪律(读到最后一页再下判断)是同一条。

本卡的结论不受影响。「until #8344 落地,这个宽度不可达」的那个 until 确实已经过去,但那一条
是直接从树上读的(.changeset/8344-node-recursion-point-redirect.md 与
packages/types/src/__tests__/node-recursion-point-8344.test.ts 都在 origin/main),与 #8344
的卡面状态无关。选项、四棱与推荐 A 一字未改。

Generated by Claude Code

Ruling: batch #137 item 4 · letter A · maintainer 「同意」 2026-09-15T22:57Z

Director seat, session session_01WCEaPsmKY4UyoivKkkaUHt. Presented on 2026-09-15 with this seat's own recommendation A; the maintainer's one-word reply adopts the batch as recommended.

What is ruled

A — ChatbotSchema.body becomes a retirementTombstone (REFUSED by name), its text pointing the author at requestBody, the same shape the same node's children already carries (objectui#9256 / ADR-0049, whose reason sentence names body as equally unread). The published record-shaped body arm on this node retires; body means one thing platform-wide.
B (keep and document) is refused; C (narrow to the children slot) is refused — the node reads neither content channel, so C would keep a legal, inert key.

Execution

needs-user-decision → pm:queue, domain:spec @ objectui, no assignee. Dispatch scope: the tombstone on both faces (TS ?: never and the zod mirror); the two load-bearing pins in node-recursion-point-8344.test.ts (the root is still ACCEPTED case and the ArmsNotAssignableToSchemaNode ≡ 'chatbot' type pin) are inverted by this ruling, ⛔ not deleted; zod-mirror-parity.test.ts's KnownDrift row becomes 「ruled」; Clause-②: yes; changeset minor on @object-ui/types with the BREAKING banner and the ADR-0087 disposition line. The report names the delivery surface of the refusal (parse-time) and whether any consumer path skips the root parse.

Prior rulings read: chatbot,body,requestbody,tombstone,record,params → 54 hits; ADR-0094 D1/D2, ADR-0110 D2, ADR-0006 D1, ADR-0020 D4, ADR-0029 D9.6, ADR-0044 D1/D3, ADR-0056 D9 — none governs this card; objectui#9256 / ADR-0049 (the children tombstone) is the governing ruling.

Generated by Claude Code

Claim: PM loop round 5 (第二张,补本轮腾出的那个位)
Session: session_01VCpmqvacV4BypY48QdoxcE
Branch: claude/issue-8572-chatbot-body-retire
Worktree: objectui-issue-8572
Domain: domain:spec
File surface: packages/types/src/complex.ts · packages/types/src/zod/complex.zod.ts · packages/types/src/__tests__/node-recursion-point-8344.test.ts(两条 pin 反转,⛔ 不删)· packages/types/src/__tests__/zod-mirror-parity.test.ts(KnownDrift 行改「ruled」)· .changeset/(stop on breach; explain in the report)
Container & model: M,mode:subagent,model: default judgment tier (opus) —— 强制条款②规定的是施工在判断档,复核才必须达档。
Clause-②: yes —— 裁决自己写死了这一行(「Clause-②: yes」),本席不另判。needs:contract-review 已在与本认领同一笔标签写入里挂上卡;PR 侧载体在 PR 一存在就挂(⚠️ 受管面守卫的 merge_group 那条腿读 PR 标签,只挂卡不够 —— 本轮 PR #9627 上正是本席补挂的)。
Thread-read: 5689209004
Serial constraints cleared: ⭐ 本条纠正了本席继承台账里的一处过宽记载。台账记「Chain A = objectql.ts · objectql.zod.ts · zod-mirror-parity.test.ts」,由 PR #9540 冻结持有。实测 PR #9540 的文件清单(/pulls/9540/files,⛔ 不是回忆):.changeset/9309-…md · __tests__/object-gallery-filter-9309.test.ts · objectql.ts · objectql.zod.ts —— 四个,不含 zod-mirror-parity.test.ts。⇒ 台账把持有面写宽了一个文件,该 pin 自由。另扫本仓全部 open PR 对本卡四个文件的命中:0(逐 PR 取 /pulls/N/files)。本轮在飞的 #9263(components/renderers/layout/page.tsx)与 #9627(plugin-detail/**、types/src/views.ts)与本面均不相交。

⚠️ 复核路径。本席被服务 claude-opus-5,低于 CONTRACT_REVIEW_TIER = claude-fable-5-1。按 2026-09-16 维护者裁决(「起达档复核子代理」,objectstack#18418),条款②复核由隔离的达档复核子代理出具,只喂卡片、其裁决与 PR —— ⛔ 不喂本派发令、⛔ 不喂本席结论、⛔ 不喂开发报告 —— 并且逐字采纳或整体作废。其档位以它自己 transcript 的 message.model 为读数。该路径本轮已跑通两次并双双落地:objectui#8800 / PR #9622(88/88)与 objectui#8801 / PR #9621(79/79),两个 PR 都已 MERGED。

派发范围逐字来自裁决,⛔ 不由本席改写

批次 #137 第 4 项 · 字母 A · 维护者「同意」(5689209004):

Execution — Dispatch scope: the tombstone on both faces (TS ?: never and the zod mirror); the two load-bearing pins in node-recursion-point-8344.test.ts (the root is still ACCEPTED case and the ArmsNotAssignableToSchemaNode ≡ 'chatbot' type pin) are inverted by this ruling, ⛔ not deleted; zod-mirror-parity.test.ts's KnownDrift row becomes 「ruled」; Clause-②: yes; changeset minor on @object-ui/types with the BREAKING banner and the ADR-0087 disposition line. The report names the delivery surface of the refusal (parse-time) and whether any consumer path skips the root parse.

B(保留并写进契约)已否;C(收窄到 children 槽)已否 —— 该节点两个内容通道都不读,C 会留下一个合法但惰性的键。治本卡的是 objectui#9256 / ADR-0049(children 那个碑文)。

裁决最后那句是本卡最难的一问,⛔ 不要糊弄它

The report names the delivery surface of the refusal (parse-time) and whether any consumer path skips the root parse.

⇒ 碑文的拒绝发生在解析时。所以「有没有哪条消费路径绕过根解析」直接决定这次退休到底拦不拦得住人:若存在绕过根解析的路径,那条路上作者写的 body 仍会被接受,退休就只在文档上成立。⭐ 这一问要实测、要带发火对照、要把通道追到端,⛔ 不接受「大概都走根解析」。

⚠️ 已知的相邻事实,说在前面免得当成新发现

该节点的渲染器读的是 requestBody,从不读 body(前序席位实测:packages/plugin-chatbot/src 里 schema.requestBody 3 处、schema.body 0 处)。
卡面记载 objectui#8344 落地后,这个宽度在根层与子槽都可达。⭐ 那是更早的读数,main 之后走了很多;⛔ 重取,不要引用。

⛔ 两条来自本轮的教训,直接适用于你

计数证明不了退休。 本轮两次撞到:碑文注释会让 grep -c 读出假阳(feat(types)!: retire AIInsightsSchema and the ai-insights node type (objectui#8800) #9622 上目标符号与对照同为 1;refactor(types)!: retire ObjectKanbanSchema.allowCollapse on both faces (#8801) #9621 上目标符号在注释里还有 3 处)。⇒ 用结构性读法(剥注释、按声明形状),并在同一文件的同一写法上配点亮对照。
面外即停,⛔ 不顺手改。 报告里写清楚,由 PM 立卡。本轮三条面外发现都已立卡(objectui#9624 / finding(changeset): .changeset/8178-...md says AIInsightsSchema.objectName is untouched and pin-held — objectui#9622 deletes the schema and the pin, and both changesets publish in the SAME release #9625 / finding(types,plugin-kanban): KanbanColumn.collapsed is declared on both faces and honoured only by KanbanEnhanced, which no production source imports — the key objectui#8801's tombstone points authors at #9628)。

仪器陷阱(本仓适用,逐个判,⛔ 无统一规律)

dispatch-gates.mjs 拒答跨仓 ⇒ 门禁清单自 objectui 的 check:* 与 .github/workflows/ 手推,--ran 对账写 UNAVAILABLE,⛔ 不伪造。
check-expected-skips.mjs 带的是 objectstack 的花名册,指向本仓会静默答错 ⇒ ⛔ 不要用。
⛔ 共享检出 /home/user/objectui 没有 node_modules ⇒ 任何要解析依赖的探针在那里是哑火零。一律在自己的 worktree 里 pnpm install 之后再测。
⛔ 永不 grep 掉工具输出;退出码在任何管道之前落盘。报告缺席永不读作成功。

— PM domain:spec · 认领时刻 2026-09-16T15:27Z

Generated by Claude Code

os-dev-report

{
"issue": 8572,
"status": "blocked",
"branch": "claude/issue-8572-chatbot-body-retire",
"pr": "#9639",
"session": "session_01VCpmqvacV4BypY48QdoxcE",
"premise_still_valid": true,
"summary": "Ruling A is implemented in full on the declared surface and is green there: ChatbotSchema.body is an ADR-0049 retirement tombstone on both faces (TS ?: never, zod retirementTombstone naming requestBody), the two load-bearing pins in node-recursion-point-8344.test.ts are INVERTED not deleted, zod-mirror-parity's KnownDrift row records the ruling and drops the key, and the changeset is minor on @object-ui/types with the BREAKING banner and the ADR-0087 D2 disposition line. BLOCKED on ONE thing only the PM can grant: the file surface. Three assertions OUTSIDE it are inverted by this same ruling and are red until they are edited - content-channel-family-d-9256.test.ts's runtime LIVE CONTROL ('chatbot still accepts body') and its type-level CONTROL ('chatbot still TYPE-CHECKS with body'), and strict-authoring-face-8345.test.ts's 'the chatbot body clause still fires at a child slot', whose ROOT half asserts acceptance. Two more out-of-surface texts are now false: defineNodeComponentUnion's comment block in zod/base.zod.ts ('a root chatbot with a record body still parses', plus its instruction not to narrow ChatbotSchema because objectui#8572 has not ruled) and the requestBody row in content/docs/plugins/plugin-chatbot.mdx ('writing body sets the base schema's children container'), which AGENTS.md #2 makes a required companion edit. Per the dispatch order ('anything else: stop and report'), none of the five was edited. PM assignee field: untouched by me, as ruled; the card already carried assignee os-elon-musk's dispatch and hotlong's claim.",
"premise_notes": "Re-taken, not cited: the card's older reading said the width was reachable at the root AND in child slots. On this base only the ROOT was reachable - PR objectui#8501 had already eliminated the nested half with a superRefine on the installed arm - so what this card retires is the root. The two other premises hold: the zod arm was still the record, and the renderer still reads requestBody and never body.",
"tests": "At HEAD ccf64eb, tracked tree clean. (1) pnpm exec vitest run packages/types/: 2 failed / 4605 passed across 195 files - both failures are the out-of-surface hold-out controls named in summary; every in-surface pin passes. (2) pnpm --filter @object-ui/types run type-check: exit 2, ONE error, the out-of-surface satisfies ChatbotSchema control at content-channel-family-d-9256.test.ts(572). (3) pnpm --filter @object-ui/types run build: exit 0, dist completeness 130 emitted files. (4) DELIVERY SURFACE, on BUILT artifacts (cli + closure built from this branch): objectui validate on a root chatbot carrying body exits 1 with 'Path: body', 'Code: invalid_type' and the message naming requestBody; the same document spelled requestBody exits 0 (LIT CONTROL); nested in card.body[] exits 1. (5) ROOT-PARSE QUESTION, measured: YES, consumer paths skip the root parse. SchemaRenderer validates in dev builds only and through @object-ui/core's hand-written validateSchema, which never consults these mirrors; rendered with console.warn spied, a chatbot+body node warns 'schema.children.type: type is required' (the generic child walker mistaking the API params for a child node - wrong key named, wrong remedy), identical before and after this change, and nothing at all in a production build. Controls in the same pass: chatbot+requestBody and plain chatbot warn nothing (LIT), a card child without a type warns (FIRING). The VS Code extension has its own walker and never imports the zod face. Root-parse consumers that DO get the refusal: @object-ui/cli's validate and check (safeValidateSchema on the document root) and tsc. (6) ABLATION from the committed state, two legs, each mutation proven on disk (scoped structural counters with untouched-neighbour controls, plus on-disk hash vs HEAD blob) BEFORE any colour was read. Leg A (zod arm back to the record): the inverted root pin fails, the ArmsNotAssignableToSchemaNode type pin fails, and the parity WIDER ratchet names complex.zod.ts#ChatbotSchema and its key body. Leg B (delete the TS member, keep the zod tombstone): vitest stays green (a ?: never is erased before a test runs) and the KnownDrift ratchet fails naming the same pair - the other direction. Both legs restored with git checkout HEAD -- path, verified by an empty git diff HEAD AND the file hash returning to its HEAD blob. DECLARED: the FIRST ablation attempt was a NULL MEASUREMENT - its anchor matched the first body: retirementTombstone( in the file, which belongs to CalendarViewSchema - so it reddened pins about another pair and said nothing about this change; it is declared in the PR body rather than silently re-run. (7) Gates: check-changeset-presence exit 0; changeset:check exit 0; check:changeset-claims exit 0 (20 named, all 20 answered in the PR body: 20 still true, 0 falsified); check:new-line-citations 0 new; check-control-bytes exit 0; check:doc-types, check:prompt-keys, check:component-surface-parity exit 0; check:doc-snippets first read PRECONDITION NOT MET (exit 2, measured nothing) and after the scoped 35-task build was RE-RUN to exit 0, 670/670 blocks. NOT MEASURED locally: the whole-repo gate farm (pnpm lint and the rest) - left to CI. --ran reconciliation: UNAVAILABLE (dispatch-gates.mjs refuses cross-repo; the list was hand-derived from this repo's own check:* scripts and workflows, and check-expected-skips.mjs was deliberately NOT pointed at this repo). DECLARED NARROWING: the ablation and the final runs ran WITHOUT the shared verify lock - two consecutive acquisitions hit queue-timeout (exit 99, 9m00s each, slot dev-8572 kept) behind a live holder (pid 21804, a sibling worktree's full packages/ vitest, held 22m+ and still running); the narrowed runs are two named test files plus one tsc project and every reading cited is a verdict, not a wall-clock figure.",
"mcp_calls": "0 - no MCP GitHub tool was called; every GitHub read and write went through the REST proxy with curl",
"api_writes": "2 REST writes: POST /repos/objectstack-ai/objectui/pulls (draft PR 9639), POST /repos//issues/8572/comments (this report). Plus 2 git pushes (the empty-branch routing probe, then the commit). NO label write: needs:contract-review is the PM seat's on both carriers per the dispatch order, and objectui's changeset-exemption label is a phantom no workflow reads, so applying one would have been a write with no mechanism behind it.",
"open_questions": [
{
"question": "The file surface has to grow by two test files for this ruling to land green. Which route?",
"options": [
"A - extend this PR's surface to content-channel-family-d-9256.test.ts and strict-authoring-face-8345.test.ts and invert their three hold-out assertions here (they are the same inversion the ruling already orders for the two named pins)",
"B - land those inversions as a separate card, leaving this PR red until both merge together",
"C - dispatch them to another seat in parallel"
],
"recommendation": "A, because the three assertions encode the HOLD-OUT that this ruling ends, they are mechanically forced (each is one boolean or one ts-expect-error), and splitting them leaves main red between the two merges - B and C both buy a window where the tree does not build."
},
{
"question": "Two non-test texts are falsified by this change and are also out of surface: defineNodeComponentUnion's comment block in packages/types/src/zod/base.zod.ts, and the requestBody row in content/docs/plugins/plugin-chatbot.mdx. AGENTS.md #2 makes the docs one a required companion edit for any refactor.",
"options": [
"A - add both to this PR's surface (prose only, no behaviour change)",
"B - file them as follow-on cards",
"C - docs here, base.zod.ts on its own card together with whatever is decided about the now-idle superRefine clause"
],
"recommendation": "C, because the docs row is a required companion under AGENTS.md #2 and is prose-only, while base.zod.ts's comment sits on top of a real design question - the superRefine chatbot clause now has nothing left to narrow - and that question deserves the file's own card rather than a comment patch smuggled in here."
}
],
"out_of_scope_findings": [
"surface extension required (not a finding, a blocker): content-channel-family-d-9256.test.ts - runtime 'LIVE CONTROL - chatbot still accepts body' and type-level 'CONTROL - chatbot still TYPE-CHECKS with body'; strict-authoring-face-8345.test.ts - 'the chatbot body clause still fires at a child slot' asserts the ROOT is accepted on both faces. All three are inverted by this ruling in the same sense as the two named pins.",
"to file (dedupe words: base.zod.ts, defineNodeComponentUnion, superRefine, chatbot, body, narrowing): the comment block on defineNodeComponentUnion states that ChatbotSchema.body mirrors the API params as a record and that 'a root chatbot with a record body still parses', and instructs readers not to narrow ChatbotSchema because objectui#8572 has not ruled. All three sentences are false as of this PR. The clause itself is left installed and the 'fill is LIVE' pin still reads its installation.",
"to file (dedupe words: plugin-chatbot.mdx, requestBody, table row, body, children container): the docs row says writing body 'sets the base schema's children container and never reaches the API'; the remedy is right and the mechanism is now wrong - it is refused by name. AGENTS.md #2 (docs-driven) makes this a required companion edit.",
"noted, not filed: the superRefine chatbot clause in defineNodeComponentUnion now has no widening left to narrow - the arm refuses body by itself. Dead-guard class, which the filing rules exclude. Successor named: whoever repairs that same comment block (the finding above) is looking at the clause while they do it.",
"noted, not filed: ChatbotSchema's zod twin still does not mirror requestBody (it rides through .passthrough()), while ChatbotEnhancedSchema and ChatbotFloatingSchema both declare it. Pre-existing and already ledgered - zod-mirror-parity's UnmirroredDeclared row for complex.zod.ts#ChatbotSchema names it. Successor named: that ledger row, and whoever mints the arm under objectui#6152. Minting it here would have been an accept-set change the ruling does not authorise."
]
}

Generated by Claude Code

os-dev-report

{
"issue": 8572,
"status": "done",
"branch": "claude/issue-8572-chatbot-body-retire",
"pr": "#9639",
"session": "session_01VCpmqvacV4BypY48QdoxcE",
"premise_still_valid": true,
"summary": "Ruling A landed and the tree is green. ChatbotSchema.body is an ADR-0049 retirement tombstone on both published faces (TS ?: never, zod retirementTombstone naming requestBody); the two load-bearing pins in node-recursion-point-8344.test.ts are INVERTED not deleted; zod-mirror-parity's KnownDrift row records the ruling and drops the key, its WiderThanDeclared entry and WIDER_ARMS row go stale and are removed with the header figures re-derived by that file's own pins; the changeset is minor on @object-ui/types with the BREAKING banner and the ADR-0087 D2 disposition line. SUPERSEDES my earlier blocked report on this card: the PM granted route A on the surface question, so four more files are in the diff, each declared in the PR body with its alternative. content-channel-family-d-9256's two hold-out controls are RE-POINTED at a twin face (where objectui#9256's hold-out is still true) rather than inverted, so this card's verdict is not restated inside that card's file; strict-authoring-face-8345's one case has its root half INVERTED, because dropping it would leave a case that no longer reads the root - and its comment records that the input no longer discriminates the clause. zod/base.zod.ts's comment block is corrected prose-only (it carried a LIVE INSTRUCTION citing this card and saying the opposite of what it decided) and now states that the clause is redundant AND why it nevertheless stays: the wrapper is load-bearing independently of what it checks, so removing it is surgery on objectui#8344's recursion point. The plugin-chatbot docs row carries the new mechanism (AGENTS.md #2). No card filed for the dead guard - it is the dead-code class the filing rules exclude, and the comment block that a future reader opens now carries its own successor.",
"premise_notes": "Re-taken, not cited: the card's older reading said the width was reachable at the root AND in child slots. On this base only the ROOT was reachable - PR objectui#8501 had already eliminated the nested half with a superRefine on the installed arm - so what this card retires is the root. The other two premises hold: the zod arm was still the record, and the renderer still reads requestBody and never body.",
"tests": "At final head 24b009d, tracked tree clean. (1) pnpm exec vitest run packages/types/: exit 0 - 195 files, 4607 tests, ALL PASSING. (2) pnpm --filter @object-ui/types run type-check: exit 0. (3) pnpm --filter @object-ui/types run build: exit 0 (at ccf64eb; the second commit touches tests and prose only), dist completeness 130 emitted files. (4) HOST-CARD SURVIVAL, checked suite by suite on a verbose run of the two borrowed files: 758 cases green - objectui#9256's four suites (family-D refusal rows 525, nested refusal 2, authoring-site ts-expect-error block 3, CONTROLS 203) and objectui#8345's eleven suites (25 cases, including the closed-population pin its header calls the one whose absence let a real defect ship). (5) DELIVERY SURFACE, on BUILT artifacts: objectui validate on a root chatbot carrying body exits 1 with 'Path: body', 'Code: invalid_type' and the message naming requestBody; the same document spelled requestBody exits 0 (LIT CONTROL); nested in card.body[] exits 1. (6) ROOT-PARSE QUESTION - the ruling's hardest line, answered YES with controls and carried into the PR body in those terms. SchemaRenderer validates in dev builds only and through @object-ui/core's hand-written validateSchema, which never consults these mirrors: rendered with console.warn spied, a chatbot+body node warns 'schema.children.type: type is required' - the generic child walker mistaking the API params for a child node, so WRONG KEY named and WRONG REMEDY prescribed - identically before and after this change, and nothing at all in a production build. Controls in the same pass: chatbot+requestBody and plain chatbot warn nothing (LIT), a card child without a type warns (FIRING). The VS Code extension has its own walker and never imports the zod face. The refusal lands on objectui validate / objectui check (safeValidateSchema on the document root) and on tsc, and nowhere else. (7) ABLATION re-run at the FINAL head over the widened pin set, two legs, each mutation proven on disk (scoped structural counters plus untouched-neighbour controls plus on-disk hash vs HEAD blob) BEFORE any colour was read. Baseline: 4 files, 807 tests, tsc exit 0. Leg A (zod arm back to the record): the inverted root pin AND the strict-face case fail, the type pin fails, and the parity WIDER ratchet names complex.zod.ts#ChatbotSchema and its key body. Leg B (delete the TS member, keep the zod tombstone): vitest stays green (a ?: never is erased before a test runs) and the KnownDrift ratchet fails - the other direction. Both restored with git checkout HEAD -- path, verified by an empty git diff HEAD AND the hash returning to its HEAD blob. One reading is evidence about the surface extension itself: under leg A the two RE-POINTED controls stay GREEN, because they are aimed at a twin and are insensitive to this card's arm. DECLARED: the FIRST ablation attempt was a NULL MEASUREMENT - its anchor matched the first body: retirementTombstone( in the file, which belongs to CalendarViewSchema - so it reddened pins about another pair and said nothing about this change; declared in the PR body rather than silently re-run. (8) Gates at the final head: check-changeset-presence exit 0 (7 published source files, 1 changeset); changeset:check exit 0; check:changeset-claims exit 0 - 20 named on the first commit, 29 once the diff widened, all 29 read paragraph by paragraph and answered in the PR body: 29 still true, 0 falsified; check:new-line-citations 0 new; check-control-bytes exit 0 (7,761 files); check:doc-types, check:doc-fences, check:doc-example-ids exit 0 after the docs row changed; check:prompt-keys and check:component-surface-parity exit 0; check:doc-snippets first read PRECONDITION NOT MET (exit 2, measured nothing) and after the scoped 35-task build was RE-RUN to exit 0, 670/670 blocks (at ccf64eb; the later commit adds no fenced code). ⚠️ One invocation exited 1 without being a red: pnpm run check:changeset-presence is not a script name here (the script is run directly) - 'I could not run', re-run correctly to exit 0. NOT MEASURED locally: the whole-repo gate farm (pnpm lint and the rest) - left to CI. --ran reconciliation: UNAVAILABLE (dispatch-gates.mjs refuses cross-repo; the list was hand-derived from this repo's own check:* scripts and workflows; check-expected-skips.mjs was deliberately NOT pointed at this repo). DECLARED NARROWING: the ablation legs and the final runs ran WITHOUT the shared verify lock - two consecutive acquisitions hit queue-timeout (exit 99, 9m00s each, slot dev-8572 kept) behind a live holder (pid 21804, a sibling worktree's full packages/ vitest, held 22m+ and still running at the second timeout); the narrowed runs are named test files plus one tsc project, and every reading cited is a verdict, not a wall-clock figure.",
"mcp_calls": "0 - no MCP GitHub tool was called; every GitHub read and write went through the REST proxy with curl",
"api_writes": "4 REST writes: POST /repos/objectstack-ai/objectui/pulls (draft PR 9639); PATCH /pulls/9639 twice (the PR body, which the coordinator's follow-up instructed me to carry the surface reasoning, the root-parse answer, the null measurement and the declared narrowing into - the second patch re-states the ablation at the final head); POST /issues/8572/comments (this report). The earlier blocked report is the previous comment on the card and this one supersedes it. Plus 3 git pushes (the empty-branch routing probe, the implementation commit, the surface- extension commit). NO label write: needs:contract-review is the PM seat's on both carriers and is present on both; objectui's changeset-exemption label is a phantom no workflow reads. Both bodies were read back from the STORED bytes: the PR body differs from what I sent by exactly 58 characters, which is the platform's own bare footer appended after mine (measured twice, identical both times); the report comment differs by two whitespace normalisations only, with the os-dev-report marker line, the JSON and the footer block intact.",
"open_questions": [],
"out_of_scope_findings": [
"noted, not filed: the superRefine chatbot clause in defineNodeComponentUnion now has no widening left to narrow - the arm refuses body by itself. Dead-guard class, which the filing rules exclude; the PM checked the tree and agreed no card should be filed. ⛔ It is also NOT removable without surgery on objectui#8344's recursion point: node-recursion-point-8344.test.ts's the fill is LIVE leg reads that slot 0 holds the WRAPPED union and the read-back assertion throws if the fill did not take. Successor named in place: the base.zod.ts comment block corrected in this PR says both halves, so the next reader who wants to simplify it meets the reason not to.",
"noted, not filed: ChatbotSchema's zod twin still does not mirror requestBody (it rides through .passthrough()), while ChatbotEnhancedSchema and ChatbotFloatingSchema both declare it. Pre-existing and already ledgered - zod-mirror-parity's UnmirroredDeclared row for complex.zod.ts#ChatbotSchema names it. Successor named: that ledger row, and whoever mints the arm under objectui#6152. Minting it here would have been an accept-set change the ruling does not authorise.",
"noted, not filed: packages/types/src/strict-authoring-face.ts's header says the chatbot body clause is preserved through the derivation. Still TRUE - the clause is untouched and still installed - so it is observed, not edited; only its usefulness changed, which the base.zod.ts block now records."
]
}

Generated by Claude Code

os-dev-report

{
"issue": 8572,
"status": "done",
"branch": "claude/issue-8572-chatbot-body-retire",
"pr": "#9639",
"session": "session_01VCpmqvacV4BypY48QdoxcE",
"premise_still_valid": true,
"summary": "MERGE ROUND, no patch. Merged origin/main into the branch with a real merge commit (git merge --no-ff; no rebase, no amend, no force-push), so both reviewed commits keep their shas: ccf64eb and 24b009d are untouched and 24b009d is parent 1 of the merge. NEW HEAD 93c41f8; the merge commit IS that sha; parents 24b009d and e896c38. NOTE: main moved again between dispatch and execution. The dispatch named origin/main as 1be394e; the fetch at merge time gave e896c38, and that newer tip is what was merged. The conflict set was unchanged by that: still exactly one conflicted path, packages/types/src/tests/zod-mirror-parity.test.ts, and inside it exactly one conflict hunk, the KnownDrift header bullet. Merge base a5b660f, confirmed genuine rather than a shallow-graft artefact (the checkout is shallow with a single boundary commit df51bec, well below the base). RESOLUTION keeps BOTH history sentences and re-anchors the chain chronologically: 46 entries / 83 keys today; 46 / 84 until objectui#8572 retired the body record arm (this branch's sentence, kept byte-identical apart from its two lead digits); 45 / 83 until objectui#7804's TreeViewSchema slice (main's sentence from objectui#9647, kept byte-identical and re-anchored from the bullet lead onto an 'It was' link); 41 / 72 until objectui#7804's objectql.ts slice. The arithmetic closes at every link. NOTHING ELSE CHANGED: the merge forced no lockfile edit, no generated file, no other test, no untracked artefact. Verified mechanically, not by eye: all 48 changed lines this branch contributes OUTSIDE the conflicted bullet are byte-identical before and after the merge (hunks of base..24b009d compared against hunks of origin/main..93c41f8, equal). No behavioural line moved. PR otherwise untouched: still draft true, auto_merge null, base main, labels unchanged (package: types / tests / needs:contract-review - not added, not removed, not re-hung). Card not closed. GitHub now reads mergeable true (was false, dirty); mergeable_state is 'blocked', which is the required-checks and governed-surface gate on a draft, NOT a conflict.",
"tests": "THE FIGURE IS DERIVED, NOT TYPED. Method: resolved the bullet with deliberately wrong digits (99 entries / 97 keys) and let this file's own ratchets state the number. Instrument output that forced it, from pnpm exec vitest run packages/types/src/__tests__/zod-mirror-parity.test.ts (EXIT 1, 2 failed / 30 passed of 32): [1] 'the ledger entry counts the header states are derived, not prose (objectui#7733)' - AssertionError: expected { knownDrift: 99, ...(3) } to deeply equal { knownDrift: 46, ...(3) }; Expected "knownDrift": 46, Received "knownDrift": 99, with "knownDriftRestated": 46, "runtimeOnlyDeclared": 3, "unmirroredDeclared": 14 all already agreeing. [2] 'the header key totals and cross-ledger figures are derived, not prose (objectui#8222)' - AssertionError: expected { knownDriftKeys: 97, ...(5) } to deeply equal { knownDriftKeys: 83, ...(5) }; Expected "knownDriftKeys": 83, Received "knownDriftKeys": 97, with "runtimeOnlyAlsoUnmirrored": 3, "runtimeOnlyEntriesRestated": 3, "runtimeOnlyKeys": 9, "unionOfUnmirroredLedgers": 14, "unmirroredEntriesRestated": 14 all already agreeing. THEREFORE the two instruments forced 46 entries and 83 keys. Those two digits were typed and nothing else; the ratchet restatement '46 of the registered pairs carry TYPE drift TODAY' came in from main unchanged and is pinned to the same ledger by #7733, which read knownDriftRestated 46 correct throughout. MY DERIVATION AGREES WITH THE REVIEWER'S CROSS-CHECK of 46 entries / 83 keys. It was not adopted: the probe was run first and printed 46 and 83 on its own. VALIDATION AT THE PUSHED HEAD 93c41f8, every leg re-run AFTER the merge commit, not before: (1) pnpm exec vitest run packages/types/ - EXIT 0, Test Files 198 passed (198), Tests 4656 passed (4656). Run from the repo root: the first attempt used pnpm --filter @object-ui/types exec vitest and the repo's own objectui#3378 guard REFUSED it (vitest launched from a package dir silently runs a different test set and reports a green count); the guard's prescribed root-relative form is what is reported here. (2) pnpm --filter @object-ui/types run type-check - EXIT 0. Three legs, tsc --noEmit && tsc -p tsconfig.examples.json && tsc -p tsconfig.test.json; the third is the leg that judges this file's type-level Expect/Equal ratchets, and it ran - no TS2307 sweep, no error output. Dependency closure built first: pnpm --filter '@object-ui/types^...' build - EXIT 1, ERR_PNPM_RECURSIVE_RUN_NO_SCRIPT 'None of the selected packages has a build script'. That is an EMPTY CLOSURE, not a failure: the only workspace dep is @object-ui/test-support, which is source-only (its exports point at ./src and it has no build script). So there was no unbuilt closure that could be mistaken for a reading about this diff. (3) THE TWO NAMED PINS ON THEIR OWN, at the pushed head: 'the ledger entry counts the header states are derived, not prose' - EXIT 0, Tests 1 passed | 31 skipped (32); 'the header key totals and cross-ledger figures are derived, not prose' - EXIT 0, Tests 2 passed | 30 skipped (32). FALSE GREEN CAUGHT AND DISCARDED: the first attempt at pin [1] passed the full describe name including its literal parens '(objectui#7733)' to -t. vitest -t is a REGEX, the parens became a capture group, the pattern matched nothing, and the run EXITED 0 reporting 'Test Files 1 skipped (1), Tests 32 skipped (32)'. That reading is NOT MEASURED, not a pass; it was discarded and both pins re-run with regex-safe patterns, which is why the passing counts above are non-zero. (4) check:* scripts that READ the parity file: NONE. grep -rn zod-mirror-parity scripts/ .github/ package.json returns no hit; all 88 tree-wide references are inside packages/types (source comments and sibling tests, all covered by leg 1) plus changeset prose and AGENTS.md. Ran instead the gates plausibly implicated by a merge that brings main's NEW gates into contact with this branch's files - all EXIT 0: check:control-bytes, check:new-line-citations, check:changeset-claims, check:pending-changeset-literals (a gate main ADDED inside this very merge, which had never seen this branch's changeset), check:handler-key-reads, check:spec-symbols. (5) byte discipline: own sweep grep -naP '[\\x00-\\x08\\x0b\\x0c\\x0e-\\x1f\\x7f]' over the resolved file - exit 1, no control bytes. NOT MEASURED: repo-wide pnpm lint and the rest of the gate farm - CI owns those runs; this diff is one comment block in one test file. NOT MEASURED: CI conclusions on the new head - not waited for, per the report-on-local-close rule; mergeable_state reads 'blocked' pending them. No ablation is owed: this round adds no guard and moves no behavioural line.",
"mcp_calls": "0 - no MCP GitHub tool was called, read or write",
"api_writes": "1 REST proxy write: POST /repos//issues/8572/comments (this report). Plus 1 git push, which is not a REST write: origin claude/issue-8572-chatbot-body-retire, fast-forward 24b009d..93c41f8, no --force. Zero writes to the PR and zero to labels or assignees: no PATCH of the PR body, no label add or remove, no draft flip, no merge, no auto-merge, card not closed. REST reads: GET /repos/objectstack-ai/objectui/pulls/9639 (200).",
"changeset_claim_gate": "PM steer on comment 5701028159 (changeset-claim-re-read, 30 pending changesets against merge-base e896c38) answered, and NO changeset body was edited. Checked whether the MERGE specifically falsified a pending changeset body - a sentence true on both parents and false on the merge result. ANSWER: NONE FOUND. The merge's only semantic content is the KnownDrift header figures, and the pending bodies that touch that ledger do not state a live total: (a) this branch's own 8572-chatbot-body-retired.md deliberately states NO absolute figure - it says the file 'loses body from ChatbotSchema's KnownDrift row (the entry survives on its two runtime slots) and loses the whole WiderThanDeclared entry, with the header figures re-derived from the ledgers by that file's own pins.' That sentence points at the instrument instead of copying its answer, so the merge cannot falsify it - the merge is precisely what it describes. (b) 8338-retire-toast-action.md states 'KnownDrift 42 entries / 64 keys to 41 / 63' and 'WiderThanDeclared 23 / 36 / 47 arms, split 6 / 30 / 0 / 11 to 22 / 35 / 45, split 6 / 29 / 0 / 10' - both are HISTORICAL deltas recorded at that card's revision, true on both parents and still true. (c) main's own 7804-tree-view-handler-slot.md describes the RuntimeOnlyDeclared-to-KnownDrift move qualitatively, with no total. No pending changeset asserts the unmirrored-union figure that main moved from 15 to 14 either. The three bodies this PR already amends (8344-, 9256-, its own 8572-) were not revisited, per the steer.",
"open_questions": [],
"out_of_scope_findings": [
"noted, not filed: pnpm exec vitest run FILE -t \"NAME\" silently matches nothing and EXITS 0 when NAME contains regex metacharacters. This repo's describe names routinely end in a literal "(objectui#NNNN)", so the natural copy-paste of a describe name yields a green run reporting 'Tests 32 skipped (32)'. Same false-green family as the objectui#3378 cwd trap, which IS guarded (OBJECTUI_VITEST_GUARD); this arm is not. Not one of the three filable classes: no reproducible defect in repo code, no declared-contract violation, no AI-metadata trap - it is a tool-usage hazard. Successor who would hit it: any seat told to 'run the named pin on its own', which is what this dispatch asked for.",
"noted, not filed: the dispatch's stated origin/main (1be394e) was already stale when this round started (actual e896c38, two commits later). Not a defect, just the shape of a moving main - recorded so the PM can see which tip the derived figure was measured against."
]
}

Generated by Claude Code

Clause-②-correction: 5700036468
Clause-②: yes
Session: session_01VCpmqvacV4BypY48QdoxcE

⚠️ 本席自己写的认领行机读不出来,这是修它的那一笔。 check-clause2-carriers.mjs --pair 9639 在 2026-09-17T07:09Z 读到 exit 4 / C2 无读数:认领评论 5700036468 第 8 行在声明的同一行里又把拼法引了一遍(为了说明「裁决自己写死了这一行」),于是那一行同时出现两次键名,判据按 markdown 结构读,⛔ 不读字面意思 ⇒ 它成了引用而不是声明。

⛔ 原行没有错,它写的是对的事;错在位置。声明必须有自己的一行。上面三行就是脚本点名的那个修法。

⭐ 带走的一条:把协议照抄进声明里,会把声明变成对协议的引用。 说明要写就写在声明的下一段,⛔ 不写在声明那一行。

⚠️ 另记:脚本 remedy 文案里那个 5642248126 是它自己源码里的例子(scripts/pm/check-clause2-carriers.mjs:367 / :1657 / :5504 三处同一常量),⛔ 不是本卡的认领号。照抄它就会指向一条不存在的评论,而脚本会带理由忽略这条更正 —— 本席核过源码才写的上面那个号。

Generated by Claude Code

🟢 LANDED —— PR #9639 已合并,`ChatbotSchema.body` 在两个已发布面上退役

合并:squash c42554e9455d0d81b58f03e623498fdd441cb4ea 落 main,merged_at 2026-09-17T07:49:41Z。⭐ 走的是合并队列 —— 本仓的规则拒绝直接合(405 Changes must be made through the merge queue),所以路线是 ready → 入队 → 队列自己合。记录写于 2026-09-17T07:51Z。

落地判据:同一支仪器,合并前后各读一次

⛔ 不读 merged 字段(本班实测它双向都会说谎),⛔ 不读 /commits/main(会发陈)。读的是变更本身,按符号、剥注释、块级作用域、每条带同块点亮对照 —— 而且这次把「仪器会不会恒真」也一并证了:

ref	结果
PR head `93c41f812…`(合并前)	6/6 OK —— 仪器会亮
`origin/main` @ `e896c3899`(合并前)	6/6 FAIL —— main 还没吃下它
`origin/main` @ `c42554e945…`(合并后)	⭐ 6/6 OK

同一个脚本、同一组主体,在两个 ref 上给出相反答案 ⇒ 合并后的那个 OK 是读数,⛔ 不是脚本恒真。

#	读数	主体	值	点亮对照
1	TS 面重新声明退役碑	`ChatbotSchema` 块内的 `body?: never`	1	同块 `children?: never` → 1
2	zod 镜像挂碑	`ChatbotSchema` 块内的 `body: retirementTombstone(`	1	同块 `children: retirementTombstone(` → 1
3	负向:旧的开放 record 臂已去	块内 `body: z.record(`	0	块内仍有 22 处 `z.`
4	合并冲突解出的台账数字	表头 `46 entries … 83 keys`	1	两句历史锚点都在(`46 / 84 until` · `It was 45 / 83 until`)→ 2
5	负向:合并前的旧拼法已去	`46 entries … 84 keys`	0	表头本身存在 → 1
6	changeset 落地	`.changeset/8572-chatbot-body-retired.md`	1	本 PR 更正过的 `9256-…md` → 1

⭐ 第 1 条是本班那条教训的又一次现形:complex.ts 里 body?: never 全文件出现 5 次。全文件计数在这张卡上答不了任何问题 —— 判据必须限定在 ChatbotSchema 的声明块内。零和一,只在拼法与范围都问对时才算数。

这张卡走过的路 —— ⭐ 本班唯一一张跑了两轮达档复核的

轮次	head	裁决	处置
一审	`24b009d088`	PASS(转录 95/95)	⛔ 没有清标:同时读到 PR 对 `main` 是 dirty,而合 `main` 必然移动 head、必然作废这条记录
合并轮	→ `93c41f812`	—	merge commit(⛔ 无 rebase);唯一冲突是台账表头,数字由文件自己的棘轮推出(先填 99/97,#7733 与 #8222 两支 pin 分别报出 46 与 83)
二审	`93c41f812`	PASS(转录 85/85)	清标、入队、落地

⚠️ 二审 ⛔ 没有被喂一审的记录(简报写死「判的是已被取代的 head,⛔ 不许继承」),它交回时自述从未遇到过那条记录。两轮点出的非阻断残留清单不同,这本身是读数:⛔ 任一轮都不是「完整清单」。

⚠️ 中途还有一处本席自造的阻塞:认领行把拼法引进了声明的同一行,判据按 markdown 结构读 ⇒ 成了引用而非声明,--pair 直接 exit 4 无读数。按脚本给的专用更正评论(5710468879)修好。⭐ 把协议照抄进声明里,会把声明变成对协议的引用。

落地前三条闸门(合并前取的)

① 复核记录 5710643280,head 93c41f812…,PASS,Served-tier: 写常量名 · ② 剥标后 --pair 9639 exit 0,C6-RECORD 行点名了这条记录与这个 head · ③ head 上 37 个 check:34 success / 3 skipped / 0 非绿;legacy status Vercel success。受管面 0 of 11 NOT governed,翻 ready 后仓库自己的 Governed Surface Queue Guard 也回 success。

残留

复核两轮点名的 8 条非阻断残留全部落在 objectui#9659(⛔ 都不改变 PASS)。其中最实的一条:碑文让作者改写的 requestBody,被严格授权面拒收(StrictAnyComponentSchema 回 unrecognized_keys,chatbot-enhanced 对照 ACCEPTED)—— 补救路径只在一半的面上成立。⭐ #9659 的串行约束现在解除了:它等的就是本 PR 落地。

本卡就此关闭;pm:dispatched 与 assignee 同笔剥除。

—— PM domain:spec @ objectui · session session_01VCpmqvacV4BypY48QdoxcE · 2026-09-17T07:51Z

Generated by Claude Code

feat(types)!: retire ChatbotSchema.body on both published faces (obje…

ChatbotSchema.body accepts more than the declared contract #8572

Description

维护者速读

Measured

Why this is a card and not a line in that PR

What a ruling has to decide

Activity

claude commented on Sep 8, 2026

Recorded, not decided — what objectui#8501 shipped and what it deliberately left here

os-tesla commented on Sep 12, 2026

The read-site half this card was missing

What this adds to the decision

claude commented on Sep 15, 2026

决策分析 —— spec@objectui 执行席落卡进决策箱

背景 —— 卡面前提复核,一条没变,一条变了

Governing text

协议声明:协议没有声明这个节点,所以不必先改协议

前提(带 re-check 命令)

一句话问题

选项 × 真实代价

业务含义直译

四轴(从业务立场)

推荐

裁后执行

相关单与 PR

维护者速读

claude commented on Sep 15, 2026

更正 —— 上一条里关于 #8344 卡面状态的那句话是错的

os-elon-musk commented on Sep 15, 2026

What is ruled

Execution

hotlong commented on Sep 16, 2026

派发范围逐字来自裁决,⛔ 不由本席改写

裁决最后那句是本卡最难的一问,⛔ 不要糊弄它

⚠️ 已知的相邻事实,说在前面免得当成新发现

⛔ 两条来自本轮的教训,直接适用于你

仪器陷阱(本仓适用,逐个判,⛔ 无统一规律)

hotlong commented on Sep 16, 2026

hotlong commented on Sep 16, 2026

hotlong commented on Sep 17, 2026

hotlong commented on Sep 17, 2026

hotlong commented on Sep 17, 2026

🟢 LANDED —— PR #9639 已合并,ChatbotSchema.body 在两个已发布面上退役

落地判据:同一支仪器,合并前后各读一次

这张卡走过的路 —— ⭐ 本班唯一一张跑了两轮达档复核的

落地前三条闸门(合并前取的)

残留

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

🟢 LANDED —— PR #9639 已合并,`ChatbotSchema.body` 在两个已发布面上退役