[Decision] spec(ui): the five ComponentPropsMap members #21464's last stage stopped on — object-metric `drillDown.report`, object-form `customFields`, form `sections`, object-timeline `items`, `action:group` / `action:menu` members

[objectstack-fleet]

Ruled: 5979239990 · forks 4 and 5 letters B, A (batch #277) — all five forks ruled: 1 B, 2 B, 3 B, 4 B, 5 A · 2026-10-04T10:58Z
Ruled: 5978663135 · forks 1, 2, 3 letters B, B, B (batch #276); forks 4 and 5 pending batch #277 · 2026-10-04T09:50Z

Filing gate: ② decisions only the maintainer can make. Each fork picks the published shape of a page-component member, and no existing ruling decides it. Filed by the director seat, summon #32 (session_016tKoy8NJa35Yih1FdzrVmn), holder of #21464's S-objectui-held claim 5976829576, from that stage's dev report 5977481177 (open_questions 1–5). The at-tier contract review of record on #21699 (5977620713, PASS) judges each a genuine fork, not owed work skipped, and takes no position among the options. PR #21699 records all five in the enumeration pin's new fork stage (component-props-unknown-members.pin.test.ts), each with its shapes. The five are folded into one card under the filing quota: they belong to one card, and the ruled letters ship as one more #21464 stage. ⛔ Not a claim.

Reader who acts: the maintainer, one letter per fork (five letters). Then the director seat dispatches one #21464 stage carrying the ruled shapes. Fork 1's B waits on #21702.

维护者速读

• 背景： spec(ui): the ComponentPropsMap rows still type renderer-read members as z.unknown() — navigation on object-map / object-gantt / object-tree and conditionalFormatting on object-kanban accept 42 — the family close-out after #21445 #21464 把页面组件里「平台没声明、写错也不报错」的属性逐个补上类型。最后一段（PR feat(spec)!: object-gantt markers, object-timeline mapping and both forms' fields take the shape each block reads; five objectui-held contracts reported as forks (#21464, S-objectui-held) #21699）补了 4 个，另外 5 个各有两种以上都说得通的写法，现有裁决都没覆盖，所以按「停阀」规则停下来，交给你选。
• 五个分叉，一次回五个字母：
  1. 指标卡下钻的报表 drillDown.report：推荐 B，先修报表契约（[finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702），再直接引用报表契约。
  2. 表单的自定义字段 customFields：推荐 B，平台写一份封闭的、驼峰命名的字段定义，只收表单真正会画的属性。
  3. 表单分区 sections：推荐 B，页面块自己的分区形状，不动已存储的表单视图。依赖第 2 项。
  4. 时间线条目 items：推荐 B，两种条目都封闭，content 先当不透明成员，按 variant 校验条目种类。
  5. 操作组和操作菜单的成员：推荐 A，沿用按钮的已测词汇，未决的键先拒绝。outcomeMessages 对四个操作块一起决定，A 下四块都暂不声明。
• 你要做的： 回复形如 1B 2B 3B 4B 5A。

Measured (the dev's census, read as claims; the contract review on #21699 judges them)

• Corpora: objectstack 7d0781482d, objectui pin ab1879721595 and main 94985a92ba, hotcrm 4054ec2680, cloud b2d7a7f6f8. The instrument is a TypeScript-AST walk over code and fenced docs. It reproduces stage 5's drillDown population exactly.
• Fork 1, drillDown.report: 3 objectui values. 1 is drawn and parses ReportSchema. 2 are it.each probes, not drawn, both refused by ReportSchema. objectstack 0, hotcrm 0, cloud 0.
• Fork 2, customFields: 27 values, 31 entries, all objectui. Each is { name, label, type } plus a few draw keys. 0 snake_case grid keys. 2 keys outside objectui's 45 (group, defaultValue), in two tests.
• Fork 3, sections, inline "shape 3" entry: written 7 times, all in plugin-form's README wizard and its submit-target tests. Form-view writers author none.
• Fork 4, timeline items: 14 values, all objectui tests (8 feed entries, 5 gantt rows), 0 with content. objectstack 1 (a spec test), hotcrm 0, cloud 0.
• Fork 5, action group/menu members: 21 static members. They use only name, label, type, locations, target, bodyExtra, bodyShape and objectName, plus one host auto-trigger test. None uses an undecided key.

Governing text

• ReportSchema's refinement comment and content/docs/ui/reports.mdx: "each block dataset-bound". This is not enforced; see [finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702.
• AGENTS.md naming: "TS config keys → camelCase".
• objectui types/src/form.ts (about :1770): FormField, 45 members plus an index signature (about :1906).
• objectui#11550 kept the inline section entry, ObjectFormSection.fields: (string | FormField)[].
• objectui#6356 ruled the timeline arms, TimelineFeedItem / TimelineGanttItem, paired by variant.
• component.zod.ts action rows, on feat(spec)!: object-gantt markers, object-timeline mapping and both forms' fields take the shape each block reads; five objectui-held contracts reported as forks (#21464, S-objectui-held) #21699's head: outcomeMessages is forwarded since objectui#11344 and not declared, "for action:button's reason". objectui leaves the input unpublished until a row declares it. endpoint has been refused since finding(spec): action:button / action:icon accept endpoint, the key ActionSchema refuses with the prescription endpoint → target: one concept, two verdicts in one release #21005.
• Prior rulings read: drillDown report, customFields, FormField spec, form section inline field, timeline items variant, UIActionSchema, outcomeMessages → 0 hits on an open objectstack card. MCP search_issues, scoped, for 「outcomeMessages action:button action block contract decision」 → 0. For 「object-form customFields sections inline runtime form field spec declaration」 → 2, both closed and unrelated (Studio field designer is hand-coded, not metadata-driven — bring design-time authoring into the protocol (fieldForm / metadata-form-registry) #3257, finding: a section-less public form accepts ANY non-anchor field from an anonymous visitor (the write-side twin of #6601) #6920). ADR: none. Thread: spec(ui): the ComponentPropsMap rows still type renderer-read members as z.unknown() — navigation on object-map / object-gantt / object-tree and conditionalFormatting on object-kanban accept 42 — the family close-out after #21445 #21464.

---

分叉 1：指标卡下钻的报表 object-metric drillDown.report

一句话问题： 指标卡点开下钻时，能挂一张报表；平台该按「报表」的正式契约来校验它，还是另写一套只给下钻用的规则？

选项	做什么	客户可感知的后果
A 直接引用现有报表契约	现在就按 ReportSchema 校验	「各分块都没绑数据集」的联合报表能通过校验，打开下钻却只列出记录，不画报表，没有任何提示（实测）
B 先修报表契约，再引用	先做 #21702：联合报表的每个分块必须绑数据集。再按 ReportSchema 校验	校验通过就一定画得出来。实测唯一的联合报表（showcase 的 TaskOverviewReport）每块都绑了数据集，不受影响
C 下钻专用形状	另写一个只看下钻抽屉两条判断的形状	下钻能用，但平台多出第二种「报表」方言，和报表元数据的校验各说各话

业务含义直译：

• A 等于收货单上写着「每箱都要有货」，验货却只数箱子。
• B 等于先把验货规则改成真的开箱看，然后下钻和报表共用同一个验货员。
• C 等于给下钻单独雇一个验货员，标准和仓库的不一样。

四轴（业务立场）：

• 实际业务需求： 实测 1 个真实画出的写法，A 和 B 下都能通过。B 的收紧不拒绝任何实测写法。
• 项目长远合理性： B 只留一份报表契约，在生产端修。A 把「收了却画不出」固化成契约。C 是第二种报表方言。
• 防 AI 犯错： 出错时谁看到什么？A 下，AI 写出没绑数据集的联合报表能通过校验，用户打开下钻只看到记录列表。B 下，objectstack validate 当场报错。
• 创业阶段不扩散： B 不加新 schema，只在已有 schema 上加一条校验。C 加一个 schema。

os-decision-facets

• ① 项目长远合理性：B 一份报表契约、在生产端修；A 固化「收了画不出」，C 增一种报表方言。
• ② 实际业务拉动：1 个实测画出的写法，B 下照样通过；第二种形状零拉动。偏 B。
• ③ 防 AI 犯错：B 在校验时响亮拒绝；A 静默列记录。偏 B。
• ④ 创业阶段不扩散：B 只加一条校验；C 加 schema。偏 B。

推荐：B。 两年后的样子：下钻、仪表盘、报表页都引用同一份报表契约，「能通过校验」就等于「画得出来」。参照 Salesforce：仪表盘组件引用的是同一份报表定义，不另立格式。
回退：A。 先引用现有契约，未绑数据集的漏洞留给 #21702 单独修。
自检： 只看①选 B；②③④ 是否翻转：否。
置信缺口： hotcrm 和 cloud 的联合报表还没普查，#21702 的认领会先普查。

裁后执行：

• B： [finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702 先落地，按收紧规范发版。之后 spec(ui): the ComponentPropsMap rows still type renderer-read members as z.unknown() — navigation on object-map / object-gantt / object-tree and conditionalFormatting on object-kanban accept 42 — the family close-out after #21445 #21464 的下一段把 report 写成对 ReportSchema 的引用。
• A： 下一段直接引用。[finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702 照常排队。
• C： 下一段写一个下钻专用形状。

---

分叉 2：表单的自定义字段 object-form customFields

一句话问题： 不绑对象的表单可以直接写一组字段。平台该怎么声明「一个字段」？

选项	做什么	客户可感知的后果
A 照搬 objectui 的 FormField	45 个成员加任意键，包括 8 个蛇形命名的表格控件键	什么都能通过，写错键名也不报错；spec 里第一次出现蛇形命名的配置键
B 平台写一份封闭的字段定义	只收表单真正会画的成员，驼峰命名。表格控件的 8 个键等 objectui 改成驼峰再收	实测 27 个写法全部通过；写错或没声明的键当场报错
C 复用表单视图的 FormFieldSchema	把按对象字段覆盖的那套定义改成按 name 索引	一个词汇两用，但「覆盖已有字段」和「定义新字段」被混成一件事，objectui 的合并逻辑也得改键

业务含义直译：

• A 等于报名表留了「其他：____」一栏，写什么都收。
• B 等于报名表只印真正会处理的栏目，填错栏会被退回。
• C 等于拿「修改已有档案」的表格，去登记新人。

四轴（业务立场）：

• 实际业务需求： 实测 27 个值、31 个条目，都是 { name, label, type } 加少量绘制键。零个蛇形表格键。开放字段袋和表格键都没有拉动。
• 项目长远合理性： 封闭、驼峰、声明过的契约才能长期维护。A 把蛇形键写进 spec，还留着「任意键」这个静默丢弃的漏洞。C 混淆了两件不同的事。
• 防 AI 犯错： B 下，AI 拼错或编造的键当场报错。A 下静默丢弃。
• 创业阶段不扩散： B 只声明有写法的成员，表格键等 objectui 改名再进。代价是在 spec 里写一份四十多个成员的定义，一次写完。

os-decision-facets

• ① 项目长远合理性：B 封闭、驼峰、可维护；A 引入蛇形键与任意键，C 混淆覆盖与定义。
• ② 实际业务拉动：27 个实测写法 B 下全过；开放袋与表格键零拉动。偏 B。
• ③ 防 AI 犯错：B 响亮拒绝拼错的键；A 静默丢弃。偏 B。
• ④ 创业阶段不扩散：B 只声明有写法的成员；A 声明 45 个加任意键。偏 B。

推荐：B。 两年后的样子：无对象表单的字段和对象字段一样有封闭契约，Studio 能据此做字段面板。参照 Salesforce Screen Flow：屏幕组件有固定、具名的属性表。
回退：C。
自检： 只看①选 B；②③④ 是否翻转：否。
置信缺口： 私有客户可能用了表格控件的蛇形键。B 下这些写法会被拒绝，需要等 objectui 改名后再收进来。

裁后执行：

• B： 下一段在 spec 里声明一份封闭的运行时表单字段，驼峰命名，按收紧规范发版。objectui 那边另开一张卡，把表格控件的 8 个键改成驼峰。
• A / C： 按所选形状写进下一段。

---

分叉 3：表单分区 object-form / object-master-detail-form sections

一句话问题： 表单分区里除了字段名，还可以直接内嵌一个完整的字段定义（objectui#11550 保留了这种写法）。平台的分区形状要不要连「已存储的表单视图」一起放宽？

选项	做什么	客户可感知的后果
A 引用表单视图的 FormSectionSchema，并放宽其字段条目	页面块和已存储的表单视图都接受内嵌字段定义	已存储的视图元数据多出第三种字段写法，这本身是另一份契约变更
B 页面块自己的分区形状	FormSectionSchema 的键，加上表单实际读取的三种条目，只收规范写法；表单视图不动	实测 7 个内嵌写法全部通过；视图元数据不受影响；已废弃的写法会被拒绝

业务含义直译：

• A 等于为了一个临时表单，改了公司档案柜的归档规则。
• B 等于临时表单用自己的格式，档案柜规则不变。

四轴（业务立场）：

• 实际业务需求： 内嵌写法实测 7 次，全在 plugin-form 的 README 向导和它的测试里。表单视图的写法零次。没有放宽视图的拉动。
• 项目长远合理性： 内嵌字段是页面块的能力。放进已存储的视图元数据，是一份单独的契约变更，分诊已说过那需要另裁。
• 防 AI 犯错： B 把分区封闭在页面块上，不教视图第三种写法。已废弃的拼法也会当场被拒，而不是原样交给渲染器。
• 创业阶段不扩散： B 不放宽任何其他契约。

os-decision-facets

• ① 项目长远合理性：B 把内嵌字段留在页面块；A 顺带改了存储视图的契约。
• ② 实际业务拉动：7 个内嵌写法都在 objectui 自己的向导与测试；视图端零拉动。偏 B。
• ③ 防 AI 犯错：B 封闭且拒绝废弃拼法；A 让视图多一种写法。偏 B。
• ④ 创业阶段不扩散：B 不放宽其他契约。偏 B。

推荐：B，排在分叉 2 之后。 内嵌条目引用分叉 2 选出的字段定义。两年后的样子：存储的视图只引用对象字段，临时表单的内嵌字段只在页面块上出现。
回退：A。
自检： 只看①选 B；②③④ 是否翻转：否。
置信缺口： 如果分叉 2 选 A，B 的内嵌条目也是开放的，封闭的好处就打折扣。

裁后执行：

• B： 与分叉 2 同一段落地，内嵌条目引用分叉 2 的字段定义。
• A： 放宽 FormSectionSchema，另开一张存储视图的契约卡，先于本段。

---

分叉 4：时间线条目 object-timeline items

一句话问题： 时间线有「动态流」和「甘特条」两种条目，由父组件的 variant 决定是哪种。平台要不要校验条目和 variant 对得上？动态条目里的子内容 content 怎么处理？

选项	做什么	客户可感知的后果
A 两种封闭，content 作为插槽，按 variant 配对	页面遍历器把 content 当作子组件位置去检查	子内容也会被检查；但为一个没人填的成员，改动每一次页面遍历
B 两种封闭，content 先不透明，按 variant 配对	校验条目与 variant 匹配；content 暂不深入检查	实测 14 个写法都通过；写错种类当场报错
C 两种封闭，content 不透明，不配对	条目只要符合任一种即可	动态流里混进甘特条也能通过，objectui 已裁定的配对规则被丢掉

业务含义直译：

• A 等于为一个还没人用的抽屉，重新设计整栋楼的巡检路线。
• B 等于先检查抽屉标签和柜子类型对得上，抽屉里面以后有人放东西再查。
• C 等于只要是抽屉就收，不管放进了哪个柜子。

四轴（业务立场）：

• 实际业务需求： 实测 14 个值都是 objectui 的测试，零个带 content。插槽机制零拉动。
• 项目长远合理性： 两种条目和配对规则是 objectui#6356 已裁定的契约，原样接过来。为没人填的成员新增插槽位置，会改变每一次页面遍历。
• 防 AI 犯错： B 和 A 都在种类混用时响亮报错。C 不报错。
• 创业阶段不扩散： B 不加插槽位置。等出现带 content 的写法，再把它升级成插槽。

os-decision-facets

• ① 项目长远合理性：B 原样承接 objectui#6356 的裁定；A 为零写法的成员改动页面遍历，C 丢掉配对。
• ② 实际业务拉动：14 个实测写法零个带 content；插槽零拉动。偏 B。
• ③ 防 AI 犯错：B、A 在种类混用时响亮报错；C 静默接受。偏 B。
• ④ 创业阶段不扩散：B 不加插槽位置。偏 B。

推荐：B。 两年后的样子：等有人真的往动态条目里放子内容，再把 content 升级成插槽。到那时有写法可测，形状是测出来的。
回退：A。
自检： 只看①选 B；②③④ 是否翻转：否。
置信缺口： 「content 暂不检查」意味着里面写错也不会报错，直到升级成插槽为止。今天这里零写法。

裁后执行：

• B： 下一段落地两种封闭条目，加上按 variant 的配对校验。content 在台账里记为不透明成员，写明原因。
• A / C： 按所选形状写进下一段。

---

分叉 5：操作组和操作菜单的成员 action:group / action:menu actions[]

一句话问题： 操作组和操作菜单里的每个成员，就是一个内联操作。平台按哪套词汇校验它？

选项	做什么	客户可感知的后果
A 沿用按钮的已测词汇	和 action:button 同一套键；未决的键（outcomeMessages、成员 className、properties.params）先拒绝，endpoint 按既有提示改写到 target	实测 21 个成员全部通过；写了未决键会被拒绝并提示
B A 加三个未决键	再声明 outcomeMessages、成员 className、properties: { params }	三个键都没有实测写法，声明后就是长期义务
C 直接引用对象元数据的 ActionSchema	按「已注册操作」的契约校验内联成员	这是另一种声明（必须有名字、按名字注册），本文件的分节说明已经排除了它

业务含义直译：

• A 等于菜单里的每一项，和单独的按钮用同一份说明书。
• B 等于为了还没人用的三个功能，先把它们写进说明书。
• C 等于把临时按钮当成正式注册的流程来审批。

四轴（业务立场）：

• 实际业务需求： 21 个实测成员只用 8 个键，零个用到未决键。B 没有拉动。
• 项目长远合理性： 内联操作只保留一套词汇，就是按钮那套实测出来的。C 混淆了两种声明。
• 防 AI 犯错： A 封闭，并带着按钮那几行的改写提示。
• 创业阶段不扩散： B 声明三个零写法的键。

os-decision-facets

• ① 项目长远合理性：A 一套内联操作词汇；C 混淆已注册操作与内联操作。
• ② 实际业务拉动：21 个实测成员零个用到未决键。偏 A。
• ③ 防 AI 犯错：A 封闭且带改写提示。偏 A。
• ④ 创业阶段不扩散：B 声明三个零写法的键。偏 A。

推荐：A。 outcomeMessages 是四个操作块（按钮、图标、组、菜单）共同的问题。A 下四块都暂不声明，等出现内联写法再一次性声明。两年后的样子：内联操作和按钮共用一套封闭词汇，新增键四块一起进。
回退：B。
自检： 只看①选 A；②③④ 是否翻转：否。
置信缺口： objectui#11344 已经在渲染端转发 outcomeMessages，并注明等 spec 的行声明后再公开输入。A 下这个公开会继续等。如果你希望现在就让四个操作块都接受 outcomeMessages，回 5A+，四块一起声明。

裁后执行：

• A： 下一段落地封闭成员形状，与 action:button 同一套键和提示。
• 5A+： 同上，另在四个操作块上一起声明 outcomeMessages，按 ActionSchema 引用。
• B / C： 按所选形状写进下一段。

---

Related

#21464 · PR #21699 · #21702 · objectui#11550 · objectui#6356 · objectui#11344 · #21005 · ADR-0021 · ADR-0087.

Dedupe words: drillDown.report, customFields, FormField, sections inline field, timeline items variant, action:group members, action:menu members, outcomeMessages, UIActionSchema. Searches are listed under Governing text.

---

Generated by Claude Code

[objectstack-fleet]

Ruling: batch #276 items 3–5 · forks 1, 2 and 3 · letters B, B, B · maintainer 「同意」 2026-10-04T09:48Z

Director seat, summon #32, session_016tKoy8NJa35Yih1FdzrVmn. Written as objectstack-fleet[bot] through the relay.

• How it was ruled. Batch 🔗 Broken links detected in documentation #276 was presented in the live director chat. Items 3, 4 and 5 were this card's forks 1, 2 and 3, each with the options this card lays out. The director recommended B for each. The maintainer answered the batch as presented: 「同意」.
• Forks 4 and 5 (timeline items, and the action:group / action:menu members) are not ruled here. They go to batch 🔗 Broken links detected in documentation #277, and this card keeps needs-user-decision until then.
• Freshness gate: this card has no comment since filing.

The rulings

Fork 1, object-metric drillDown.report: B.

• [finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702 lands first: ReportSchema's joined branch refuses a block with no dataset.
• Then report is typed by reference to ReportSchema, so that admitted implies drawn.
• Not taken: A, referencing the schema as it stands; C, a drill-report shape of its own.

Fork 2, object-form customFields: B.

• The spec declares a closed runtime form field, in camelCase, holding only the members the form draws.
• The grid widget's eight snake_case field-level keys stay out until objectui camelCases them.
• Not taken: A, objectui's open FormField; C, the form view's FormFieldSchema re-keyed by name.

Fork 3, both forms' sections: B, after fork 2.

• A page-block section shape of its own: FormSectionSchema's keys plus the three entry arms the form reads, with the inline arm referencing fork 2's field.
• Canonical spellings only. The stored form view is untouched.
• Not taken: A, widening the stored form view.

四棱(本裁决新记录)

• ① 长远:一份报表契约在生产端修;一份封闭、驼峰的表单字段契约;内嵌字段留在页面块,存储视图不动。
• ② 拉动:分叉 1 实测 1 个画出写法,B 下照过;分叉 2 实测 27 值全过、零蛇形键;分叉 3 实测 7 个内嵌写法全在 objectui 自家向导与测试。
• ③ 防 AI:未绑数据集的联合报表、拼错的字段键、废弃的分区拼法都在校验时响亮拒绝。
• ④ 不扩散:分叉 1 只加一条校验;分叉 2 只声明有写法的成员;分叉 3 不放宽其他契约。
• 只看①:三项均选 B;②③④ 是否翻转:否。

Execution

• This card: needs-user-decision stays until forks 4 and 5 are ruled.
• Ships as one more stage of spec(ui): the ComponentPropsMap rows still type renderer-read members as z.unknown() — navigation on object-map / object-gantt / object-tree and conditionalFormatting on object-kanban accept 42 — the family close-out after #21445 #21464, after PR feat(spec)!: object-kanban conditionalFormatting takes the list view's own rules, and its hold exits (#21464, S-kanban-cf) #21711 (S-kanban-cf) lands, because both edit ui/component.zod.ts and the enumeration pin.
  • Forks 2 and 3 go together: the runtime form field declared once, then the section shape referencing it.
  • Fork 1 rides the same stage if [finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702 is on main by then; otherwise it is its own short stage after [finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702.
  • Forks 4 and 5, once ruled, join the same stage where they can.
• Each typing is a narrowing: Clause-②: yes (narrowing), D3 entries, the ADR-0087 marker, minor, and an at-tier contract review, in the earlier stages' shape. Census first. A measured writer refused under the ruled shape goes back to the maintainer before landing.
• objectui side: an objectui card renames the grid widget's eight snake_case field-level keys to camelCase. It is filed with the stage, so fork 2's deferred keys have a carrier.
• The enumeration pin's fork lines for forks 1–3 leave as each is typed.

---

Generated by Claude Code

[objectstack-fleet]

Ruling: batch #277 items 1–2 · forks 4 and 5 · letters B, A · maintainer 「277 同意」 2026-10-04T10:57Z

Director seat, summon #32, session_016tKoy8NJa35Yih1FdzrVmn. Written as objectstack-fleet[bot] through the relay.

• How it was ruled. Batch 🔗 Broken links detected in documentation #277 was presented in the live director chat. Its items 1 and 2 were this card's forks 4 and 5, each with the options this card lays out. The director recommended B for fork 4 and A for fork 5, and offered A+ (declare outcomeMessages on all four action blocks now). The maintainer answered 「277 同意」, so fork 5 is plain A.
• All five forks are now ruled: 1 B, 2 B, 3 B (5978663135); 4 B, 5 A (this record).
• Freshness gate: no comment on this card since 5978663135.

The rulings

Fork 4, object-timeline items: B.

• Both arms are closed: TimelineFeedItem and TimelineGanttItem, objectui#6356's ruled contract, taken as it stands.
• Feed content is an opaque member for now, with its own ledger reason. It becomes a slot position only when a writer appears.
• A row refinement pairs each entry with the arm its parent's variant selects.
• Gantt-bar dates are a string or a number.
• Not taken: A, content as a slot position the page walks judge; C, a plain union with no variant pairing.

Fork 5, the action:group / action:menu members: A.

• The measured read set, action:button's keys by type, with the rows' prescriptions. endpoint goes through the rows' target alias, and enabled and autoTrigger follow the rows.
• outcomeMessages, a member className and properties.params are refused.
• outcomeMessages stays undeclared on all four action blocks (action:button, action:icon, action:group, action:menu), as one decision. It is declared on all four together once an inline writer appears.
• Not taken: B, declaring the three undecided keys; C, the object-metadata ActionSchema by reference.

四棱(本裁决新记录)

• ① 长远:分叉 4 原样承接 objectui#6356 的裁定,不为零写法成员改页面遍历;分叉 5 内联操作只留按钮那一套实测词汇。
• ② 拉动:分叉 4 实测 14 值零个带 content;分叉 5 实测 21 个成员零个用到未决键。
• ③ 防 AI:条目种类混用、菜单成员写未决键都在校验时响亮拒绝。
• ④ 不扩散:不加插槽位置;不声明三个零写法的键。
• 只看①:分叉 4 选 B、分叉 5 选 A;②③④ 是否翻转:否。

Execution

• This card is closed completed in this act: every fork is ruled, and the execution is spec(ui): the ComponentPropsMap rows still type renderer-read members as z.unknown() — navigation on object-map / object-gantt / object-tree and conditionalFormatting on object-kanban accept 42 — the family close-out after #21445 #21464's.
  • Forks 2 and 3: stage S-forms, claim 5979114945, in progress.
  • Forks 4 and 5: the stage after S-forms, serial on ui/component.zod.ts and the enumeration pin.
  • Fork 1: after [finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702 lands.
• The enumeration pin's items[] fork line names two shapes where this card offered three. That line leaves when fork 4 is typed, so nothing to align.
• Each typing is a narrowing, in the earlier stages' shape: census first, Clause-②: yes (narrowing), D3 entries, the ADR-0087 marker, minor, and an at-tier contract review. A measured working writer refused under the ruled shape goes back to the maintainer before landing.

---

Generated by Claude Code

[objectstack-fleet]

Closed completed by the director seat, summon #32 (session_016tKoy8NJa35Yih1FdzrVmn), 2026-10-04T10:58Z. Every fork on this card is ruled: 1 B, 2 B and 3 B (5978663135, batch #276); 4 B and 5 A (5979239990, batch #277).

The execution is #21464's:

• forks 2 and 3: stage S-forms, claim 5979114945, in progress;
• forks 4 and 5: the next stage;
• fork 1: after [finding] spec(report): a type: 'joined' report whose blocks bind no dataset parses and passes objectstack validate, while ReportSchema's own refinement comment and reports.mdx say "each block dataset-bound" #21702 lands.

#21464 stays open until all five are typed.

---

Generated by Claude Code