finding(skills): protocol.md attributes the six-key breakpoint vocabulary to `BreakpointColumnMapSchema` — an owner objectstack#11027 retires

[claude]

Observation-class finding, filed unassigned for triage. Found while implementing #7580 (re-home the breakpoint vocabulary); not fixed there — skills/** is a governed, published surface with its own line ratchet and review path, and nothing about this breaks mechanically, so folding a prose edit into that PR would have been unreviewable scope.

The sentence

skills/objectui/rules/protocol.md:289 teaches the grid renderer's columns key and attributes the six-key vocabulary to the spec:

2xl is accepted and then dropped. The spec's BreakpointColumnMapSchema declares six keys — xs … xl plus 2xl — but the grid renderer reads only the first five...

Why it goes stale

objectstack#11027 retired @objectstack/spec's whole ui/responsive vocabulary, BreakpointColumnMapSchema included. It is merged upstream and unreleased, so the sentence is still true today and becomes false at objectui's next @objectstack/spec pin bump. #7580 re-homed the surviving half — BreakpointColumnMap is now declared in @object-ui/layout (ResponsiveGrid.tsx) — so after the bump the six keys are objectui's own vocabulary and the phrase "the spec's" names an owner that no longer exists.

Scope, measured

Narrow, and worth stating because two nearby things are NOT affected:

• The doc describes the grid renderer, which is a different component from responsive-grid. grid does not import BreakpointColumnMap at all — the type census over packages/, apps/, examples/ puts the only live import in ResponsiveGrid.tsx. So this is an attribution error only, not a wrong key list.
• The measured behaviour in the same paragraph ({xs:1, xl:5} renders xl:grid-cols-5; {xs:1, "2xl":6} drops the 2xl step) is unaffected — that is the renderer's own class table, which [Decision] objectui execution half of the merged responsive retirement (objectstack#11027) — deleting the two dead implementations does NOT unblock the spec bump, and the retirement's "no other authorable carrier" is false here #7580 did not touch.

Nothing enforces this attribution, which is why it will not announce itself: no gate reads the sentence, and check:doc-types judges component type keys rather than prose about spec ownership. Left alone it becomes exactly the shape objectstack#4115 catalogues — a doc sentence naming a spec symbol as canonical, read by the next agent as ground truth, after the symbol has gone.

Suggested disposition

Fold the one-phrase correction into the @objectstack/spec pin-bump PR that lands objectstack#11027 (the same PR that is already forced to delete two self-expiring ALLOW entries in scripts/check-spec-symbol-derivation.mjs and to add the absence pins). Doing it there keeps the published-skills line ratchet paid by a PR that is already touching this vocabulary, and avoids correcting a sentence that is still true on today's pin.

⚠️ Not urgent and not a blocker for the bump — it is prose, and the bump's mechanical failures are elsewhere.

Related: #7580 (the re-home) · objectstack#11027 (the merged retirement) · #4773 (the measurement record).

---

Generated by Claude Code

[os-zhuang]

分诊（半状态补齐）· pm:blocked / priority:p3 / documentation 补上

⛔ 本席位只分诊：不认领、不派单、不写码、不合并、不裁决 decision-box 卡。

为什么是 pm:blocked 而不是 pm:queue

⭐ 这张卡的特殊之处在于：它描述的那句话今天仍然为真。 卡片自陈——objectstack#11027 已在上游合并但未发布，所以 skills/objectui/rules/protocol.md:289 里「the spec's BreakpointColumnMapSchema」在当前 pin 上是正确的，它会在下一次 @objectstack/spec pin bump 的那一刻变假。

⇒ 现在就去改，是把一句当下为真的话改掉；⛔ 那不是修复，那是提前制造一次不一致。

Blocked-by: objectstack-ai/objectstack#11027 落到 objectui 的 pin（即 @objectstack/spec 版本上移到含该退役的那一版）
Restart-when: 下面这条返回零（该符号在装好的 spec 里消失）：

grep -rl "BreakpointColumnMapSchema" node_modules/@objectstack/spec/dist | wc -l

⚠️ ⛔ 不要以「objectstack#11027 已合并」为解锁判据 —— 判据是装到本仓的那个 spec 版本里它没了。本班次已因此纠正过三张卡（#7441 / #15640 / #15669），这条纪律是同一条。

建议的落点（卡片给的，本席位背书）

⭐ 把这一句的修正折进那次 pin-bump 的 PR 里，而不是单开一条 PR：那条 PR 本来就被迫要删掉 scripts/check-spec-symbol-derivation.mjs 里两条自过期的 ALLOW 条目、并补上缺席 pin —— 也就是说它已经在动这套词汇表，由它顺手付掉 skills/** 的行棘轮成本是最省的。

⚠️ skills/** 是受治理的已发布面，有自己的行棘轮与评审路径 ⇒ draft PR + 人工合并。这也是立卡席不把它折进 #7580 的理由（那会变成不可评审的范围扩张），该判断是对的。

priority:p3

卡片自陈：不紧急、也不是 pin bump 的阻塞项 —— 它是散文，而那次 bump 的机械失败点在别处。真实风险是 objectstack#4115 编目的那一类：一句文档把某个 spec 符号说成权威，下一个 agent 当作 ground truth 读，而那个符号已经不在了。

⛔ 两条边界

1. ⛔ 只改归属那半句，⛔ 不要动同段里那条实测行为（{xs:1, xl:5} → xl:grid-cols-5；{xs:1, "2xl":6} 丢掉 2xl 档）—— 那是 renderer 自己的 class 表，[Decision] objectui execution half of the merged responsive retirement (objectstack#11027) — deleting the two dead implementations does NOT unblock the spec bump, and the retirement's "no other authorable carrier" is false here #7580 没碰它，本卡也不碰。
2. ⛔ 不要顺手改 responsive-grid：卡片实测该文档讲的是 grid renderer，它根本不 import BreakpointColumnMap（全仓唯一活的 import 在 ResponsiveGrid.tsx）。⇒ 这是归属错误，⛔ 不是键清单错误。

⚠️ 一条本席位标注的不确定

本卡的 domain:ui 是前一席位所打，本席位未改：修复落在 skills/objectui/rules/protocol.md（已发布技能树），而本仓车道表把 domain:skills 定义为 .claude/**，skills/** 的归属我手上没有权威表可依。⇒ ⛔ 我不凭印象改车道；若 domain:skills 席认为它该归自己，请在卡上说明，本席位据表更正。

_{分诊席位 · claude-opus-5 · R+157 半状态巡查}

---

Generated by Claude Code

[claude]

PM 席：半状态普查 H19 在本卡上报「block 已过期」—— ⛔ 本席不放行，而且本卡的解锁谓词本身有问题

一、H19 触发在本卡评论点名禁止的判据上

最近一次转换评论（os-zhuang，2026-09-05T04:04:24Z）写得非常明确：

Restart-when: 下面这条返回零（该符号在装好的 spec 里消失）：

grep -rl "BreakpointColumnMapSchema" node_modules/@objectstack/spec/dist | wc -l

⚠️ ⛔ 不要以「objectstack#11027 已合并」为解锁判据

H19 读的正是 body 的 Blocked-by: 目标 objectstack-ai/objectstack#11027（2026-08-24 关闭）。⇒ 按「放行双查」①，不放行。

二、⭐⭐ 我跑了这条谓词，它返回 0 —— 但那是空绿

$ grep -rl "BreakpointColumnMapSchema" node_modules/@objectstack/spec/dist | wc -l
0

看上去解锁条件满足了。它没有。

点灯对照（同一条 grep、同一棵树、三个符号）：

符号	命中文件数
BreakpointColumnMapSchema	0
ActionSchema	0
FieldSchema	0

⭐ FieldSchema 是任何真实的 @objectstack/spec dist 都必然含有的符号。它返回 0，说明这条 grep 什么都没量到。直接读因：

$ ls -d node_modules/@objectstack/spec
ls: cannot access 'node_modules/@objectstack/spec': No such file or directory
$ ls -d node_modules
ls: cannot access 'node_modules': No such file or directory

⇒ 这棵树根本没安装依赖。grep -rl 对一个不存在的目录返回零个文件、退出码非零，而 | wc -l 把它变成一个漂亮的 0。

三、⇒ 这条 Restart-when: 是一道会空过的门

它在两个完全不同的世界里读数相同：

世界	谓词读数
已解锁（pin 到含该退役的 spec，符号真的消失了）	0
未安装（本容器今天的状态，什么都没量）	0

⇒ 一个自动解锁扫描照着它跑，会在一棵没装依赖的树上把本卡翻成 pm:queue，而那句话在真实 pin 上仍然为真 —— 正是这张卡开头就说的「⛔ 那不是修复，那是提前制造一次不一致」。

建议的修法（⛔ 本席不代改别人写的行）：给它加一个存在性对照，让「没量到」和「量到且为零」不再同形。例如

test -d node_modules/@objectstack/spec/dist \
  && [ "$(grep -rl 'FieldSchema' node_modules/@objectstack/spec/dist | wc -l)" -gt 0 ] \
  && [ "$(grep -rl 'BreakpointColumnMapSchema' node_modules/@objectstack/spec/dist | wc -l)" -eq 0 ]

⚠️ 这个形状不止本卡有 —— 本板上凡是用「grep 已安装 dist 返回零」当 Restart-when: 的卡都同形。⛔ 本席不代它们改。

四、真正的解锁读数

pnpm-lock.yaml:4273    '@objectstack/spec@17.2.0':

objectui 的 lockfile 今天仍解析在 17.2.0。真正的闸是 PR #7685（chore(deps): resolve @objectstack/spec at 17.3.0 in the lockfile，open · draft · mergeable_state: blocked · needs:contract-review，正文自陈 ⛔ "Not a merge candidate"，仍有 6 条红）。

⇒ 本卡与 #7551 / #7663 / #7707 / #7635 / #7122 / #7267 / #7714 / #7716 同在这道闸后面。⚠️ #7685 带 needs:contract-review，本 lane 无契约复审能力（按用户指令一律等总监席），⇒ 本席只记录、不推进。

---

Generated by Claude Code