docs(catalog): add IP-034 unstaffable team lane is a typed gap

[BigDataDZ]

Goal And Delivered Outcome

• Goal/source and gap: GH-C37 asks the interaction pattern catalog to carry one new public-safe good/bad case. fix(manager): keep an unstaffable team lane a typed gap and offer the admitted plan as a card #4569 shipped the behavior that keeps a lane this host cannot staff a typed gap while still offering the admitted plan as a card, but no catalog entry named that boundary: IP-018 owns plan-to-todo writeback and IP-024 owns the repair delta, and neither covers part of an admitted plan having no honest owner.
• Observable before → after: before, the catalog has no pattern for a partially unstaffable team plan, so a surface that invents a lane, fails the whole confirmation, or quietly omits the lane has no named pattern contradicting it; after, IP-034 names the boundary in the Planning Governance family and examples/interaction-pattern-catalog-smoke.py fails if the entry's core sentence, its family table row, or its family-to-canary matrix cell is removed.
• Issue/task and intended base: Closes [Task]: add IP-034 unstaffable team lane is a typed gap to the interaction pattern catalog (GH-C37) #4638 (GH-C37); base main.

Scope And Continuation

• Completed scope and remaining work: complete within this scope — one catalog entry with Trigger, Expected behavior, Visual Model, Bad smell, and Validation; the Planning Governance table row; the family-to-canary matrix entry; and three durable assertions in the catalog smoke. No runtime behavior changes.
• Slice boundary / successor: N/A with reason — the accepted GH-C37 slice is one new pattern and this PR delivers it whole.

Validation

• Tested revision: 84c9b49e2 (rebased onto current main)
• Run state: finished
• Input classes: none

Check kind	Result	Public-safe evidence / limitation
static	passed	python3 examples/interaction-pattern-catalog-smoke.py — ok; asserts IP-034 owns exactly one family table row and one matrix cell and that "An unstaffable lane is a typed gap, not an invented lane." plus the missing-registration wording survive
unit	passed	python3 -m pytest -q tests/test_manager_team_plan_guidance.py tests/test_steward_team_plan_apply.py — 19 passed; test_a_lane_whose_kind_the_host_does_not_ship_creates_nothing is the shipped behavior IP-034 documents (plan admitted, one Todo created for the lane that can run, none for the unstaffable lane)
real_entrypoint	passed	python3 -m loopx.entrypoint check --scan-path docs/concepts/interaction-pattern-catalog.md — ok, errors=0, public boundary scan clean; the 2 warnings are the pre-existing missing local registry on this checkout, unrelated to the diff
regression_parity	not_applicable	documentation-only change; no runtime path, payload, or CLI surface changed

• Coverage and gaps: the two unit files cover both halves the entry claims — the manager guidance wording (gap versus invention, no automatic regeneration) and the apply-side admission behavior. The catalog smoke protects the entry itself from silent deletion. No runtime path is touched, so no integration or backend gate applies. Not covered: the product surface that renders the gap as a card, which is maintainer-preview work and outside this slice.

Frontend / Visual Evidence

• UI impact: none

Type of Change

• Documentation update

LoopX Area

• Public docs or presentation surface (README, protocols, dashboard)

Technical Direction

• Direction / acceptance reference, when applicable: Core control-plane hardening; contributor task GH-C37, tracked by [Task]: add IP-034 unstaffable team lane is a typed gap to the interaction pattern catalog (GH-C37) #4638.

Shared-authority RFC fixture impact

• N/A

Boundary Checklist

• Neither the diff nor this PR body/comments/attachments disclose private state, credentials, raw traces or verifier output, internal links, or local machine paths (including .loopx/, .codex/goals/, and live ACTIVE_GOAL_STATE.md).
• I did not duplicate maintainer-owned benchmark work unless a maintainer split out a public issue for it.
• I kept the change scoped to the linked issue/task.
• I completed the visual evidence section for UI changes, or marked UI impact none.
• Every commit includes a DCO Signed-off-by trailer (git commit -s).

[huangruiteng]

Reviewed exact head: 84c9b49e296640b3e6b0f831558ea914e37b73d0 (docs/ip-034-unstaffable-lane-gap).

动机

交互模式目录（docs/concepts/interaction-pattern-catalog.md）记录可复用的 user/agent/state 交互形状，它自己的维护规则写明：只有当这套行为本身是可复用交互、且有 trigger / 渠道 / 状态契约 / bad smell / validation 时才分配新 IP；若已被某个 IP 覆盖，就应该把新的 smoke、fixture 或说明追加到那个 IP 里。steward/manager 的 team plan 路径在已发布代码里已经把「准入（admission）」和「配人（staffing）」分开处理，但目录里没有这条形状，于是 reviewer 每次都要重新推导，canary/readiness 的分组表也没有可指向的条目。

我按这条维护规则先去找「不加新 IP」的更强理由：目录里已有 33 个模式，最近邻是 IP-031（manager context 不是 turn authority，讲的是权威绑定）、IP-018（plan→todo writeback）和 IP-024（repair delta），它们分别拥有权威、写回与前沿修复，没有谁拥有「已被准入的计划里有一条 lane 没有诚实归属」这件事。新章节也明确写出与这两个相邻 IP 的边界，所以新增 IP 是正确层级，而不是往旧条目里塞。

改动思路

改动只有两个文件：目录文档与它的守卫 smoke。文档侧把 IP-034 注册到该出现的三处（Planning Governance 分组行、P1 行列表、完整模式章节），结构与其兄弟条目完全一致（Trigger / Expected behavior / Visual Model / Bad smell / Validation）；examples/interaction-pattern-catalog-smoke.py 增加三条必需字符串，让「条目被删掉」变成红灯——这和站内其它 IP 的登记方式一致（我核对了 IP-033 的三处出现方式，完全相同）。

行为侧我不看作者总结，而是直接找被点名的证据层次：apply 侧由 tests/test_steward_team_plan_apply.py 拥有（计划仍被准入、只创建一条可运行 lane 的 Todo、不可配的 lane 不落 Todo），preview 侧由 tests/test_steward_team_plan_preview.py 拥有（该 lane 的 staffing == "gap"、gap_reason_code、保留 declined_first_todo、preview["gaps"] 列出它），引导文案由 tests/test_manager_team_plan_guidance.py 对 manager_skill_text() 做 prose 断言。这三层我都实际跑过。

具体改动

2 个文件、+86/-1：docs/concepts/interaction-pattern-catalog.md +83/-1（分组行就地扩展、P1 行新增、完整章节新增）、examples/interaction-pattern-catalog-smoke.py +3（三条必需字符串）。没有生产代码、没有 canary、没有授权面变化。

关键代码讲解

• 分组行（docs/concepts/interaction-pattern-catalog.md:93）：把 IP-034 放进 Planning Governance，使这条形状能被 canary/readiness 分组选择到——这是「写目录」而不是「写一段说明」的实际作用点。
• 完整章节（:2771 起）：四条 expected behavior 正好对应「不发明 lane/Agent/capability/action kind」「整份计划仍然被准入」「不静默丢弃 lane」「gap 不会自愈」，并配 mermaid 图与 bad smell，符合目录模板。
• smoke（examples/interaction-pattern-catalog-smoke.py:149）：三条 require 字符串把这个条目纳入既有守卫，而不是新建一份只服务于本 PR 的检查。

对主干的风险

我跑了：python examples/interaction-pattern-catalog-smoke.py → ok；pytest tests/test_steward_team_plan_apply.py tests/test_manager_team_plan_guidance.py -q → 19 passed。头部 CI 14 项成功、0 失败，review 时 4 项仍在排队（test-shard 1/2、stage2c e2e 1、checks），merge_state=BEHIND 属于待更新分支。

我也扫了同类产物：rg 显示 "unstaffable" 只出现在新章节、smoke 和那两个测试文件里，没有第二份文档重复描述这条形状；作者批量上，BigDataDZ 的目录类 PR 分布在 09-13、09-14、09-17 多天（IP-029/IP-028/IP-031/IP-034），中间夹着其它合并，不属于几分钟内同形状连发（同日的 #4642 是已关闭的 CLI 重构，与本文档无关）。

两条非阻塞问题：

1. Validation 一节的层次归属不准（P2）。它写「tests/test_manager_team_plan_guidance.py pins the preview contract」，但那个文件断言的是 SKILL.md 的文案，真正机器强制的 preview 契约由 tests/test_steward_team_plan_preview.py::test_an_unsupported_action_kind_gaps_that_lane_instead_of_refusing_the_plan 拥有（staffing == "gap"、gap_reason_code == "action_kind_not_supported"、保留 declined_first_todo、first_todo 不存在、preview["gaps"] 列出该 lane），而这一条没有被引用。读者按 Validation 去找证据，会把机器强制的半边误当成只有文案守护——这正是仓库反复强调的 guidance 与 obligation 不能混。建议补上该 preview 测试的引用，并把 guidance 测试明确标注为文案契约。

2. 分组行的 evidence 列没有跟着更新（P3）。Planning Governance 行现在选择了 IP-034，但该行的典型证据仍只写「two-turn stalled fixture plus repair/writeback delta assertion」，没有体现这条形状的证据是「preview 的 staffing gap + apply receipt」。这行存在的意义就是把行映射到 canary archetype，建议把这句补进 evidence 列。

我的整体评价

结论 APPROVE。这是一次边界正确、层级正确的目录补充：它没有为新 IP 而新 IP（先论证了三个最近邻 IP 都不拥有这条形状），登记位置与该目录既有约定一致，并且我逐条验证了它宣称的行为在 head 上有真实可运行的证据（preview / apply / guidance 三层，19 项测试通过）。smoke 的 3 行新增是把条目纳入既有守卫，而不是再造一份只服务本 PR 的检查，符合「薄而持久」的要求。

两条问题都不影响这个判断：它们都是「读者会顺着目录找到错误层次/不完整映射」的文档精度问题，最小修法已经写明，不需要改运行时行为。按仓库规则，本评审只给出 exact-head 结论；合并就绪度（CI 收敛、BEHIND 更新）属于另一道关卡。

English verdict: APPROVE - exact head 84c9b49; IP-034 is registered in the same three places as its siblings, its claims are backed by shipped tests I ran (apply-side exactly-one-Todo, preview-side per-lane staffing gap, guidance prose; 19 passed plus the catalog smoke ok), and the 3-line smoke addition extends the existing guard instead of adding a one-off check. Two non-blocking findings: the Validation section attributes the preview contract to the guidance-text test and omits tests/test_steward_team_plan_preview.py that actually enforces it (P2), and the Planning Governance group row's evidence column was not extended for the new pattern (P3).