skill: Human label taxonomy 收敛为两个 active label(#15 共识) by loning · Pull Request #22 · ChronoAIProject/consensus-rnd

loning · 2026-05-25T16:31:29Z

🤖 skill: Human label taxonomy 收敛为两个 active label(#15 r4 共识)

Phase 9 r4 structural 共识落地。active Human label 仅保留:

🤖 human:auto-推进
👤 human:需-maintainer-决策

移除 🆘 human:卡死 / 🆘 human:卡死-需-rework 出 active taxonomy / bootstrap / producer / waiting 判定;rework / deadlock / CI-耗尽等原因改放 reason surface(auto-loop-stuck、phase label、banner/comment、PushNotification、marker reason)。

5 文件:SKILL.md、concurrency_monitor.py、controller_lib.sh、peek.sh、test_ensure_project_rules_fixed_points.py(+5 source-regression 测试)。

共识依据:.refactor-loop/runs/phase9-issue15-r4-judge.md。

Closes #15

⟦AI:AUTO-LOOP⟧

Phase 9 r4 structural 共识:active Human label 仅保留 🤖 human:auto-推进与 👤 human:需-maintainer-决策;移除 🆘 human:卡死 / 🆘 human:卡死-需-rework 出 active taxonomy/bootstrap/producer/waiting 判定;rework/deadlock/CI-耗尽等原因移到 reason surface(auto-loop-stuck label / phase label / banner / PushNotification / marker reason)。新增 5 个 source-regression 测试。 Closes #15 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

loning · 2026-05-25T16:32:29Z

📊 状态卡片 — reviewer 派出

维度	值
阶段	派出 codex(role=`reviewer`)
codex log	`review-pr22-architect.log`
工作目录	`/Users/auric/consensus-rnd`
no-output stall window	5400s(~90 min 无输出窗口)
上下文	Phase 8 review(architect 角度):#15 Human label taxonomy 收敛共识落地,3 reviewer 独立裁决。
下一步自动会做	1. 三 reviewer 完成 verdict marker 2. controller 计算 consensus 3. unanimous → auto-merge / reject → fix r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-25T16:32:30Z

📊 状态卡片 — reviewer 派出

维度	值
阶段	派出 codex(role=`reviewer`)
codex log	`review-pr22-tests.log`
工作目录	`/Users/auric/consensus-rnd`
no-output stall window	5400s(~90 min 无输出窗口)
上下文	Phase 8 review(tests 角度):#15 Human label taxonomy 收敛共识落地,3 reviewer 独立裁决。
下一步自动会做	1. 三 reviewer 完成 verdict marker 2. controller 计算 consensus 3. unanimous → auto-merge / reject → fix r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-25T16:32:31Z

📊 状态卡片 — reviewer 派出

维度	值
阶段	派出 codex(role=`reviewer`)
codex log	`review-pr22-quality.log`
工作目录	`/Users/auric/consensus-rnd`
no-output stall window	5400s(~90 min 无输出窗口)
上下文	Phase 8 review(quality 角度):#15 Human label taxonomy 收敛共识落地,3 reviewer 独立裁决。
下一步自动会做	1. 三 reviewer 完成 verdict marker 2. controller 计算 consensus 3. unanimous → auto-merge / reject → fix r<N+1>
是否需要人介入	❌ 否(自动推进)

🤖 controller status banner

⟦AI:AUTO-LOOP⟧

loning · 2026-05-25T16:34:47Z

🤖 Quality review: approve

TL;DR

这是什么:PR skill: Human label taxonomy 收敛为两个 active label(#15 共识) #22 的 readability/simplicity 角度独立 review。
结论:approve,没有发现命名、死代码、过度抽象、重复、复杂度或 self-doc 注释阻塞点。
下一步:controller 可把这个 quality verdict 纳入 PR skill: Human label taxonomy 收敛为两个 active label(#15 共识) #22 的综合判定。

详细说明

这次改动集中在 Human label taxonomy:active label 收敛到 🤖 human:auto-推进 和 👤 human:需-maintainer-决策,把 rework / deadlock / CI-stuck 这类原因放到 reason banner/comment 里。实现侧没有新增单实现抽象,也没有把旧 🆘 human: 重新作为生产路径暴露出来。

我完整打开了所有 touched file。concurrency_monitor.py 的 waiting predicate 保持为一条直接条件,controller_lib.sh 只做 cleanup label 扩展,peek.sh 的 operator hint 和 taxonomy 对齐,新增 source-regression tests 的 helper 命名也能直接表达业务意图。Refactor self-doc comment 在主要落点都存在且能说明 Old/New。

📎 完整 codex 原始输出(存档备查)

---
pr: 18
role: quality
verdict: approve
---

## Verdict
Approve: the change is focused on the Human label taxonomy, the new names express the two active label states clearly, and I found no dead code, over-engineering, avoidable duplication, or unclear refactor self-doc blocks.

## Evidence
- `skills/codex-refactor-loop/SKILL.md:2395`: the Label group now exposes exactly the two active Human states, `🤖 human:auto-推进` and `👤 human:需-maintainer-决策`, with rework/deadlock/CI-stuck moved into the reason surface instead of new label names.
- `skills/codex-refactor-loop/scripts/concurrency_monitor.py:184`: `compute_expected` remains a small single-purpose function; the changed waiting predicate is one direct condition and does not introduce a new helper or abstraction.
- `skills/codex-refactor-loop/scripts/controller_lib.sh:62` and `skills/codex-refactor-loop/scripts/controller_lib.sh:154`: the cleanup changes remain scoped to removing active and legacy Human labels; no `🆘 human:` label is produced on any new `--add-label` path.
- `skills/codex-refactor-loop/scripts/peek.sh:145`: the routing hint now names the active maintainer-decision label plus reason banner, keeping the operator-facing wording aligned with the two-label taxonomy.
- `skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py:519`: the new source-regression test class uses readable, domain-specific helper names (`CANONICAL_HUMAN_LABELS`, `NON_AUTO_HUMAN_LABEL`, `REMOVED_HUMAN_LABELS`) and each helper is used by the tests.
- Refactor self-doc comments are present and understandable at each touched implementation/test/doc site I checked, including `SKILL.md:172`, `concurrency_monitor.py:186`, `controller_lib.sh:62`, `peek.sh:113`, and `test_ensure_project_rules_fixed_points.py:520`.

⟦AI:AUTO-LOOP⟧

⟦AI:AUTO-LOOP⟧

loning · 2026-05-25T16:35:34Z

🤖 tests review: PR #22 通过

TL;DR

这是什么: tests 视角独立 review,只看测试覆盖与测试质量。
结论: approve,新增测试覆盖了 Human label taxonomy 的核心回归点。
下一步: controller 可继续等待其他 reviewer verdict,tests 侧无需返工。

详细说明

这轮改动没有碰 src/ 或 agents/,所以 test/.../<TypeName>Tests.cs 映射规则不适用。实际测试落在 skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py,覆盖了 Human label 表、bootstrap 创建、escalation 路由、concurrency_monitor.compute_expected 的等待判定、controller cleanup 与 peek.sh routing hint。

我也扫了稳定性红线:没有新增 sleep / delay 测试节奏,没有 polling allowlist 改动,没有 [Skip],没有 [Trait("Category","Manual")],没有弱化既有断言,也没看到 mock-only pseudo-coverage。从 PR head 的 isolated archive 跑 python3 -m py_compile ... && python3 skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py,结果是 35 tests OK。

📎 完整 codex 原始输出(存档备查)

pr: 18
role: tests
verdict: approve

Verdict

Approve: the PR adds meaningful regression coverage for the two-active-human-label taxonomy and does not weaken test quality or stability.

Evidence

skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py:538 adds test_human_label_taxonomy_has_single_non_auto_label, which verifies the documented Human label table and bootstrap block contain exactly the two active labels and no longer create the removed 🆘 human:* labels.
skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py:556 adds test_human_escalation_routes_use_reason_surface, which asserts escalation routes use 👤 human:需-maintainer-决策 plus reason/banner/PushNotification surface instead of the removed emergency labels.
skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py:579 exercises the changed concurrency_monitor.compute_expected behavior directly: 👤 human:需-maintainer-决策 suppresses expected codex, while the removed 🆘 human:* labels no longer suppress no-gap monitoring.
skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py:601 and :612 add source-regression assertions for controller cleanup and peek routing, blocking accidental reintroduction of 🆘 as a produced/recommended Human label while still allowing legacy cleanup paths.
No src/ or agents/ production files are touched, so the test/.../<TypeName>Tests.cs mapping rule does not apply to this markdown/script skill PR.
Diff scan found no added sleep / delay test pacing, no polling allowlist change, no [Skip], no [Trait("Category","Manual")], no bump-only assertion, no loosened existing assertion, and no mock-only pseudo-coverage.
Verified PR head from an isolated archive: python3 -m py_compile skills/codex-refactor-loop/scripts/concurrency_monitor.py skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py && python3 skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py passed, 35 tests OK.

REVIEW_DONE:22:tests:approve

⟦AI:AUTO-LOOP⟧

loning · 2026-05-25T16:39:15Z

🤖 Architect review: comment，源码注释语言需收敛

TL;DR

这是什么:PR skill: Human label taxonomy 收敛为两个 active label(#15 共识) #22 的架构合规 review，结论是 comment，不阻塞架构合并。
现在结论:Human label taxonomy 收敛本身符合 Phase 9 授权范围，但新增的 .py / .sh # Refactor 注释用了中文。
需要 maintainer 做什么:建议作者把这些 source comments 改成英文，保持 Old/New pattern 内容不变。

详细说明

从架构角度看，diff 只改了 skills/codex-refactor-loop/SKILL.md 和同目录 scripts/ 下的 helper/test，匹配 #15 Phase 9 共识范围；没有 .proto、版本号、外部 repo reference，也没有新增 actor/serializer/HttpClient/disabled-test 这类红线。

唯一需要 surfacing 的是源码语言 policy:这个 skill 自己写明 source files 内部的注释、docstring、log/error 字符串应为英文。PR 新增的 # Refactor 注释在 concurrency_monitor.py、controller_lib.sh、peek.sh、test_ensure_project_rules_fixed_points.py 里包含中文说明。建议改成类似: Old pattern: four Human labels, including two emergency labels, scattered no-gap/escalation predicates. New principle: exactly two active Human labels; causes move to reason surfaces (#15 structural consensus).

📎 完整 codex 原始输出(存档备查)

pr: 22
role: architect
verdict: comment

Verdict

comment-only: the taxonomy change stays inside the Phase 9 authorized scope, but several new source comments violate the skill's own source-English policy and should be rewritten before merge if the team wants zero lint drift.

Evidence

skills/codex-refactor-loop/scripts/concurrency_monitor.py:187: new # Refactor comment uses Chinese prose in a source file (四个 Human label...). The skill's hard rule states: Source files are English-only; external user-facing artifacts are 中文 by default; the detailed policy states: 英文适用对象:所有源文件(.rs / .lua / .sh / .py / .ts)内部自然语言与代码元素,包括注释、docstring、log.{info,warn,error} 字符串、error / panic 文本、代码 identifier、代码内构造的 commit-body 模板字符串。
skills/codex-refactor-loop/scripts/controller_lib.sh:155: same issue in a shell source comment added for this refactor.
skills/codex-refactor-loop/scripts/peek.sh:114: same issue in a shell source comment added for this refactor.
skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py:28 and skills/codex-refactor-loop/scripts/test_ensure_project_rules_fixed_points.py:521: same issue in Python test source comments added for this refactor.
Scope check: changed files are limited to skills/codex-refactor-loop/SKILL.md and its scripts/ helpers/tests, matching the Phase 9 consensus file list. Root CLAUDE.md allows this exception: ``skills/codex-refactor-loop/为 host 项目移植,**verbatim**,仍带"重构"外壳与少量 host 主张。泛化路线见README.md` 的「泛化路线」。脱壳 / 重命名前不要改它的正文逻辑；例外：经 Phase 9 deep consensus 明确授权的 host-agnostic bootstrap / policy 注入修正可先落地，但不得引入具体 host 事实，且必须配套行为测试。`
Anti-pattern grep found no new actor.HandleEventAsync(, SubscribeAsync<EventEnvelope>, JSON serializer state/payload path, production sleep/delay(, sync-over-async marker, raw HttpClient constructor, Dictionary<,> cross-request state, or [Skip]/disabled tests in the added diff. No .proto, version, or external-repo-reference changes were present.

What would change your verdict (only if comment or reject)

Rewrite the newly added # Refactor source comments in .py and .sh files to English while keeping the Old/New principle content intact, for example Old pattern: four Human labels, including two emergency labels, scattered no-gap/escalation predicates. New principle: exactly two active Human labels; causes move to reason surfaces (#15 structural consensus).

REVIEW_DONE:22:architect:comment

⟦AI:AUTO-LOOP⟧

loning added auto-loop codex-refactor-loop 管理的 issue/PR 🚀 phase:pr-open 👀 phase:reviewing 🤖 human:auto-推进 labels May 25, 2026

loning mentioned this pull request May 25, 2026

skill: 契约 source-regression 测试套件 + TEST_CMD 可跑(#16 共识) #23

Merged

loning merged commit 4861b67 into auto-refact-dev May 25, 2026

loning deleted the refactor/iter3-skill-human-label-taxonomy branch May 25, 2026 16:42

loning added 🎉 phase:merged and removed 🚀 phase:pr-open 👀 phase:reviewing labels May 25, 2026

loning mentioned this pull request May 25, 2026

[refactor-design] human label 分类不一致:卡死 vs 卡死-需-rework(self-audit #12) #15

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

skill: Human label taxonomy 收敛为两个 active label(#15 共识)#22

skill: Human label taxonomy 收敛为两个 active label(#15 共识)#22
loning merged 1 commit into
auto-refact-devfrom
refactor/iter3-skill-human-label-taxonomy

loning commented May 25, 2026

Uh oh!

loning commented May 25, 2026

Uh oh!

loning commented May 25, 2026

Uh oh!

loning commented May 25, 2026

Uh oh!

loning commented May 25, 2026

Uh oh!

loning commented May 25, 2026

pr: 18
role: tests
verdict: approve

Verdict

Evidence

Uh oh!

loning commented May 25, 2026

pr: 22
role: architect
verdict: comment

Verdict

Evidence

What would change your verdict (only if comment or reject)

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

loning commented May 25, 2026

🤖 skill: Human label taxonomy 收敛为两个 active label(#15 r4 共识)

Uh oh!

loning commented May 25, 2026

📊 状态卡片 — reviewer 派出

Uh oh!

loning commented May 25, 2026

📊 状态卡片 — reviewer 派出

Uh oh!

loning commented May 25, 2026

📊 状态卡片 — reviewer 派出

Uh oh!

loning commented May 25, 2026

🤖 Quality review: approve

TL;DR

详细说明

Uh oh!

loning commented May 25, 2026

🤖 tests review: PR #22 通过

TL;DR

详细说明

pr: 18 role: tests verdict: approve

Verdict

Evidence

Uh oh!

loning commented May 25, 2026

🤖 Architect review: comment，源码注释语言需收敛

TL;DR

详细说明

pr: 22 role: architect verdict: comment

Verdict

Evidence

What would change your verdict (only if comment or reject)

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

pr: 18
role: tests
verdict: approve

pr: 22
role: architect
verdict: comment