Cmochance

用户显式触发的本地代码审查:按 diff 特征自动从 lens 池选 3-6 个相关审查维度(不是固定 3-agent、也不是几十 agent 的 workflow),跑 Route→Fan-out→Verify→Synthesize,对抗式 verify 杀假阳性后按严重度出报告。替代项目「固定 3-agent 本地审查」约定。三档 quick/standard/thorough,默认 standard;thorough 复用现有 /code-review Workflow 不重造。仅用户显式触发,不自动跑。

分级路由的代码审查:把多维度审查的「广度发现」层 offload 到 grok CLI(用 grok 订阅省 Claude 额度),Claude 只做高档对抗验证 + 综合。当多 agent 审查(pr-review-toolkit:* 子 agent)或 code-review workflow 被触发时,PreToolUse hook 会 deny 并把发现层改道到本 skill 的 grok 编排器。grok-only MVP(agy 第三腿暂缓)。

用户说「merge 收尾」时的固定收尾 SOP(step0-6):stacked child PR base 解耦、squash merge、.app 转移加 test 前缀、清理远端/本地/worktree 并逐项 verify、关联 GitHub issue、Linear MOC transition + step6.2 强制扫描未完成 issue 挂强相关。含搭车改动补 Done issue、外部 contributor workflow approve、fork PR 收尾差异、深层 stacked squash、reviewThreads 分页坑等变体。没听到「merge 收尾」明确指令绝不主动 merge。

发版/release 全流程 SOP:dispatch 前预发布微 PR(版本号全盘 grep bump、双语 release notes 硬模板 2500-3200 字符、release-notes/vX.Y.Z.md CI 门禁)、release.yml workflow_dispatch 版本校验、merge cleanup 顺序、build 完保持 draft 等用户显式确认才转 Latest。用户说「发版/准备 release/触发 release workflow」时使用。

本地 Plane work item 索引(Plane Cloud)。需要定位、查找、回忆 Plane work item 时(按编号/关键词/状态/标签/project/归档),先查本地索引再按需 MCP 拉详情;也负责索引的增量/全量刷新。与 linear-index 同构,但适配 Plane(原生归档、无限 issue、PQL、external_id 回链 Linear)。

通过 CDP 注入驱动真实运行的 Codex Desktop 跑一轮对话并读回 assistant 回复,对 codex-app-transfer 的 Codex 集成做 E2E/冒烟测试(验证 provider 路由 / 协议适配 / 注入 / 代理链路端到端是否真的工作),替代很多手搓 E2E。默认新建独立会话测试、结束精准归档,绝不动用户已有会话。当需要"在真实 Codex 里发一句 prompt 看实际效果"而不仅是跑单测时使用。

本地 Linear issue 索引。需要定位、查找、回忆 Linear issue 时(按编号/关键词/状态/标签/里程碑/归档状态),先查本地索引再按需 MCP 拉详情;也负责索引的增量/全量刷新。merge 收尾扫描相关 issue 时同样先走本索引筛候选。

大型任务拆多个 stacked PR + 双工具 review 的完整 SOP(A-H 章节):CI trigger 适配 stacked PR、单 PR 全程双轨记录、CI 绿即 mark ready、PR-comment Monitor 与子代理 fix 判断表、stacked 链 retarget/force-push、hotfix 盘点、review 反馈循环、SOP 自进化机制。规划或推进多 PR / stacked PR 的重构、清理类大型任务时使用。

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Use when you have a written implementation plan to execute in a separate session with review checkpoints

Use when starting any conversation - establishes how to find and use skills, requiring skill invocation before ANY response including clarifying questions

Use when creating new skills, editing existing skills, or verifying skills work before deployment

Submission-grade Nature/high-impact journal figure workflow for Python or R. Use whenever the user asks to create, revise, audit, or polish manuscript figures, multi-panel scientific plots, figures4papers-style matplotlib plots, or journal-ready SVG/PDF/TIFF outputs, especially for Nature-family or other high-impact journals. Before plotting, define the figure's conclusion, evidence logic, export needs, and review risks. If the user has not chosen Python or R, ask "Python or R?" and stop. Use only the selected backend for figure generation, previewing, exporting, and QA. Supports matplotlib/seaborn and ggplot2/patchwork/ComplexHeatmap. Not for dashboards or Illustrator/Figma-first infographics. Also trigger on general academic-writing figure needs even without the word "Nature", such as making figures/plots for a paper, scientific/academic plotting, data visualization for a manuscript, and Chinese phrasings like 论文配图、学术写作配图、科研绘图、科研作图、画图、作图、出图、论文图表、可视化.

Multi-source literature search, citation verification, MeSH search strategy, citation file management (.nbib/.ris/.bib conversion), and reference management (BibTeX, related articles, ID conversion) via MCP tools (PubMed, CrossRef, arXiv, Scopus, ScienceDirect). Use when the user needs coordinated multi-step literature workflows beyond a single MCP call. Also trigger on general literature-search needs during academic writing even without the word "Nature", such as searching for papers/literature, doing a literature review, verifying a citation, converting citation files, and Chinese phrasings like 文献检索、查文献、找文献、 文献综述检索、查论文、引文核对、参考文献管理、文献去重.