菜单
Rigorous plan review via the Momus consultant. Validates clarity, verifiability, and completeness before execution; returns OKAY or REJECT.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Runtime debugging loop, reproduce, form ranked hypotheses, instrument, fix, verify. Use for crashes, wrong output, intermittent failures, race conditions, or anything that works locally but not in prod. Triggers: "debug this", "why is this failing at runtime", "intermittent failure", "race condition", "works locally but not in prod", "flaky test", "silent failure". NOT for build failures, type errors, or toolchain issues; those belong to the hephaestus flow. This skill owns runtime misbehavior only.
Strip AI-generated code slop from changed files: double-guards, dead fallbacks, redundant re-validation, narrating comments, speculative flexibility, premature abstraction, and more. Each category pairs with a KEEP rule and a trust-boundary proof requirement, so cleanup never strips load-bearing validation or error handling. Triggers: 'remove ai slop', 'clean up generated code', 'de-slop', 'remove defensive clutter', 'strip ai slop', 'clean up ai-generated code'.
Designer-turned-developer who crafts stunning UI/UX even without design mockups. Use for UI components, styling, layout, animations, and visual design work.
MUST USE for ANY git operations. Atomic commits, rebase/squash, history search (blame, bisect, log -S). Triggers: 'commit', 'rebase', 'squash', 'who wrote', 'when was X added', 'find the commit that'.
Configure ~/.claude/ for oh-my-claudeagent (deps check, block injection, settings, statusline).
Create a detailed context summary for seamless continuation in a new session.
| name | momus |
| description | Rigorous plan review via the Momus consultant. Validates clarity, verifiability, and completeness before execution; returns OKAY or REJECT. |
| context | fork |
| agent | oh-my-claudeagent:momus |
| user-invocable | true |
| argument-hint | [plan file path] |
| effort | high |
Review the work plan at: $ARGUMENTS
No path specified → ask the user for the plan file path. Accept a FILE PATH only, not an inline plan, todo list, or text summary.
Follow momus workflow: read the plan, deep-verify every file reference, apply the five evaluation criteria, run falsification on the 2 most critical tasks, and return the Final Verdict (OKAY / REJECT with confidence, justification, and priority-tiered issues).
Output: A single OKAY/REJECT verdict in the Final Verdict Format defined in agents/momus.md. This feeds the prometheus review loop.
