菜单
Exhaustively clarify a prompt until 100% confident — asks every question, confirms every assumption. Use for high-stakes or ambiguous tasks where you want zero guesswork before work begins. More aggressive than `clarify`.
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
After a code change, test it end-to-end (bash, browser, or computer use — whichever fits) and then run `/simplify` to review and clean up the changed code. Use to confirm a change works and is tidy before wrapping up.
Create a git commit and push it
Create a git commit
Ask the user for maximum clarifications before acting on a prompt. Use when the user wants to ensure all ambiguity is resolved before work begins.
Write or refresh a README that earns the reader's attention — short, honest, with a real visual. Tailored to the project type.
Greenfield product spec for an empty repo. Clarifies scope, then challenges the idea — checking whether it already exists and whether the pain is real — before writing a concise spec.md.
| name | interrogate |
| description | Exhaustively clarify a prompt until 100% confident — asks every question, confirms every assumption. Use for high-stakes or ambiguous tasks where you want zero guesswork before work begins. More aggressive than `clarify`. |
| argument-hint | ["prompt or task description"] |
| disable-model-invocation | true |
| allowed-tools | AskUserQuestion |
| model | opus |
| effort | max |
$ARGUMENTS
Reach full confidence about what the user wants before acting. Do NOT implement during interrogation — only after every question is answered and the spec is confirmed.
AskUserQuestion: 4 questions per round, 2-4 answers each. Cover scope, edge cases, error handling, constraints, file location, dependencies, naming, testing, UX, migration — and anything else you'd otherwise guess. Do NOT skip dimensions because they "seem obvious"; confirm them. Mark the most reasonable default as the first option labeled (Recommended).