菜单
<What thinking mode this activates — verb phrase.> Use for: <comma-separated use cases.> Invoke when user mentions: <trigger keywords.>
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
Reviews a branch for quality, tests, error handling, and security. Returns PASS or FAIL with actionable findings. Use for: reviewing code, evaluating diffs, checking tests, checking error handling, checking security, reviewing comments, writing verdicts. Invoke when user mentions: fleet, review, code review, verdict, findings, code quality, test quality, error handling, security, check comments, review-pr.
Builds code in an isolated worktree — implements features, fixes bugs, writes tests, and submits for review when done. Use for: implementing features, fixing bugs, writing tests, building in worktrees, broadcasting changes, sending/replying to agent messages, submitting for review, writing task summaries. Invoke when user mentions: fleet, build, implement, code, fix, test, worktree, broadcast, message, reply, submit, review, summary, publish, verify.
Splits work into parallel tasks and swarms agents across worktrees. Plans, decomposes, monitors, merges, and ships. Use for: planning features, writing specs, decomposing tasks, splitting issues, running parallel agents, swarming worktrees, monitoring progress, merging branches, resolving conflicts, finishing branches, creating PRs. Invoke when user mentions: fleet, swarm, orchestrate, decompose, parallel, worktrees, split work, break into tasks, spawn agents, fleet go, merge, conflicts, finish branch, PR, plan, spec, commit.
Builds code in an isolated worktree — implements features, fixes bugs, writes tests, and submits for review when done. Use for: implementing features, fixing bugs, writing tests, building in worktrees, broadcasting changes, sending/replying to agent messages, submitting for review, writing task summaries. Invoke when user mentions: fleet, build, implement, code, fix, test, worktree, broadcast, message, reply, submit, review, summary, publish, verify.
Splits work into parallel tasks and swarms agents across worktrees. Plans, decomposes, monitors, merges, and ships. Use for: planning features, writing specs, decomposing tasks, splitting issues, running parallel agents, swarming worktrees, monitoring progress, merging branches, resolving conflicts, finishing branches, creating PRs. Invoke when user mentions: fleet, swarm, orchestrate, decompose, parallel, worktrees, split work, break into tasks, spawn agents, fleet go, merge, conflicts, finish branch, PR, plan, spec, commit.
Reviews a branch for quality, tests, error handling, and security. Returns PASS or FAIL with actionable findings. Use for: reviewing code, evaluating diffs, checking tests, checking error handling, checking security, reviewing comments, writing verdicts. Invoke when user mentions: fleet, review, code review, verdict, findings, code quality, test quality, error handling, security, check comments, review-pr.
| name | <skill-name> |
| description | <What thinking mode this activates — verb phrase.> Use for: <comma-separated use cases.> Invoke when user mentions: <trigger keywords.> |
| allowed-tools | <tool permissions> |
| globs | <file patterns> |
<Behavioral instructions — what the agent temporarily becomes when this skill activates. Domain-specific reasoning strategies, constraints, and decision criteria. Keep brief; detailed procedures belong in shortcuts.>
| Command | Purpose |
|---|---|
<cli> <command> |
| Command | Purpose |
|---|
| Command | Purpose |
|---|
| Command | Purpose |
|---|
