菜单
Performance profiling and analysis: identify hot paths, memory leaks, N+1 queries, benchmark before/after. Use when: (1) user reports slowness, (2) optimizing a critical path, (3) benchmarking changes, (4) diagnosing memory issues. NOT for: general debugging (use debug), code review (use code-review).
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
Analyze and document codebase architecture: dependency graphs, module boundaries, API surfaces, tech debt. Use when: (1) user asks 'how is this codebase structured', (2) onboarding to a new project, (3) planning a large feature that spans modules, (4) identifying tech debt or coupling issues. NOT for: specific code changes (use refactor), reviewing a PR (use code-review).
Structured code review for PRs, diffs, or changed files. Use when: (1) reviewing a pull request, (2) checking code quality before merge, (3) auditing changes for correctness/security/performance, (4) user asks to review code or a diff. NOT for: debugging failures (use debug), running tests (use verify), or general code reading.
Manage dependencies: audit outdated packages, resolve conflicts, evaluate alternatives, handle breaking upgrades. Use when: (1) updating dependencies, (2) resolving version conflicts, (3) checking for vulnerabilities in deps, (4) evaluating a new library. NOT for: general code changes, security vulnerabilities in own code (use security-audit).
Generate or update documentation: API docs, README, CHANGELOG, ADRs. Use when: (1) user asks to document code, (2) writing a README for a new project, (3) updating docs after a change, (4) creating architecture decision records. NOT for: inline code comments (just add them), reviewing docs (use code-review).
Advanced git operations: rebase strategies, cherry-pick workflows, bisect automation, worktree management, conflict resolution, branch cleanup. Use when: (1) complex git operations beyond basic add/commit/push, (2) resolving merge conflicts, (3) reorganizing commit history, (4) managing multiple working branches. NOT for: basic git (use git tools directly), GitHub API operations (use github skill).
Language-specific idiom guides for Go, Python, TypeScript, Rust. Covers non-obvious patterns: project layout, error handling, testing idioms, build quirks, common pitfalls. Use when: (1) writing code in a specific language and need idiomatic patterns, (2) user asks 'what's the Go/Python/etc way to do X', (3) porting code between languages. NOT for: syntax questions (the model already knows syntax).
| name | perf-profile |
| description | Performance profiling and analysis: identify hot paths, memory leaks, N+1 queries, benchmark before/after. Use when: (1) user reports slowness, (2) optimizing a critical path, (3) benchmarking changes, (4) diagnosing memory issues. NOT for: general debugging (use debug), code review (use code-review). |
| when_to_use | Use when the user asks about performance, profiling, benchmarks, or optimization. |
| user-invocable | true |
| disable-model-invocation | true |
Measure first, optimize second. Never optimize without data.
Identify the actual bottleneck and quantify the improvement.
go test -bench . -benchmem for microbenchmarksgo tool pprof for CPU and memory profilesruntime/pprof or net/http/pprof for production profilinggo test -trace trace.out for goroutine analysiscProfile / profile for function-level profilingpy-spy for sampling profiler (no code changes)memory_profiler for memory usagetimeit for microbenchmarks--prof flag + --prof-process for V8 profilesclinic.js for automatic profilingconsole.time / performance.now for manual timing