菜单
Profile Rust code using samply and related evidence to identify CPU bottlenecks, allocation hot paths, and allocator-fragmentation symptoms. Use when performance is slow, RSS grows unexpectedly, before optimizing, or when the user asks to profile.
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
Context pressure checkpoint, self-handoff, and durable lesson capture workflow. Use when an agent is near the context limit, context usage is above 98%, auto-compaction or context warnings appear, the user says high tide, before `/clear`, or after a long confusing/debugging session where lessons learned should be preserved for the next fresh session or written to project-local `docs/.ai/lessons/`.
Generate execution-ready, model-specific prompts for handing work to a different LLM agent or fresh session. Use for shared-workspace delegation, cold-start continuation, cross-model transfer, one-shot worker directives, or requests such as "create a handoff prompt", "delegate this", "hand this off", and "prepare context for another agent". Supports current Anthropic Claude, OpenAI GPT/Codex, Google Gemini, and xAI Grok families.
Surface hidden unknowns before, during, and after implementation using structured techniques — clickable mocks, intervention brainstorms, semantics maps, tweakable plans, implementation notes, buy-in docs, and more. Use PROACTIVELY when entering an unfamiliar codebase or domain, when a feature request is ambiguous or underspecified, when the user says "I'm not sure", "I'll know it when I see it", or has no design direction, when porting from a reference implementation, when writing an implementation plan, before merging a large diff, or when packaging work for reviewer sign-off. Also triggers on explicit asks: "blindspot pass", "interview me", "quiz me before I merge", "show me design directions", "teach me <domain>", "what am I missing?".
Measure and improve Go program performance on modern Go (1.24+). Use when profiling Go code, diagnosing CPU or memory bottlenecks, investigating latency or contention, writing or fixing benchmarks, comparing benchmark results, using pprof or trace data, applying PGO, or tuning hot-path Go code.
Behavior-preserving cleanup for recently changed code. Use after implementing or modifying code, when explicitly invoked as `$code-simplifier`, or when the user asks to simplify, clean up, or refactor the current diff without widening scope.
Decompose large agent-built features into atomic, reviewable chunks using a diff-size budget (~1500 lines) as the signal. Use this whenever implementing a non-trivial or net-new feature with an agent, when an agent has produced a huge diff/PR that is painful to review, when the user says a change is "too big", "too much to review", or "should be broken up", or when planning how to parallelize feature work across multiple agents. Reach for it on any "build this whole feature" request, even when the user does not explicitly ask to decompose — getting under the review-ability threshold is the goal. Skip it for small, self-contained features that will obviously land well under the threshold, and for large-but-mechanical sweeps (renames, codemods, lockfile churn) where the diff is big but trivially reviewable.
| name | rust-profiling |
| description | Profile Rust code using samply and related evidence to identify CPU bottlenecks, allocation hot paths, and allocator-fragmentation symptoms. Use when performance is slow, RSS grows unexpectedly, before optimizing, or when the user asks to profile. |
Profile Rust binaries to find CPU bottlenecks and allocation pressure using samply, then use OS-level memory evidence when RSS or fragmentation is the symptom.
# 1. Ensure profiling profile exists in Cargo.toml (see reference.md)
# 2. Build with debug symbols
cargo build --profile profiling
# 3. Profile (opens Firefox Profiler UI)
samply record ./target/profiling/<binary> [args...]
# 4. Or save for CLI analysis
samply record --save-only -o profile.json ./target/profiling/<binary>
python3 ~/.agents/skills/rust-profiling/scripts/analyze_profile.py profile.json
| File | Purpose |
|---|---|
reference.md | Cargo.toml setup, samply options, troubleshooting |
examples.md | Common profiling scenarios and analysis patterns |
optimization.md | Post-profiling fixes: source patterns, release-profile tuning, PGO, BOLT, what doesn't work |
scripts/analyze_profile.py | CLI tool to analyze saved profile.json files |
| Pattern | Meaning | Action |
|---|---|---|
| High self-time | Function itself is slow | Direct optimization target |
| High total-time | Called often or slow callees | Check call frequency |
malloc/alloc in hot path | Allocation overhead | Preallocate, reuse, pool, or use bounded/shared buffers |
| RSS grows then plateaus while profiles show allocation churn | Possible allocator fragmentation, not necessarily a leak | Compare allocators, then reduce high-rate small allocations |
pthread_mutex/parking_lot | Lock contention | Reduce lock scope or use lock-free |