M1

m10-performance

Assists with measuring performance, locating hotspots, and implementing memory-efficient structures.

Install

mkdir -p .claude/skills/m10-performance && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/9229" && unzip -o skill.zip -d .claude/skills/m10-performance && rm skill.zip

Installs to .claude/skills/m10-performance

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

CRITICAL: Use for performance optimization. Triggers: performance, optimization, benchmark, profiling, flamegraph, criterion, slow, fast, allocation, cache, SIMD, make it faster, 性能优化, 基准测试
189 chars✓ has a “when” trigger
Advanced

Key capabilities

  • Measure performance
  • Identify hotspots
  • Reduce allocations
  • Improve cache locality
  • Benchmark with criterion

How it works

The skill follows a profile-first approach to identify bottlenecks and apply optimizations like zero-copy or pre-allocation.

Inputs & outputs

You give it
Slow code
You get back
Optimized implementation

When to use m10-performance

  • Reducing memory allocations
  • Improving cache locality
  • Setting up benchmarks with criterion
  • Identifying performance bottlenecks

About this skill

Performance Optimization

Layer 2: Design Choices

Core Question

What's the bottleneck, and is optimization worth it?

Before optimizing:

  • Have you measured? (Don't guess)
  • What's the acceptable performance?
  • Will optimization add complexity?

Performance Decision → Implementation

GoalDesign ChoiceImplementation
Reduce allocationsPre-allocate, reusewith_capacity, object pools
Improve cacheContiguous dataVec, SmallVec
ParallelizeData parallelismrayon, threads
Avoid copiesZero-copyReferences, Cow<T>
Reduce indirectionInline datasmallvec, arrays

Thinking Prompt

Before optimizing:

  1. Have you measured?

    • Profile first → flamegraph, perf
    • Benchmark → criterion, cargo bench
    • Identify actual hotspots
  2. What's the priority?

    • Algorithm (10x-1000x improvement)
    • Data structure (2x-10x)
    • Allocation (2x-5x)
    • Cache (1.5x-3x)
  3. What's the trade-off?

    • Complexity vs speed
    • Memory vs CPU
    • Latency vs throughput

Trace Up ↑

To domain constraints (Layer 3):

"How fast does this need to be?"
    ↑ Ask: What's the performance SLA?
    ↑ Check: domain-* (latency requirements)
    ↑ Check: Business requirements (acceptable response time)
QuestionTrace ToAsk
Latency requirementsdomain-*What's acceptable response time?
Throughput needsdomain-*How many requests per second?
Memory constraintsdomain-*What's the memory budget?

Trace Down ↓

To implementation (Layer 1):

"Need to reduce allocations"
    ↓ m01-ownership: Use references, avoid clone
    ↓ m02-resource: Pre-allocate with_capacity

"Need to parallelize"
    ↓ m07-concurrency: Choose rayon or threads
    ↓ m07-concurrency: Consider async for I/O-bound

"Need cache efficiency"
    ↓ Data layout: Prefer Vec over HashMap when possible
    ↓ Access patterns: Sequential over random access

Quick Reference

ToolPurpose
cargo benchMicro-benchmarks
criterionStatistical benchmarks
perf / flamegraphCPU profiling
heaptrackAllocation tracking
valgrind / cachegrindCache analysis

Optimization Priority

1. Algorithm choice     (10x - 1000x)
2. Data structure       (2x - 10x)
3. Allocation reduction (2x - 5x)
4. Cache optimization   (1.5x - 3x)
5. SIMD/Parallelism     (2x - 8x)

Common Techniques

TechniqueWhenHow
Pre-allocationKnown sizeVec::with_capacity(n)
Avoid cloningHot pathsUse references or Cow<T>
Batch operationsMany small opsCollect then process
SmallVecUsually smallsmallvec::SmallVec<[T; N]>
Inline buffersFixed-size dataArrays over Vec

Common Mistakes

MistakeWhy WrongBetter
Optimize without profilingWrong targetProfile first
Benchmark in debug modeMeaninglessAlways --release
Use LinkedListCache unfriendlyVec or VecDeque
Hidden .clone()Unnecessary allocsUse references
Premature optimizationWasted effortMake it work first

Anti-Patterns

Anti-PatternWhy BadBetter
Clone to avoid lifetimesPerformance costProper ownership
Box everythingIndirection costStack when possible
HashMap for small setsOverheadVec with linear search
String concat in loopO(n^2)String::with_capacity or format!

Related Skills

WhenSee
Reducing clonesm01-ownership
Concurrency optionsm07-concurrency
Smart pointer choicem02-resource
Domain requirementsdomain-*

When not to use it

  • Premature optimization
  • Benchmarking in debug mode

Prerequisites

criterionperf

Limitations

  • Requires release mode for benchmarks
  • Profiling adds overhead

How it compares

It prioritizes algorithmic and data structure changes over micro-optimizations.

Compared to similar skills

m10-performance side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
m10-performance (this skill)06moNo flagsAdvanced
debug-lldb17moReviewIntermediate
jit-overview27moNo flagsAdvanced
rspack-sftrace15moReviewAdvanced

Try saying

Example prompts that trigger this skill in your AI assistant.

More by actionbook

View all by actionbook

rust-learner

actionbook

Use when asking about Rust versions or crate info. Keywords: latest version, what's new, changelog, Rust 1.x, Rust release, stable, nightly, crate info, crates.io, lib.rs, docs.rs, API documentation, crate features, dependencies, which crate, what version, Rust edition, edition 2021, edition 2024, cargo add, cargo update, 最新版本, 版本号, 稳定版, 最新, 哪个版本, crate 信息, 文档, 依赖, Rust 版本, 新特性, 有什么特性

83

arxiv-viewer

actionbook

View, search, and download academic papers from arXiv. Supports API queries, web scraping via Actionbook, and HTML paper reading via ar5iv. Use when user asks about arxiv papers, academic papers, research papers, paper summaries, latest papers, or wants to search/download/read papers.

49

rust-call-graph

actionbook

Visualize Rust function call graphs using LSP. Triggers on: /call-graph, call hierarchy, who calls, what calls, 调用图, 调用关系, 谁调用了, 调用了谁

31

rust-code-navigator

actionbook

Navigate Rust code using LSP. Triggers on: /navigate, go to definition, find references, where is defined, 跳转定义, 查找引用, 定义在哪, 谁用了这个

35

rust-daily

actionbook

CRITICAL: Use for Rust news and daily/weekly/monthly reports. Triggers on: rust news, rust daily, rust weekly, TWIR, rust blog, Rust 日报, Rust 周报, Rust 新闻, Rust 动态

33

rust-router

actionbook

CRITICAL: Use for ALL Rust questions including errors, design, and coding. HIGHEST PRIORITY for: 比较, 对比, compare, vs, versus, 区别, difference, 最佳实践, best practice, tokio vs, async-std vs, 比较 tokio, 比较 async, Triggers on: Rust, cargo, rustc, crate, Cargo.toml, 意图分析, 问题分析, 语义分析, analyze intent, question analysis, compile error, borrow error, lifetime error, ownership error, type error, trait error, value moved, cannot borrow, does not live long enough, mismatched types, not satisfied, E0382, E0597, E0277, E0308, E0499, E0502, E0596, async, await, Send, Sync, tokio, concurrency, error handling, 编译错误, compile error, 所有权, ownership, 借用, borrow, 生命周期, lifetime, 类型错误, type error, 异步, async, 并发, concurrency, 错误处理, error handling, 问题, problem, question, 怎么用, how to use, 如何, how to, 为什么, why, 什么是, what is, 帮我写, help me write, 实现, implement, 解释, explain

36

Search skills

Search the agent skills registry