performance-optimizer_skill

Overview

This skill provides practical performance analysis and optimization guidance for code and systems, focused on identifying bottlenecks and delivering measurable improvements. It combines profiling tactics, optimization patterns, and verification steps to reduce resource usage and speed up real workloads. Use it when you need targeted, evidence-based performance work rather than guesswork.

How this skill works

I guide you to establish baseline metrics, run the right profilers for your stack (Python, Node, shell, system), and isolate hot paths using data, not intuition. Once the bottleneck is identified, I recommend targeted optimizations (algorithmic changes, caching, I/O batching, vectorization, parallelism) and provide verification steps—benchmarks, regressions checks, and monitoring. The process emphasizes measurable gains and maintaining correctness.

When to use it

• When a service or script is too slow and you need the root cause
• Before investing time in large refactors or low-level rewrites
• When memory usage grows unexpectedly or leaks are suspected
• To establish performance budgets and SLA targets
• When deploying to production and you need safe profiling and monitoring

Best practices

• Always measure first: profile to find real bottlenecks before changing code
• Fix the highest-impact hotspot first (80/20 rule) and avoid micro-optimizations
• Prefer algorithmic improvements over micro-tweaks; document trade-offs
• Use representative data, warm-up runs, and multiple iterations for benchmarks
• Verify changes with tests, benchmarks, and monitoring to catch regressions

Example use cases

• Speed up a Python data pipeline by replacing O(n²) patterns with set lookups or vectorized NumPy operations
• Reduce I/O latency by batching writes and switching to buffered I/O for large files
• Diagnose high CPU in production using sampling profilers, then apply caching or async processing
• Optimize shell scripts by replacing unnecessary subshells and using xargs for parallel work
• Improve database performance by adding indexes, fixing N+1 queries, and introducing a caching layer