golang-performance

The golang-performance skill provides a structured methodology for optimizing Go applications after performance bottlenecks have been identified through profiling or benchmarking. It focuses on practical performance engineering patterns rather than speculative micro-optimizations, emphasizing a disciplined workflow of measuring, diagnosing, optimizing, and re-measuring. The skill is designed to help developers identify the correct optimization strategy for specific bottlenecks, including CPU inefficiencies, excessive memory allocations, garbage collection overhead, poor memory layout, hot-path inefficiencies, and unnecessary I/O or concurrency costs. It also reinforces the principle that external systems such as databases or APIs should be ruled out as bottlenecks before attempting application-level optimizations.

The skill includes multiple operating modes tailored to different optimization scenarios. Review mode supports architecture-level analysis to detect structural performance anti-patterns such as unbounded goroutines, missing connection pools, inefficient data structures, and caching gaps. A separate hot-path review mode focuses on tight loops or specific functions identified through profiling. Optimize mode guides users through an iterative improvement cycle that prioritizes atomic benchmarking, single-change experimentation, and benchmark comparison using tools such as pprof, fgprof, benchstat, staticcheck, and fieldalignment. The skill also encourages documenting optimization rationale and benchmark results to preserve long-term maintainability.

This skill is intended for Go developers, backend engineers, and AI coding agents working on performance-sensitive services or applications. It is especially useful during profiling-driven optimization work, performance-oriented code reviews, and scalability improvement efforts. The skill is compatible with Claude Code and similar AI coding agents, and is designed for projects using Golang with access to standard Go performance tooling.