What problem does it solve? Go services often suffer from hidden bottlenecks — excessive allocations, GC pauses, cache misses, lock contention, or external I/O waits — and intuition about the cause is wrong most of the time. This Skill provides a disciplined, profile-first methodology that maps each identified bottleneck to the correct optimization pattern, preventing wasted effort on micro-optimizations that don't address the real problem. ## Core Features & Use Cases - Decision-tree diagnosis: Maps pprof and fgprof signals (alloc_objects, CPU profile, goroutine blocks) to the right optimization domain — memory, CPU, runtime tuning, I/O, or caching. - Iterative optimization workflow: Enforces a measure-baseline, change-one-thing, benchstat-compare cycle with statistical significance checks and audit-trail report files. - Deep-dive references: Covers allocation reduction, sync.Pool patterns, struct alignment, cache locality, false sharing, SIMD, HTTP transport tuning, JSON performance, and GC tuning with GOMEMLIMIT. - Use Case: Your Go API has p99 latency of 2s but only 5% CPU usage. The Skill directs you to fgprof for off-CPU analysis, identifies blocked goroutines on database calls, and guides connection pool tuning instead of pointless CPU micro-optimization. ## Quick Start Ask the agent to review your Go package for performance anti-patterns or to optimize a specific function that profiling has identified as a bottleneck.
