What problem does it solve? Go developers often optimize code based on intuition, which is wrong most of the time. This Skill enforces a measure-first discipline: define a metric, write an isolated benchmark, capture a baseline, diagnose with pprof, apply one optimization at a time, and verify gains with statistical rigor before committing. ## Core Features & Use Cases - Benchmark authoring and execution: Write atomic benchmarks with b.Loop(), sub-benchmarks for input sizes, and -benchmem -count=10 runs that record time, bytes, and allocations per operation. - Profile-driven diagnosis: Match pprof signals (alloc_objects, CPU, goroutine, block profiles) to bottleneck types and route to targeted fixes for memory, CPU, I/O, runtime, or caching issues. - Statistical comparison and review: Use benchstat to prove improvements are significant (rejecting ~ results), document numbers in perf(scope): commits, and detect regressions in CI with benchdiff, cob, or gobenchdata. - Use Case: A service's JSON parsing endpoint is slow. Benchmark the parser, capture a CPU and memory profile, discover reflection-driven allocations, switch to a code-generated marshaler, and confirm a 40% reduction with benchstat before merging. ## Quick Start Ask the AI to benchmark a specific Go function, profile it with pprof, and recommend one measured optimization with benchstat verification.
