What problem does it solve? Performance claims are often made from intuition or a single noisy timing run, leaving teams without repeatable evidence when a regression is suspected or a release needs benchmark proof. This Skill enforces a disciplined procedure that produces executed, reproducible performance evidence tied to a real budget or baseline. ## Core Features & Use Cases - Structured Benchmark Contract: Requires a defined workload, dataset or fixture, baseline or budget, and regression threshold before any result is reported. - Environment and Variance Discipline: Captures machine, OS, runtime versions, warmup, repetitions, variance policy, and outlier handling so results are reproducible. - qa-result Evidence Emission: Outputs a structured evidence record with exact commands, artifacts, residual risk, and release-blocker status. - Use Case: A reviewer suspects a recent change slowed down message processing. Use this Skill to pin the ref, run the benchmark with warmup and repetitions, compare against the predeclared threshold, and emit qa-result evidence a reviewer can inspect. ## Quick Start Ask the agent to run a performance benchmark for the suspected regression in the message store workload and produce qa-result evidence with the exact commands and artifacts.
