What problem does it solve? Teams often waste effort optimizing code that is not the real bottleneck, or ship changes that never measurably improve performance. This Skill enforces a measure-first discipline: profile the actual hot path, record a baseline, change one thing at a time, and re-measure before keeping any optimization. ## Core Features & Use Cases - Profile-Guided Hot Path Detection: Reproduce a representative workload and use CPU, allocation, I/O, or query-time profiling to find where time actually goes instead of guessing. - Baseline-Verified Change Loop: Record a baseline number, hypothesize the cause, change one variable, and re-measure against the baseline, reverting anything that does not move the metric. - Correctness and Audit Trail: Keep tests green, add tests for behavior the optimization could affect, and record baseline and post-change numbers in the commit, PR, or task. - Use Case: A user asks why an API endpoint is slow. The Skill guides profiling to discover 70% of wall time in N+1 queries, fixes the query shape, and verifies the latency drop against the recorded baseline. ## Quick Start Ask the agent to profile this slow endpoint and optimize the real hot path using the performance-optimization skill.
