What problem does it solve? Performance testing often degrades into running a load tool and eyeballing numbers, with no pass/fail criteria, so regressions slip through CI unnoticed. This Skill enforces a discipline of defining SLO/SLA thresholds first, then measuring against them and failing the build when exceeded. ## Core Features & Use Cases - Seven measurement test types: Covers performance, load, stress, spike, soak (endurance), scalability, and capacity testing, mapped to ISO/IEC 25010 performance efficiency. - Load test procedure and completion checklist: Step-by-step workflow from SLO definition through ramp-up, percentile-based measurement (p50/p95/p99), threshold comparison, and separation of load generator from target. - Performance fix workflow: Guides profiling-first diagnosis, interval narrowing, differential oracle testing against the slow reference implementation, and measurement-driven optimization decisions. - Use Case: You need to verify an API endpoint meets p99 < 200ms under 50 concurrent connections. The Skill provides an autocannon-based TypeScript example that throws and fails CI when latency or error thresholds are exceeded. ## Quick Start Ask the AI to write a load test for your API endpoint with explicit p99 latency and error-rate SLO thresholds that fails CI when exceeded.
