What problem does it solve? It prevents wasted optimization effort by enforcing a measurement-first discipline: instead of guessing why an application is slow, it establishes a reproducible baseline, identifies the real bottleneck (algorithm, query, I/O, memory, or CPU), and proves the fix with quantified before/after metrics. ## Core Features & Use Cases - Baseline Measurement: Requires concrete numbers (latency, query counts, memory usage) under a defined load before any code change, using profilers, EXPLAIN PLAN, and pg_stat_statements. - Bottleneck Pattern Library: Covers five recurring bottleneck patterns — algorithmic complexity, N+1 queries, I/O and network blocking, memory allocation, and CPU spin loops — each with cause, proof method, and common pitfalls. - Quantified Validation: Delivers a before/after metrics table (P99 latency, queries per request, peak memory) and documents accepted trade-offs so future maintainers understand why the code changed. - Use Case: When a user reports "the endpoint is slow" or "it times out under load", this Skill guides the investigation from baseline measurement through diagnosis to a validated fix, such as reducing 1000 queries per request to 1. ## Quick Start Use the otimizar-performance skill to diagnose why my endpoint takes 5 seconds and validate the fix with before-and-after metrics.
