What problem does it solve? Performance regressions like N+1 database queries, blocking calls in async handlers, unbounded queries, and memory leaks often slip into production because they are invisible during code review. This Skill acts as a pre-merge performance gate that scans code changes for measurable slowdown patterns and reports them with file:line evidence and estimated impact. ## Core Features & Use Cases - Pattern Detection: Scans diffs or hotpath files for N+1 queries, unbounded queries, SELECT *, sync-in-async calls, missing awaits, event listener leaks, and unbounded caches. - Framework-Specific Checks: Applies targeted rules for React, Node.js/Express, Python/Django, and SQL, plus bundle bloat analysis for frontend projects. - Lighthouse & Core Web Vitals Gates: Enforces PASS/WARN/BLOCK thresholds for Performance, Accessibility, LCP, INP, and CLS when a public URL is available. - AI Token Budget Audit: Detects AI API calls in loops, missing token tracking, model mismatches, and unbounded max_tokens for cost control. - Use Case: Before merging a pull request that touches database access code, run the perf gate to catch an N+1 query in an ORM loop and get a BLOCK verdict with the exact file, line, and estimated latency impact. ## Quick Start Run the perf skill on my current diff and report any performance regressions with file and line evidence.
