Agent skill · vasilyu1983
software-performance
Systematic profiling, load testing, performance budgets, and regression prevention. Use when diagnosing slow services, benchmarking systems, or preventing regressions.
What it needs
About 8k tokens when loaded.
What this skill does
Performance Engineering Use this skill for systematic performance work across APIs, web apps, services, and release gates. It owns profiling, load and stress testing, performance budgets, and regression detection. It does not replace SQL tuning, observability setup, or system-architecture design. Quick Reference Task Use ------ ----- Profiling hot paths flamegraph-first profiling with the right language toolchain Load and stress testing k6 by default; alternate tools only when the team has a clear reason Frontend or backend budgets CI thresholds plus representative benchmarks Database-related slowdown query and pool investigation first, then broader system review Continuous regression detection benchmark suites plus statistical comparison CWV thresholds, k6 patterns, Lighthouse CI budget JSON, profiling tools references/perf-budgets-and-cwv.md LCP/INP/CLS, pprof, dotnet-counters, Instruments assert CWV budgets from Lighthouse JSON scripts/checkperfbudget.py exits non-zero on breach When to Use Diagnose slow API responses, page loads, or jobs. Design and run load, stress, soak, or spike tests. Set and enforce performance budgets. Find CPU, memory, or I/O bottlenecks. Plan capacity or benchmark architectural changes. Catch regressions before they reach production. Route Elsewhere SQL query tuning and indexing: use data-sql-optimization. System architecture and scaling design: use software-architecture-design. Frontend build-tool setup: use software-frontend. Observability and telemetry setup: use qa-observability. Resilience patterns such as retries and breakers: use qa-resilience. Mobile-specific performance: use software-mobile. Defaults Measure before you optimize. Flamegraphs first for CPU questions. Benchmark before and after every claimed fix. Prefer realistic load patterns over synthetic vanity numbers. Gate regressions in CI where practical. Treat tool comparisons and best-tool claims as time-sensitive. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills vasilyu1983/software-performance