How are the benchmarking results formatted?

Results are output as a standardized JSON block between PERF_METRICS_START and PERF_METRICS_END markers, making them easy to parse programmatically.

Why does this skill enforce a 60-second minimum run?

This duration ensures the system reaches a steady state and reduces the impact of transient noise, leading to more reliable and repeatable performance metrics.

What is the purpose of the 10-second warm-up phase?

The warm-up allows caches to populate, JIT compilers to optimize, and the runtime environment to stabilize before the actual measurement phase begins.

When should I use the 30-second run exception?

The 30-second run duration is specifically reserved for binary search phases where rapid iteration is required to narrow down specific performance issues.

Can I run multiple benchmarks in parallel?

No, the skill strictly enforces sequential execution to prevent CPU, memory, and I/O contention that would otherwise skew the results.

Performance Benchmarker

Name: Performance Benchmarker
Author: ComposioHQ

byComposioHQ

•

1,152

•

Analytics & Monitoring

Executes standardized performance benchmarks with strict duration rules and sequential processing to ensure accurate system baselines.

The Performance Benchmarker skill for Claude Code provides a rigorous framework for conducting system and application performance tests. It enforces critical best practices such as mandatory warm-up periods, minimum run durations to eliminate noise, and strictly sequential execution to prevent resource contention. Ideal for establishing performance baselines, identifying regressions, or performing binary search analysis on bottlenecks, this skill ensures that metrics are captured consistently and output in a structured JSON format for automated analysis and reporting.

Key Features

01Strict sequential execution to prevent resource interference and noise

02Standardized JSON metrics output for seamless integration with monitoring tools

03Mandatory 60-second minimum runs to ensure data stability and accuracy

04Integrated 10-second minimum warm-up phase before measurement begins

051,152 GitHub stars

06Automated anomaly detection and re-run enforcement for outlier results

Use Cases

01Validating performance regressions after implementing architectural changes

02Establishing a stable performance baseline for new software releases

03Conducting binary search analysis to identify specific performance bottlenecks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add composiohq/awesome-claude-plugins benchmark

For use in Claude.ai and ChatGPT

Download Skill