Benchmark (Performance Regression Detection)

What it does

This skill provides an automated performance engineering workflow. It captures baseline metrics (TTFB, FCP, LCP, JS/CSS bundle sizes, request counts), runs targeted audits (via a browse daemon), and compares before/after metrics to detect regressions on pull requests. It generates human-readable reports and recommendations for slow resources and remediation steps.

When to use it

Use this skill as part of CI or code review when you want to ensure changes don't degrade page performance — for PR checks, trend analysis, or ad-hoc audits. Trigger with commands like /benchmark <url> or when reviewing PRs that touch frontend assets.

What's included

• Scripts: workflow and preamble snippets embedded in the skill (shell commands to prepare sessions and run perf checks).
• References: limited in-repo docs and templates.
• Instructions: phases covering setup, page discovery, metrics collection, baseline capture, comparison thresholds, slow resource analysis, and reporting.

Compatible agents

Designed for use with agent environments that can run shell snippets and the browse daemon (ostack/Claude Code contexts). Ideal for engineering-focused agents with access to repo and CI tools.