What problem does it solve?
Performance regressions slip into production when teams only rely on intuition or one-off checks, and then users notice slow page loads. This Skill measures real page performance, stores a baseline, and flags changes introduced by each PR.
Core Features & Use Cases
- Measure real browser performance: Collects navigation timing and resource data using the browse daemon and in-page JavaScript evaluation for TTFB, FCP, LCP, DOM timing, load time, requests, and transfer size.
- Baseline and compare per PR: Captures benchmark baselines with branch and timestamp context, then compares current results against the saved baseline to detect timing and bundle/resource regressions.
- Make slowdowns actionable: Identifies the slowest resources and produces performance budget and trend-style summaries to guide targeted fixes.
Use case: you are preparing a PR that changes frontend bundles and you want to verify that LCP, load time, request count, and bundle size did not worsen compared to the last known good baseline.
Quick Start
Run benchmark for your site by telling the AI: "Benchmark https://example.com with --baseline before changes, then benchmark again after the changes and report any regressions."