benchmark

Detects web performance regressions by comparing page metrics against captured baselines.

Updated Jun 22, 2026
One-click install
npx skills add https://github.com/aicodepro/ai-agent-nexi --skill benchmark-aicodepro
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: benchmark
Source: https://github.com/aicodepro/ai-agent-nexi/tree/main/agent/skills/gstack/benchmark
Command: npx skills add https://github.com/aicodepro/ai-agent-nexi --skill benchmark-aicodepro

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires bun.

What problem does it solve? Web applications slow down gradually as each pull request adds small delays and bundle bloat, and nobody notices until load times become unacceptable. This Skill measures page performance, captures baselines, and compares metrics across branches so regressions are caught before they ship. ## Core Features & Use Cases - Baseline Capture and Comparison: Record TTFB, FCP, LCP, DOM timings, transfer sizes, and bundle sizes per page, then diff current metrics against the stored baseline with regression thresholds. - Resource and Bundle Analysis: Identify the slowest resources, largest JavaScript and CSS bundles, and total network transfer using the browse daemon's perf command and JavaScript evaluation. - Diff-Aware and Trend Modes: Benchmark only pages affected by the current branch with --diff, or review historical performance trends with --trend. - Use Case: Before merging a PR, run a benchmark against your staging URL to discover that LCP doubled and the JS bundle grew 60%, then investigate the offending dependency before users feel the slowdown. ## Quick Start Ask the assistant to run a performance benchmark on your staging URL with baseline comparison, for example by requesting a benchmark of https://staging.example.com against the saved baseline.

Frequently Asked Questions about benchmark

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I detect performance regressions before merging a pull request?

Capture a baseline of page metrics on the main branch, then run the benchmark on your feature branch. The skill compares TTFB, FCP, LCP, and bundle sizes against the baseline and flags regressions using defined thresholds.

How to measure Core Web Vitals from the command line?

Use the browse daemon's perf command combined with JavaScript evaluation of the Performance API. It extracts navigation timing, paint entries, and resource timing data to compute LCP, FCP, and load metrics.

What counts as a performance regression in benchmark comparisons?

Timing increases over 50% or over 500ms absolute count as regressions, while over 20% is a warning. Bundle size growth over 25% is a regression, and request count growth over 30% triggers a warning.

Can I benchmark only the pages changed in my branch?

Yes, the --diff mode uses git diff against the base branch to identify affected files and benchmarks only the impacted pages. This avoids running full-site audits on every change.

Why does the benchmark setup report NEEDS_SETUP?

The browse daemon binary must be built once before use, which takes about ten seconds via the setup script. It requires bun, which the skill can install with a checksum-verified installer if missing.