Core Web Vitals Audit Claude Skill
This skill audits a website's Core Web Vitals and page-experience signals and produces an actionable performance report. It optionally pulls real-world CrUX field data from Google PageSpeed Insights and always collects browserless lab proxies — render-blocking CSS and JS, page weight, image dimensions and lazy-loading, TTFB, text compression, DOM size, and third-party origins — then recommends fixes tied to the measured data.
Quick Take
Point the skill at a site URL, sitemap, or URL list. It collects field and lab signals, runs fifteen severity-ranked checks mapped to LCP, CLS, and INP, and returns a prioritized report with specific fixes for each vital.
What The Skill Checks
- Real-world field vitals per page — LCP, CLS, INP, and TTFB — pulled from the Chrome UX Report via PageSpeed Insights when an API key is supplied, used as the verdict on whether each page passes.
- Browserless lab proxies that diagnose root causes: render-blocking CSS and JS counts, total and inline scripts, page weight, missing text compression, oversized DOM, request counts, and third-party origins.
- Image hygiene that drives layout shift and slow loads: missing width and height attributes, lazy-loaded LCP hero images, and missing preconnect hints, each mapped to the vital it affects and ranked by severity.
How The Skill Is Packaged
The skill follows the standard Claude Agent Skill structure: a SKILL.md file with YAML frontmatter and workflow instructions, a references/ folder with the full audit check definitions and report template, and a scripts/ folder with Python scripts that collect performance signals and audit them against your budgets. Copy the skill folder into your Claude skills directory and Claude invokes it automatically when a request matches its description.
Skill Files
Every file in the skill is embedded below directly from the Claude-SEO-Skills repository, so you can review exactly what the skill instructs Claude to do before installing it.
SKILL.md
The skill definition: frontmatter, inputs to collect (including an optional PageSpeed Insights API key and performance budgets), the five-step workflow, and the resources the skill relies on.
Audit Checks Reference (references/audit-checks.md)
Full definitions, thresholds, and rationale for all fifteen checks, from poor field metrics and missing compression to lazy-loaded LCP images and missing preconnect hints.
Report Template (references/report-template.md)
The output structure the skill follows, including pages analyzed, median metrics, per-page field vitals, and issues grouped by severity.
Vitals Collector Script (scripts/collect_vitals.py)
Collects browserless lab proxies for each page — TTFB, render-blocking resources, page weight, image metadata, DOM size, and compression — and optionally pulls CrUX field data from the PageSpeed Insights API.
Vitals Auditor Script (scripts/audit_vitals.py)
Runs the audit checks against the collected signal inventory using configurable performance budgets and outputs the structured report data mapped to LCP, CLS, and INP.