What this Skill helps with

Core Web Vitals Optimizer interprets LCP, INP, and CLS data and turns the supplied evidence into a prioritized technical fix plan. It can suggest code, image, font, caching, and server changes while keeping measured data separate from estimated impact.

Who this Skill is for

Use it for developers, technical SEO teams, site owners, and agencies reviewing Lighthouse, PageSpeed Insights, Chrome UX Report, or real-user monitoring data. It is most useful when page templates and metric details are available.

What you provide

  • PageSpeed, Lighthouse, CrUX, or real-user measurements
  • The affected URL and page template
  • Device and network context
  • Relevant HTML, CSS, JavaScript, image, and font details
  • Hosting, CDN, caching, and framework information
  • Recent releases that may have changed performance

What it produces

  • Metric status and evidence gaps
  • Likely causes grouped by LCP, INP, and CLS
  • Fixes ordered by impact, effort, and confidence
  • Code or configuration examples where evidence supports them
  • A measurement plan for before-and-after comparison
  • Risks and rollback notes

How the Skill works

The workflow separates field data from lab data, checks each metric against the supplied context, and maps symptoms to likely causes. It orders fixes that can be tested independently and labels projected gains as estimates. The final plan includes a repeatable measurement window instead of treating one Lighthouse run as proof.

Installation and first use

  1. Download and add the Skill to Claude.
  2. Attach the measurement report and identify the page template.
  3. Provide the relevant code or implementation details.
  4. Ask for diagnosis before requesting code changes.
  5. Test one change at a time and retain a rollback path.

Example workflow and expected output

Provide mobile PageSpeed data showing a slow LCP image and layout movement from an ad slot. The expected output separates lab and field evidence, proposes image loading and reserved-space changes, and supplies a measurement plan. It should not promise a specific score without a real deployment test.

Limits and review guidance

Generated code may not match the site's framework, build process, consent rules, or CDN. Test changes in staging, watch for visual regressions, and compare field data over a suitable period. FID is an older metric and should not replace current INP review. Score projections remain estimates until measured on the deployed site.

Frequently asked questions

Does a better Lighthouse score mean Core Web Vitals passed?

Not necessarily. Lighthouse is a lab test, while the public assessment commonly uses field data collected from real visits.

Can the Skill edit my website automatically?

No. It proposes changes and examples. A developer must adapt, test, approve, and deploy them.

Should I fix LCP, INP, or CLS first?

Start with failed metrics that affect important templates, then consider user impact, confidence, effort, and the ability to measure the change.