Audit and Fix Lighthouse Performance Issues in Batches

PageSpeed Enhancer turns a supplied PageSpeed Insights or Lighthouse report into a staged audit, risk assessment, and targeted fix plan across performance, accessibility, best practices, and SEO. Its workflow completes the scan and risk assessment before making changes, then applies fixes in batches and checks the results with a new report. The referenced detailed guide supplies the complete procedure and prerequisites; this overview does not replace it.

The performance work covers Core Web Vitals such as LCP, INP, and CLS alongside lab diagnostics including FCP, TBT, and SI. For an LCP investigation, the workflow examines TTFB, the identified LCP element, its loading strategy, element render delay, and the critical resource chain. Render-blocking CSS, a lazy-loaded hero image, unused JavaScript, and image optimization are examples of issues it can help investigate. Other supported topics include security headers, ARIA issues, and SEO meta tags.

Use it when a report identifies a performance bottleneck or when a site needs an organized response to accessibility, SEO, or loading diagnostics. It can narrow an investigation to LCP or support a broader review across the four Lighthouse pillars. Findings should distinguish an observed cause from a hypothesis and retain the report date, version, and device context.

The skill does not run Lighthouse or PageSpeed tests itself. A user must supply the report or URL and arrange the measurements used to validate changes. Core Web Vitals use field data at the 75th percentile; lab improvements do not establish field performance. A 100 accessibility score is not WCAG certification, and numerical improvement examples are illustrative rather than predicted outcomes.

Inputs include the report, affected project files, loading behavior, and relevant deployment configuration. The output is a prioritized risk report followed by project-specific fix batches and a re-test comparison. Paths and configuration syntax require adaptation to the actual project. Server-level work such as CDN configuration, PHP opcode caching, and database query optimization falls outside the documented scope.

Header recommendations assume control of the deployment platform, with Netlify and Vercel given as examples. CSP, COOP, and Permissions-Policy can disrupt authentication, popups, camera or geolocation access, and embedding, so those paths require testing before enforcement. CSP should progress through report-only tuning against actual resource origins. The sample policy is incomplete: unsafe-inline and a missing reporting destination prevent treating it as a finished defense or working telemetry.

HSTS subdomain and preload settings require domain-wide readiness and can persist beyond rollback. Inline stylesheet onload handlers can conflict with CSP; critical styles should remain synchronous. Image conversion requires installed tools such as cwebp, sharp, or Pillow while preserving source assets. Font redistribution depends on its license, and public audit services or sitemap submissions disclose URLs and require authorization.