Lighthouse expertise, packaged for your AI agent

What it does

Web Quality Skills is a set of instruction packs for AI coding agents. Each pack encodes rules from Google Lighthouse, Core Web Vitals, and WCAG 2.2 so an agent can audit or optimize a web project for performance, accessibility, SEO, and security. It is stack-agnostic, with tailored patterns for React, Vue, Svelte, Astro, and plain HTML.

The interesting bit

Instead of relying on an LLM’s training data to recall exact CLS thresholds or WCAG 2.2 guidelines, these skills act as structured system prompts. They trigger automatically when you use natural language like “audit my site” or “fix LCP,” effectively turning institutional web-performance knowledge into an agent-native tool.

Key highlights

• Covers 150+ Lighthouse audits across Performance, Accessibility, SEO, and Best Practices.
• Includes concrete thresholds and budgets: LCP ≤ 2.5s, INP ≤ 200ms, total page weight < 1.5 MB, JavaScript < 300 KB compressed.
• Framework-specific guidance for React/Next.js, Vue/Nuxt, SvelteKit, Astro, and static HTML.
• Supports Claude Code, Codex, Gemini CLI, and other agents via the add-skill toolchain.
• Skills auto-activate from natural language triggers; no manual flag needed.

Caveats

• The project is explicitly unofficial and community-maintained.
• These are prompt-based instruction sets, not a Lighthouse runner or automated scanner; the agent must still apply the rules correctly.
• Installation mechanics differ across platforms—some agents need manual copying, others require specific CLI versions.

Verdict

Worth a look if you want your AI assistant to stop hallucinating web-performance advice and start citing actual Lighthouse thresholds. Skip it if you are looking for a drop-in CI scanner or a standalone auditing binary.