Agent Skills for optimizing web quality based on Lighthouse and Core Web Vitals.
Shell
2,829
34 commits
updated Aug 24, 2026
An (unofficial) measurement-first collection of Agent Skills for optimizing web projects with Google Lighthouse, Chrome DevTools for agents, Core Web Vitals, WCAG, and search guidance.
Stack-agnostic. Works with any framework: React, Vue, Angular, Svelte, Next.js, Nuxt, Astro, plain HTML, and more.
While interface guidelines tell you what to build, Web Quality Skills tell you how to build it performantly, accessibly, and optimally for search engines. These skills combine:
| Skill | Description | Use when |
|---|---|---|
| web-quality-audit | Evidence-led review across Lighthouse and source categories | "Audit my site", "Review this for quality", "Check web quality" |
| performance | Field/lab measurement, traces, loading and runtime fixes | "Optimize performance", "Speed up my site", "Fix slow loading" |
| core-web-vitals | Measured LCP, INP, and CLS diagnosis and optimization | "Improve Core Web Vitals", "Fix LCP", "Reduce CLS" |
| accessibility | WCAG compliance, screen reader support, keyboard navigation | "Improve accessibility", "WCAG audit", "a11y review" |
| seo | Search engine optimization, crawlability, structured data | "Optimize for SEO", "Improve search ranking", "Fix meta tags" |
| best-practices | Security, modern APIs, code quality patterns | "Apply best practices", "Security audit", "Code quality review" |
The performance skills keep four evidence types explicit:
| Evidence | Answers |
|---|---|
| CrUX field data | How eligible real Chrome users experienced the URL or origin over the recent aggregation window |
| First-party RUM | How the site's own users experience routes, releases, devices, and interactions |
| DevTools trace or Lighthouse lab run | What happened in one controlled browser session and why |
| Static source inspection | What might be a problem when the page cannot run |
Field data determines user impact. Lab traces reproduce and diagnose it. A PerformanceObserver snippet run in one browser is a lab observation, not real-user data by itself.
add-skill is a powerful CLI tool that lets you install agent skills onto your coding agents from git repositories. Whether you're using OpenCode, Claude Code, Codex, or Cursor, the add-skill tool makes it simple to extend your agent's capabilities with specialized instruction sets. Use add-skill to automate release notes, create pull requests, integrate with external tools, and more. Simply run npx add-skill to get started.
npx skills add addyosmani/web-quality-skills
or
npx add-skill addyosmani/web-quality-skills
Or manually:
cp -r skills/* ~/.claude/skills/
Install as a versioned, namespaced plugin from inside Claude Code:
/plugin marketplace add addyosmani/web-quality-skills
/plugin install web-quality-skills@addy-web-quality-skills
Skills are then namespaced (e.g. /web-quality-skills:performance) and update with /plugin update. The plugin reads the same skills/ directory as the manual copy above — no duplication.
Install directly via the Codex plugin marketplace (Codex CLI v0.122+):
codex plugin marketplace add addyosmani/web-quality-skills
Once installed, invoke skills in chat using @ (e.g. @performance, @accessibility). See docs/codex-setup.md for local installation and troubleshooting.
Install directly via Gemini CLI extensions:
gemini extensions install https://github.com/addyosmani/web-quality-skills
Skills are auto-discovered by Gemini and activate when prompts match their description. See docs/gemini-setup.md for workspace mode and troubleshooting.
Add skills to your project knowledge or paste the SKILL.md contents into your conversation.
Skills activate automatically when your request matches their description. Examples:
Audit this page for web quality issues
Optimize performance and fix Core Web Vitals
Review accessibility and suggest improvements
Make this SEO-ready
When the agent exposes equivalent live-browser capabilities, the skills prefer them. With Chrome DevTools MCP, the current tool names are:
performance_start_trace and focused Performance Insights for performance work, including CrUX context when eligible field data existslighthouse_audit for Accessibility, SEO, Best Practices, and Agentic BrowsingThis integration is optional. Without it, the skills fall back to Lighthouse CLI, PageSpeed Insights/CrUX tools, manual browser checks, and static inspection rather than stopping the task.
The comprehensive skill that orchestrates all other skills. It starts with live evidence when a page is runnable and keeps measured failures separate from source-code hypotheses.
Trigger phrases: "audit my site", "quality review", "lighthouse audit", "check web quality"
What it checks:
llms.txt checksDeep-dive into loading and runtime performance optimization, using CrUX or first-party RUM to prioritize user impact and DevTools traces to diagnose causes.
Trigger phrases: "speed up", "optimize performance", "reduce load time", "fix slow"
Key optimizations:
Specialized skill for the three Core Web Vitals that describe loading, responsiveness, and visual stability. Google uses Core Web Vitals in its page-experience systems, but a passing score does not promise a ranking change.
Trigger phrases: "Core Web Vitals", "LCP", "INP", "CLS", "page experience"
Metrics covered:
Comprehensive accessibility audit following WCAG 2.2 guidelines.
Trigger phrases: "accessibility", "a11y", "WCAG", "screen reader", "keyboard navigation"
Categories:
Search engine optimization for better visibility and ranking.
Trigger phrases: "SEO", "search optimization", "meta tags", "structured data", "sitemap"
What it covers:
Modern web development standards and security practices.
Trigger phrases: "best practices", "security audit", "modern standards", "code quality"
Areas covered:
| Metric | Good | Needs improvement | Poor |
|---|---|---|---|
| LCP | ≤ 2.5s | 2.5s – 4.0s | > 4.0s |
| INP | ≤ 200ms | 200ms – 500ms | > 500ms |
| CLS | ≤ 0.1 | 0.1 – 0.25 | > 0.25 |
Field Core Web Vitals are assessed at the 75th percentile; these thresholds are not a claim about a single lab run.
These are initial guardrails, not universal thresholds. Set budgets from the product's target devices, networks, page types, and user journeys.
| Resource type | Budget |
|---|---|
| Total page weight | < 1.5 MB |
| JavaScript | < 300 KB (compressed) |
| CSS | < 100 KB (compressed) |
| Images | < 500 KB total above-fold |
| Fonts | < 100 KB |
| Third-party | < 200 KB |
Scores help catch regressions but do not prove accessibility, security, SEO ranking, or user experience. Preserve project-specific targets and compare runs made with the same tool version and conditions.
| Category | Example guardrail |
|---|---|
| Performance | ≥ 90 |
| Accessibility | Aim for 100 automated coverage; complete manual checks |
| Best Practices | ≥ 95 |
| SEO | ≥ 95 |
These skills are framework-agnostic, but some common patterns:
React/Next.js: Use next/image, React.lazy(), Suspense, useCallback/useMemo for INP
Vue/Nuxt: Use nuxt/image, async components, v-once, computed properties
Svelte/SvelteKit: Use {#await}, svelte:image, reactive statements
Astro: Use <Image>, partial hydration, view transitions
Static HTML: Use native lazy loading, <picture>, preconnect hints
Contributions welcome! Please follow the Agent Skills specification.
skills/{skill-name}/SKILL.mdreferences/ for details)MIT License - see LICENSE for details.
Built with insights from the Chrome DevTools team, web performance experts, and accessibility advocates to help developers create high-quality web experiences.
Shell
100.0%
Agent Skills for optimizing web quality based on Lighthouse and Core Web Vitals.
Shell
2,829
34 commits
updated Aug 24, 2026
An (unofficial) measurement-first collection of Agent Skills for optimizing web projects with Google Lighthouse, Chrome DevTools for agents, Core Web Vitals, WCAG, and search guidance.
Stack-agnostic. Works with any framework: React, Vue, Angular, Svelte, Next.js, Nuxt, Astro, plain HTML, and more.
While interface guidelines tell you what to build, Web Quality Skills tell you how to build it performantly, accessibly, and optimally for search engines. These skills combine:
| Skill | Description | Use when |
|---|---|---|
| web-quality-audit | Evidence-led review across Lighthouse and source categories | "Audit my site", "Review this for quality", "Check web quality" |
| performance | Field/lab measurement, traces, loading and runtime fixes | "Optimize performance", "Speed up my site", "Fix slow loading" |
| core-web-vitals | Measured LCP, INP, and CLS diagnosis and optimization | "Improve Core Web Vitals", "Fix LCP", "Reduce CLS" |
| accessibility | WCAG compliance, screen reader support, keyboard navigation | "Improve accessibility", "WCAG audit", "a11y review" |
| seo | Search engine optimization, crawlability, structured data | "Optimize for SEO", "Improve search ranking", "Fix meta tags" |
| best-practices | Security, modern APIs, code quality patterns | "Apply best practices", "Security audit", "Code quality review" |
The performance skills keep four evidence types explicit:
| Evidence | Answers |
|---|---|
| CrUX field data | How eligible real Chrome users experienced the URL or origin over the recent aggregation window |
| First-party RUM | How the site's own users experience routes, releases, devices, and interactions |
| DevTools trace or Lighthouse lab run | What happened in one controlled browser session and why |
| Static source inspection | What might be a problem when the page cannot run |
Field data determines user impact. Lab traces reproduce and diagnose it. A PerformanceObserver snippet run in one browser is a lab observation, not real-user data by itself.
add-skill is a powerful CLI tool that lets you install agent skills onto your coding agents from git repositories. Whether you're using OpenCode, Claude Code, Codex, or Cursor, the add-skill tool makes it simple to extend your agent's capabilities with specialized instruction sets. Use add-skill to automate release notes, create pull requests, integrate with external tools, and more. Simply run npx add-skill to get started.
npx skills add addyosmani/web-quality-skills
or
npx add-skill addyosmani/web-quality-skills
Or manually:
cp -r skills/* ~/.claude/skills/
Install as a versioned, namespaced plugin from inside Claude Code:
/plugin marketplace add addyosmani/web-quality-skills
/plugin install web-quality-skills@addy-web-quality-skills
Skills are then namespaced (e.g. /web-quality-skills:performance) and update with /plugin update. The plugin reads the same skills/ directory as the manual copy above — no duplication.
Install directly via the Codex plugin marketplace (Codex CLI v0.122+):
codex plugin marketplace add addyosmani/web-quality-skills
Once installed, invoke skills in chat using @ (e.g. @performance, @accessibility). See docs/codex-setup.md for local installation and troubleshooting.
Install directly via Gemini CLI extensions:
gemini extensions install https://github.com/addyosmani/web-quality-skills
Skills are auto-discovered by Gemini and activate when prompts match their description. See docs/gemini-setup.md for workspace mode and troubleshooting.
Add skills to your project knowledge or paste the SKILL.md contents into your conversation.
Skills activate automatically when your request matches their description. Examples:
Audit this page for web quality issues
Optimize performance and fix Core Web Vitals
Review accessibility and suggest improvements
Make this SEO-ready
When the agent exposes equivalent live-browser capabilities, the skills prefer them. With Chrome DevTools MCP, the current tool names are:
performance_start_trace and focused Performance Insights for performance work, including CrUX context when eligible field data existslighthouse_audit for Accessibility, SEO, Best Practices, and Agentic BrowsingThis integration is optional. Without it, the skills fall back to Lighthouse CLI, PageSpeed Insights/CrUX tools, manual browser checks, and static inspection rather than stopping the task.
The comprehensive skill that orchestrates all other skills. It starts with live evidence when a page is runnable and keeps measured failures separate from source-code hypotheses.
Trigger phrases: "audit my site", "quality review", "lighthouse audit", "check web quality"
What it checks:
llms.txt checksDeep-dive into loading and runtime performance optimization, using CrUX or first-party RUM to prioritize user impact and DevTools traces to diagnose causes.
Trigger phrases: "speed up", "optimize performance", "reduce load time", "fix slow"
Key optimizations:
Specialized skill for the three Core Web Vitals that describe loading, responsiveness, and visual stability. Google uses Core Web Vitals in its page-experience systems, but a passing score does not promise a ranking change.
Trigger phrases: "Core Web Vitals", "LCP", "INP", "CLS", "page experience"
Metrics covered:
Comprehensive accessibility audit following WCAG 2.2 guidelines.
Trigger phrases: "accessibility", "a11y", "WCAG", "screen reader", "keyboard navigation"
Categories:
Search engine optimization for better visibility and ranking.
Trigger phrases: "SEO", "search optimization", "meta tags", "structured data", "sitemap"
What it covers:
Modern web development standards and security practices.
Trigger phrases: "best practices", "security audit", "modern standards", "code quality"
Areas covered:
| Metric | Good | Needs improvement | Poor |
|---|---|---|---|
| LCP | ≤ 2.5s | 2.5s – 4.0s | > 4.0s |
| INP | ≤ 200ms | 200ms – 500ms | > 500ms |
| CLS | ≤ 0.1 | 0.1 – 0.25 | > 0.25 |
Field Core Web Vitals are assessed at the 75th percentile; these thresholds are not a claim about a single lab run.
These are initial guardrails, not universal thresholds. Set budgets from the product's target devices, networks, page types, and user journeys.
| Resource type | Budget |
|---|---|
| Total page weight | < 1.5 MB |
| JavaScript | < 300 KB (compressed) |
| CSS | < 100 KB (compressed) |
| Images | < 500 KB total above-fold |
| Fonts | < 100 KB |
| Third-party | < 200 KB |
Scores help catch regressions but do not prove accessibility, security, SEO ranking, or user experience. Preserve project-specific targets and compare runs made with the same tool version and conditions.
| Category | Example guardrail |
|---|---|
| Performance | ≥ 90 |
| Accessibility | Aim for 100 automated coverage; complete manual checks |
| Best Practices | ≥ 95 |
| SEO | ≥ 95 |
These skills are framework-agnostic, but some common patterns:
React/Next.js: Use next/image, React.lazy(), Suspense, useCallback/useMemo for INP
Vue/Nuxt: Use nuxt/image, async components, v-once, computed properties
Svelte/SvelteKit: Use {#await}, svelte:image, reactive statements
Astro: Use <Image>, partial hydration, view transitions
Static HTML: Use native lazy loading, <picture>, preconnect hints
Contributions welcome! Please follow the Agent Skills specification.
skills/{skill-name}/SKILL.mdreferences/ for details)MIT License - see LICENSE for details.
Built with insights from the Chrome DevTools team, web performance experts, and accessibility advocates to help developers create high-quality web experiences.
Shell
100.0%