web-quality-audit Skill (addyosmani) for Claude: What It Does and How to Install It

Makes Claude audit your site with live Lighthouse and trace evidence, then rank and fix what it finds.

Installs: 34.3K; on skills.sh
Stars: 3K; on GitHub
Files: 1; besides SKILL.md
Updated: Aug 2026; last repo commit

web-quality-audit makes Claude run an evidence-first review of a web page across performance, accessibility, SEO, best practices and agentic browsing. It prefers live measurement (a browser trace and a Lighthouse run) and then looks in the code, and it keeps measured results apart from guesses.

It triggers on requests like "audit my site" or "run lighthouse audit". It reads a guide from the performance skill in the same repo, so install the set. A bundled script, analyze.sh, does a quick static check of HTML files and needs jq.

Good for

Developers shipping a site who want a ranked list of fixes before launch
SEO and performance owners checking Core Web Vitals and Lighthouse scores
Teams that want accessibility and agent-readiness checked together

Install web-quality-audit

Claude Code

Copy this prompt.

Install the web-quality-audit skill in Claude Code. Run this command:

npx skills add addyosmani/web-quality-skills

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

web-quality-audit reads the measurement guide from the performance skill and links to the performance, core-web-vitals, accessibility, seo and best-practices skills in the same repo, so the README command installs all 6. The skill works best with Chrome DevTools MCP but falls back to Lighthouse CLI, per the README. The bundled analyze.sh needs jq.

From the project README. Source.

What's in the skill

When it loads

Run an evidence-led web quality audit covering performance, accessibility, SEO, best practices, and agentic browsing. Use when asked to "audit my site", "review web quality", "run lighthouse audit", "check page quality", or "optimize my website".

Sections in SKILL.md

How it works
Tool routing
Audit categories
Performance
Accessibility
SEO
Best practices
Agentic browsing
Severity levels
Audit output format
Audit results
Evidence
Critical issues (X found)
High priority (X found)
Summary
Recommended priority

Key facts

Author: addyosmani (the repo owner); the README calls the collection unofficial
Repository: addyosmani/web-quality-skills
Installs on skills.sh: 34,264 as of 2026-11-02
GitHub stars: 3,022 for the whole addyosmani/web-quality-skills repo, 2026-11-02
License: MIT (SKILL.md frontmatter)
Last commit to the repo: 2026-08-24
What is inside: 2 files including SKILL.md: scripts/analyze.sh
Version: 2.0 (SKILL.md metadata)
Part of: web-quality-skills: 6 skills in the repo (web-quality-audit, performance, core-web-vitals, accessibility, seo, best-practices)
Optional: Chrome DevTools MCP for live traces and Lighthouse; otherwise Lighthouse CLI and PageSpeed Insights
skills.sh audits: Gen Agent Trust Hub pass, Socket pass, Snyk pass

What it includes

Five steps from target to re-test
Pick URLs and journeys, collect a live baseline, use runtime failures to find source code, rank by impact and confidence, then re-run checks and report what is verified.
Read · How it works
Best available tool, with fallbacks
Browser trace for performance, Lighthouse for accessibility, SEO, best practices and agentic browsing, the accessibility tree for rendered semantics, and analyze.sh as a source smoke test.
Try: “Audit https://example.com and tell me what to fix first.”
Runs code · Tool routing
Static check of HTML files
Scans a file or folder for a missing doctype, charset, viewport, lang, title, images without alt, and http:// URLs. Prints JSON, caps findings at 100, and needs jq.
Runs code · scripts/analyze.sh
Performance, Accessibility, SEO, Best practices, Agentic browsing
Core Web Vitals targets of LCP under 2.5s, INP under 200ms and CLS under 0.1, plus crawlability, security headers and WebMCP review.
Read · Audit categories
Critical, High, Medium, Low
Each level has an action, from fix immediately to fix when convenient.
Read · Severity levels
Evidence table, findings with impact and fix, verification
Results are split into evidence, issues by severity, a summary per category, a recommended order and what was re-verified.
Read · Audit output format
Before deploy, weekly, monthly
Short checklists for each cadence.
Read · Quick checklists

From the official docs

Permissions

Low risk

Claude runs browser traces or Lighthouse against the pages you name and may run a shell script over your HTML. Run it on sites you own and keep authenticated pages in snapshot mode.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Navigation audits reload the page
The SKILL.md says Chrome DevTools MCP navigation mode reloads the page. Use snapshot mode to keep a logged-in or user-created state. github.com
With Chrome DevTools MCP it can run scripts in the audited page
The tool routing table names take_snapshot and focused evaluate_script for checking rendered semantics. Point it at pages you own, and use snapshot mode on logged-in pages. github.com
It can change your code
The workflow ends with making or recommending specific fixes and re-running checks. Review the diff before you commit. github.com
analyze.sh needs jq and writes nothing to disk
The script body reads files with grep and find and prints JSON. It stops with an error if jq is missing. github.com

Technical details

Turning on read-only	No read-only switch is documented in SKILL.md.
Talks to	The audited site, and PageSpeed Insights or CrUX when those are used
Scopes	Runs Lighthouse or DevTools traces against the URLs you give it scripts/analyze.sh reads HTML files in the path you pass

Questions

What does web-quality-audit do?

It makes Claude audit a site for performance, accessibility, SEO, best practices and agentic browsing, using live measurement where it can and your source code to find causes.

How do I install web-quality-audit?

Use the install box at the top of this page. It installs the whole web-quality-skills set, because this skill reads a guide from the performance skill and links to the others.

Do I need Chrome DevTools MCP?

No. The README says the integration is optional. Without it the skills fall back to the Lighthouse CLI, PageSpeed Insights and CrUX tools, manual browser checks and static inspection.

Is web-quality-audit safe to use?

It runs Lighthouse or trace tools against pages you name and can run a shell script over your HTML files. Claude may also make code changes, so review the diff. The SKILL.md warns that navigation-mode audits reload the page.

Does the audit score equal quality?

The SKILL.md says it does not treat an aggregate score as proof of quality. It separates measured findings from hypotheses found only in code.

Is it an official Lighthouse or Google skill?

No. The README describes the collection as unofficial. It is written for Google Lighthouse and Chrome DevTools tooling.

web-quality-audit vs the performance skill?

The audit is the umbrella that covers all five categories and sends you to the others. The performance skill goes deeper on speed, and the audit reads its measurement guide.

Information

updated Nov 2, 2026

Repository: addyosmani/web-quality-skills
Folder: skills/web-quality-audit

addyosmani/web-quality-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 4

addyosmani/web-quality-skills README · github.com
web-quality-audit SKILL.md · github.com
analyze.sh · github.com
performance MEASUREMENT.md · github.com