Website Doctor
Website audit for SEO, security, Core Web Vitals & AI visibility — measured against TODAY's official standards, with safe fixes.
Agent for Claude Code
Token note: agents fetch official sources live and measure for real — that uses more tokens than a simple prompt. The payoff is evidence-backed instead of guessed results. Execution runs locally with your own Claude.
Website audit tool for SEO, security headers, Core Web Vitals, GDPR and AI visibility (GEO). Measures your site or local project for real and checks every rule live against the current official source — then delivers a prioritized, evidence-backed report and safe fixes.
What this skill does
- Measures real signals: HTTP headers, redirect/status chain, raw HTML (what crawlers see), size/timing
- Live methodology: fetches the official source before each category and cites it with a retrieval date
- Security headers against OWASP (HSTS, CSP, nosniff, Referrer/Permissions-Policy)
- Performance proxies against Core Web Vitals (LCP/INP/CLS) — honestly labeled as lab
- SEO basics: title/description/canonical/robots, single H1, sitemap, status codes
- Structured data (Schema.org / JSON-LD) checked and suggested
- AI visibility/GEO: AI crawlers in robots.txt, content in raw HTML, answer-first, llms.txt
- Mobile (viewport), GDPR (third-party resources/consent), accessibility basics (lang, alt, landmarks)
- Prioritized report (critical/warning/info) with evidence + sources dossier
- Local project: fix plan + safe fixes after confirmation, with a re-check
Description
Website Doctor is a skill for website audits — it runs a full SEO, security, performance and AI-visibility check on your website or local project and verifies every rule live against today's official source (web.dev, OWASP, schema.org, WCAG) before proposing safe fixes. Website Doctor is a Claude Code agent (plugin) that does two things differently from a typical prompt: it MEASURES instead of guessing (real HTTP headers, status codes, raw HTML, sizes/timing) and it judges against the CURRENT official standard instead of training knowledge — before each category it fetches the official source live (Core Web Vitals on web.dev incl. INP instead of FID, OWASP Secure Headers, schema.org, WCAG 2.2, the current AI crawler lists) and cites source + retrieval date. It checks security headers, performance proxies (honestly labeled as lab), SEO basics, structured data, AI visibility/GEO, mobile, GDPR and accessibility basics. The result is a report sorted by severity and leverage with evidence and a label per finding (measured / checked against source (date) / estimated) plus a sources dossier. For local projects the audit also produces a fix plan; only safe fixes are applied (headers, robots.txt, llms.txt, JSON-LD, meta/viewport/canonical/lang) and only after explicit confirmation, one change at a time, with a re-check. Runs entirely locally in your Claude Code — no external runtime.
Examples
What it does not do
- Won't invent measurements — honestly flags what can't be measured/verified
- Without a field source it gives no real field data (CrUX); performance is lab/proxy
- Never changes files unprompted — fixes are local only, after explicit confirmation
- Won't tighten CSP/redirects without asking; deletes/overwrites nothing blindly
- Doesn't replace a full manual accessibility/pentest audit
Compatibility & tech
- Tested (internal)
- 4 scenarios
- Recommended runtime
- Claude Code with web fetch enabled; strongest available model for the audit run.
- Modes
- Audit (read-only, measured & backed by official sources) · Fix (safe fixes only, after explicit confirmation)
- Inputs
- Website URL · Local project path
- Output format
- Markdown report (findings by severity + evidence + labels), sources dossier, optional fix plan
- Subcategory
- Website audit / SEO / GEO / security
- License
- Proprietär
Security profile
Runs entirely on your machine with your own AI — no external runtime, no running costs.
Contains executable code or runs actions/tools — take a quick look before using.
Accesses external sources / the network in normal use (e.g. live pages, search/data APIs).
What you get
- website-doctor-1.1.0/7 files
- .claude-plugin/marketplace.json
plugins/website-doctor/7 files
- README.md
.claude-plugin/1 file
- plugin.json
agents/1 file
- website-doctor.md
commands/1 file
- website-doctor.md
skills/3 files
website-doctor-audit/2 files
- SKILL.md
references/1 file
- sources.md
website-doctor-fix/1 file
- SKILL.md
- .agents/skills/website-doctor/→ universal — same content (Codex, Cursor, Copilot, Gemini, Windsurf, Cline)
- wrappers/copilot/
- wrappers/gemini/
- wrappers/cursor/
- wrappers/codex/
- wrappers/windsurf/
- wrappers/cline/
- LICENSE.txt
Installation
Reviews
No reviews yet – be the first.
Note
Agent for Claude Code (CLI & desktop plugin). Runs locally with your own Claude. Token note: agents fetch official sources live and measure for real — this uses more tokens than a simple prompt; the payoff is evidence-backed instead of guessed results.
Changelog
- v1.1.022.06.2026Platform-native output: now actively uses your platform's native formatting — tables, checklists, code blocks and clickable/button options, with a numbered fallback on non-interactive platforms. Plus internal metadata that prepares the upcoming agent builder.
- v1.0.013.06.2026First agent: Website Doctor as a Claude Code plugin (command + audit subagent + audit/fix skills). Live methodology against official sources, measured signals, safe fixes behind a human gate.
Frequently asked questions
What does Website Doctor do?
Website audit tool for SEO, security headers, Core Web Vitals, GDPR and AI visibility (GEO). Measures your site or local project for real and checks every rule live against the current official source — then delivers a prioritized, evidence-backed report and safe fixes.
How do I run a website audit for security, SEO and performance?
Website Doctor really measures your website or local project and checks every point live against the current official source. You get a prioritized, evidence-backed report covering security, SEO, Core Web Vitals performance, schema, mobile, GDPR and AI visibility.
How do I fix website issues like missing security headers or meta tags safely?
Website Doctor only proposes safe fixes — security headers, robots.txt, llms.txt, JSON-LD, meta and viewport tags — and applies them after your confirmation, never on its own. That keeps you in control and lets you follow every step.
Are the website audit results current or outdated?
Website Doctor checks every rule and threshold live against today's official source (web.dev, OWASP, schema.org, WCAG, AI crawler lists) and cites the source plus retrieval date. So the report reflects the current state, not outdated knowledge.
Will my website be found by AI search engines like ChatGPT and Perplexity?
Website Doctor also checks AI visibility (GEO): whether AI crawlers get in, whether your content is in the raw HTML, and whether there are citable answers and structured data. You can see how well AI answer engines can read and cite your site.
Which AI tools does Website Doctor work with?
Claude Code (CLI) · Claude Code (desktop plugin) · Requires file, Bash and web-fetch tools
How do I use Website Doctor?
Website Doctor is an agent (Claude Code plugin): download the ZIP, set it up in Claude Code via '/plugin marketplace add' + '/plugin install' and run it as a slash command. It runs locally with your own Claude. The bundled skills also run natively in Cursor, GitHub Copilot, Gemini CLI, Windsurf & Cline.