angeo skills

Agent Skills for AEO — checking whether AI assistants and their crawlers can reach, parse, and cite a website.

Maintained by angeo.dev. MIT licensed.

Install

/plugin marketplace add angeo-dev/skills
/plugin install aeo-audit@angeo

Then ask:

Audit https://example.com for AI search visibility

What's here

Plugin	What it does
aeo-audit	Audits any URL for AI crawler access, llms.txt, JSON-LD structured data, sitemaps, and .well-known endpoints, then reports findings by severity with concrete fixes

What it checks

Crawler access — robots.txt rules for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, Claude-SearchBot, PerplexityBot, Perplexity-User, Google-Extended, Applebot-Extended, Bingbot and CCBot; X-Robots-Tag headers.

Machine-readable content — llms.txt, llms-full.txt, XML sitemap (including sitemap indexes), and whether the primary content exists in served HTML rather than only in JavaScript.

Structured data — JSON-LD validity, entity coverage by page type, and offer completeness: price, priceCurrency, availability, GTIN/MPN, return and shipping policies.

Page signals — canonical, title, meta description, H1 count, lang, freshness.

Agentic endpoints — presence of /.well-known/ucp, /.well-known/mcp.json and related discovery files, with the served content validated as JSON rather than assumed from a 200.

What it does not do

Stated plainly, because an audit that overclaims is worth less than none:

• No JavaScript execution. The collector reads served HTML, which approximates what a crawler receives. A client-rendered site is reported as appearing client-rendered, not as empty.
• Homepage by default. Other pages are fetched only when you name them. Product and category coverage needs a representative URL from you.
• No claims about ranking. Nobody outside OpenAI, Anthropic, Google or Perplexity knows how those systems select sources. This reports what is verifiable: access, presence, validity, completeness.
• Presence, not protocol conformance. A .well-known endpoint is reported as present or absent. No handshake is performed.
• No content quality analysis. Answer structure, topical coverage and entity consistency are out of scope for 0.1.

Reliability

Signals that could not be fetched are reported as unverified, with the status code — never as passes. Before release the collector was run against 14 sites spanning single-page apps, publishers, hosted commerce platforms, documentation sites and WAF-protected storefronts. Four classes of false verdict were found and fixed:

• inline JavaScript counted as page text, which made client-rendered sites look content-rich
• an H1 regex that never matched real markup
• soft 200s — catch-all routing serving the app shell for /llms.txt, /ai.txt and every .well-known path
• silently truncated bodies reported as complete file sizes

Requirements

Python 3 for the signal collector — standard library only, no dependencies — and network access to the audited site.

Related

The same checks run inside Magento 2 as a CLI command via angeo/module-aeo-audit, part of an MIT-licensed suite of Magento 2 AEO modules on Packagist.

Contributing

Issues and PRs welcome — particularly corrections to the crawler catalogue in plugins/aeo-audit/skills/aeo-audit/references/checks.md, which goes stale as vendors add, rename and deprecate agents.

Licence

MIT — see LICENSE.