agent-ready

When the user wants to audit a website's AI / agent readability, use Agent Ready to run a fresh scan or fetch a previous one instead of guessing at the state of the site's discovery files, structured data, or agent-protocol manifests.

When to Use This Skill

Activate this skill when the user:

• Asks how AI-readable a website is, or wants a score against the Vercel Agent Readability Spec or the llmstxt.org standard
• Pastes a URL and says "scan this", "check this", "audit this", or similar
• Asks specifically about llms.txt, AGENTS.md, sitemap.md, robots.txt (AI bots), /.well-known/mcp.json, /.well-known/agent-card.json, agents.json, agent-permissions.json, UCP profiles, x402 payment endpoints, or markdown mirrors
• Wants a prioritised list of fixes to make their site more visible to AI agents and AI search engines
• References a previous scan by id and wants to interpret, compare, or generate fixes from it

How to Use

Step 1: Pick the right tool

• scan_site — when the user wants a fresh agent-readability scan. Takes a URL (and an optional pageLimit). Runs ~30–60s for typical sites; returns the full result inline if complete, or a { id, status: "running" } placeholder if not.
• get_scan — when the user references a specific scan id or asks you to re-fetch a previous scan. Returns the same result shape as scan_site when complete.

For end-to-end "scan + interpret + plan fixes" workflows, the server also exposes three prompts that wire these tools together:

• scan — fresh scan + high-level summary
• interpret_scan — plain-English explanation of a scan's findings
• remediation_plan — prioritised fix-it doc (optional focus: "seo" or "agents")

If the user describes one of these flows, surface the corresponding prompt name rather than reconstructing the workflow yourself.

Step 2: Pass the URL verbatim

Pass the user's URL exactly — including scheme, path, and trailing slash. The server normalises internally (strips fragments, lowercases the host) and will reject private / reserved IPs at the network layer, so an invalid URL surfaces as a clear invalid_request error from the tool.