Agent Skill & CLAUDE.md

What is an Agent Skill?

An Agent Skill is a small, portable bundle that teaches an AI coding agent when to use a service and how to call it. Two files do the work: CLAUDE.md (a human-readable playbook) and aiscan-skill.json (a machine-readable manifest with trigger phrases, the API surface, and rate limits).

Install the skill

Drop both files into the root of your repo:

curl -O https://aiscan.site/CLAUDE.md
curl -O https://aiscan.site/aiscan-skill.json

Claude Code will pick up CLAUDE.md automatically on next session. Cursor users can paste the contents into .cursorrules or attach the file via the rules panel.

CLAUDE.md

View the live file. It tells the agent:

• The trigger phrases that should activate the skill ("scan this site for AI", "is this site agent-ready?", …).
• How to call the REST endpoint and what the response looks like.
• How to interpret the score and grade.
• How to apply fixes — group by dimension, tackle Discoverability and Bot Access first, follow each fixGuide's ordered steps.
• How to re-scan after applying changes and report the new score.

aiscan-skill.json

View the live file. This is the machine-readable counterpart — name, description, trigger phrases, and the HTTP surface the agent can call.

Trigger phrases

The skill activates on phrasing like:

• "scan this website for AI"
• "check if this site is agent-ready"
• "run aiscan on <url>"
• "make my site work with ChatGPT/Claude/Perplexity"

You don't need to memorise them — just say the intent in your own words and the agent will route to the skill if its CLAUDE.md is loaded.