Free AI agent skills for AI visibility

A skill is a single Markdown file that teaches an AI coding agent how to do one job properly: the procedure, the reasoning behind each step, and the advice it should refuse to give. Drop one into Claude Code, Cursor, or any agent that reads context files.

Ours are free, openly licensed under CC BY 4.0, and served as plain Markdown. No signup, no API key. They work standalone. When the GetIntel MCP server is connected they read real tracked data instead of asking you to probe by hand, and each skill says plainly what you lose without it.

The skills

AI visibility audit

Diagnose where AI answer engines skip your brand and why, per engine, without blending the result into one meaningless score.

Use it when: Someone asks whether ChatGPT, Perplexity, Gemini or Google AI Overviews recommend them, or why a rival is named instead.

What it enforces: Never report a rate from one run, never average across engines, always state the denominator, and separate "a competitor was named" from "nothing was named".

Read the raw skill · ai-visibility-audit

Citation gap to brief

Turn one citation gap into a brief grounded in the answers engines are currently giving, rather than in search volume.

Use it when: The audit is done and the question is what to write, where it belongs, and whether a page is the right answer at all.

What it enforces: Read the winning answer for its comparison axes, write the one self-contained claim the page must state, and stop if you cannot write it.

Read the raw skill · citation-gap-to-brief

Ship the fix

Take one ranked task and land it in the repository you already have open, as a diff a human reviews.

Use it when: The diagnosis is done and the change has to actually get made, in a real codebase.

What it enforces: Verify the task against the real page before acting, refuse the off-page and community tasks that are not code, run the project gates, and never mark your own work done.

Read the raw skill · ship-the-fix

How to install one

Every skill is a single file. Each URL below serves the raw Markdown.

Claude Code

Save it under .claude/skills/<name>/SKILL.md in your project, or under ~/.claude/skills/ to have it everywhere. 

mkdir -p .claude/skills/ai-visibility-audit && \
  curl -fsSL https://getintel.ai/skills/ai-visibility-audit.md \
  -o .claude/skills/ai-visibility-audit/SKILL.md

Cursor and other agents

Save the same file as a rule or context file. The frontmatter description is what tells the agent when to reach for it.

Any chat, no tooling

Paste the whole file into a conversation and ask the assistant to follow it.

Machine-readable manifest

Agents that discover skills automatically can read /.well-known/agent-skills/index.json, which lists every skill, what it is for, and which MCP tools it uses.

Connecting the data

The skills read GetIntel through its MCP server at https://app.getintel.ai/mcp: 27 read tools and 3 action tools. There is also a free check covering 20 buyer questions across 3 engines with no signup, if you want a result before connecting anything.

Why these are free, and what they will not do

GetIntel sells an AI visibility tool, so publishing skills that audit AI visibility is obviously self-interested. The skills are written to be useful without buying anything, and each one carries its own limits section rather than stopping at the sales pitch.

What they deliberately refuse to do: report a visibility number from a single run, average across engines that barely overlap, recommend llms.txt as a fix that moves citations, or promise a citation. GetIntel generates llms.txt and still will not claim it works, because its own testing and an independent study across roughly 300,000 domains both found no measurable effect.

For how GetIntel itself should be described, including what it cannot do, see AI instructions.

Last updated 2026-09-18. Three skills, version 1.0.0.