NEWFree AI visibility report. Monitoring from $125/month

Claude Code skill

citehawk-geo is a Claude Code skill that runs the whole GEO loop for you: it reads your real visibility gaps from CiteHawk, edits your own website's repo to close them, and lets the next weekly collection measure the lift. No guessing at generic SEO — every fix traces back to a specific gap CiteHawk actually observed.

What it does

The skill works in a three-step find-fix-measure loop:

  • Find — it calls the CiteHawk MCP tools (get_visibility_summary, list_recommendations, get_source_gaps, get_rankings, get_prompt_performance, get_evidence) to pull your scores, prioritised recommendations, and the exact prompts and sources where competitors beat you.
  • Fix — it works in your site's repo: adding schema.org / JSON-LD, writing AI-citable comparison and FAQ pages, tightening titles and answer-first content — mapped to the real gaps.
  • Measure — it commits the changes and tells you that CiteHawk's next Monday collection will re-measure presence, citation rate, and the affected prompts, so the lift shows up in your scores.

Install

The skill reads your data through the CiteHawk MCP, so add that first (an API key is self-served from Settings > API Keys):

1. Connect the MCP

claude mcp add --transport http citehawk https://www.citehawk.com/api/mcp \
  --header "Authorization: Bearer YOUR_KEY"

Then add the skill:

2. Add the skill

npx skills add richyparr/citehawk-skills

Confirm the connection by asking Claude to run get_visibility_summary once. If it returns available: false, your workspace has no collection yet — run a collection in CiteHawk first.

Example session

A typical run looks like this:

  1. You ask Claude Code: “Improve my AI visibility.”
  2. The skill calls get_visibility_summary and list_recommendations, and reports your scores plus the 3-5 highest-leverage gaps — say, a low presence score because two competitors are cited on a review directory you are missing from, and a losing prompt of “best tools for X”.
  3. It inspects your repo, then adds an Organization and FAQPage JSON-LD block and drafts a comparison page naming the real competitors from get_rankings, targeting the real losing prompt.
  4. It commits with a focused message and tells you the fix takes effect once deployed and re-crawled — next Monday's CiteHawk collection quantifies the lift. You re-run the skill after that to pick the next target.

Off-site gaps (getting listed on a third-party directory) can't be fixed by a repo edit — for those, the skill drafts the outreach or listing content and hands you the action. See the MCP tools reference for what each call returns.