Entity Intelligence: A Claude Skill

Entity Intelligence is a free Claude skill that takes any URL and hands back an interactive knowledge graph of every named entity on the page, with verified Wikidata QIDs and a depth diagnostic that tells you whether the page is built to win SERPs or built to lose them. About two minutes per page.

  • Built for: SEOs and content strategists trying to understand what a page is actually about
  • Delivers: A self-contained HTML knowledge graph plus a depth/drift diagnostic for your page.

Why I built it

Most entity SEO tools give you a list and stop. A list doesn’t tell you whether to add more entities, cut some, or rewrite for depth. Two things I wanted built in from the start:

  1. Verified Wikidata QIDs, not LLM guesses. Wikidata QIDs are random integers, no LLM can guess them reliably. The first draft of this skill let Claude guess, and the guesses were almost always wrong. So I wired the pipeline to hit the real Wikidata API for every entity. If there’s no real match, the panel says “Not verified” instead of linking to a fabricated page.
  2. A depth check, not just an entity inventory. My research on 10,000 healthcare SERPs found that Position-#1 pages have fewer entities but more words per entity than Position-#10 pages. Counting entities doesn’t tell you whether a page is positioned to rank. So every run also returns words-per-entity, top-3 focus, drift mass, and whether the page lands in the deep-focused or shallow-broad quadrant.

What it does

Point it at a URL. Claude scrapes via Firecrawl, chunks the content, extracts every named entity with prominence scores, deduplicates aliases, verifies QIDs against the real Wikidata API, runs the depth diagnostic, and renders an interactive HTML knowledge graph you can drag, zoom, click, and share. Fully self-contained. No CDN dependencies. Email it, embed it, drop it in a client deck.

What the diagnostic surfaces

Two real runs:

Investopedia’s Roth IRA page (2,636 words, 38 entities): 69 words/entity, top-3 focus 17.6%. Verdict: shallow & broad. Investopedia builds topical authority across thousands of pages, not within any one. The skill correctly flagged the natural consequence of that strategy at the single-page level.

AAC’s inpatient rehab page (979 words, 36 entities): 27 words/entity. Same verdict, weaker excuse. 36 entities crammed into under 1,000 words with no clear hub is a page trying to be everything to everyone and ending up nothing to anyone.

The recommendation is derived from the actual numbers: which entities to cut, how much depth to add, or whether the page is part of a site-architecture play.

Install (about 30 seconds)

  1. Download the .skill file below
  2. Open Claude, switch to Cowork mode
  3. Customize → create new skill → upload
  4. Save

Next time you mention “entity analysis” or “what’s this page about” in a conversation, Claude picks it up automatically.

Grab the skill

One favor

If you try it and hit anything weird, reply on LinkedIn or email me.

If you’re looking to build out AI workflows for your marketing team, turning the repetitive parts of your work into one-click skills that analyze, audit, report, and take action, reach out. I’d love to help.

Happy graphing, see you in the SERPs.


WillDoSEO.com

Powered By Espresso ☕️

Disclaimer: I’m not affiliated with, sponsored by, or endorsed by Anthropic, Wikidata, Firecrawl, or any other company mentioned. I’m just an SEO who built something useful and wanted to share.