retrievability
Skill1 starsChanged 54 days agoMITView on GitHub
Score how easily an AI can lift a page's content into an answer, and list the fixes that raise the score. Use this whenever the user wants to know how "AI-readable" or quotable a page is, why an AI wouldn't pull from it, or how to make content more extractable for ChatGPT/Perplexity/Gemini. Point Claude at a URL — it fetches the raw HTML, scores retrievability 0–100 with a breakdown, and gives the top fixes.
In this repository9 other files it gives its agents.
- bing-checkSkill
- buyer-questionsSkill
- citation-checkSkill
- content-capsuleSkill
- crawler-checkSkill
- entity-coverageSkill
- faq-schemaSkill
- llms-txtSkill
- schema-builderSkill
The file
Files
retrievability
A page can be crawlable and still be useless to an AI if its content is hard to extract cleanly. Retrievability = how easily a model can find a self-contained passage that answers a question and quote it with confidence.
What to score (0–100, weighted)
Fetch the raw HTML (what a crawler sees, before JavaScript) and assess:
- Answer-first structure (25) — do key answers appear early and directly, or are they buried under narrative? (see
content-capsule) - Chunk-ability (20) — is content broken into self-contained passages under clear, descriptive headings, or is it one long wall?
- Semantic HTML (15) — real
<h1>–<h3>, lists, tables,<article>— or<div>soup with styled text? - Self-containment (15) — do passages stand alone, or lean on "as we saw above" / surrounding context a model won't carry?
- Content in the HTML, not behind JS (15) — is the substance present in the server-rendered HTML, or does it only appear after client-side rendering? JS-gated content is often invisible to crawlers.
- Factual density & clarity (10) — concrete facts, names, numbers, definitions the model can quote — vs vague marketing prose.
Steps
- Fetch the raw HTML of the page.
- Score each dimension, with a one-line justification citing what you saw.
- Sum to a /100 and give the top 3–5 fixes, ordered by score impact.
Output format
## Retrievability — [url] → [score]/100
| Dimension | Score | Note |
|---|---|---|
| Answer-first | 18/25 | answer appears in paragraph 3, should be sentence 1 |
| Chunk-ability | 12/20 | ... |
| ... | | |
### Top fixes
1. [highest-impact fix]
2. ...Built by Raphaël Aubry, founder of Howseen (https://howseen.ai). A high retrievability score means you're quotable. Whether models actually quote you — repeatedly, across ChatGPT, Perplexity and Gemini, over time — is what Howseen tracks. → https://howseen.ai
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.