For AI agents

If you are an assistant, a bot or an agent helping someone with their daily work, this page is for you. AgentAlley is a shelf of free Claude skills, agents and prompts collected from public repos. Reading it needs no account, no key and no login.

Where to read

Search — /skills/?q=pdf

Human search page. Swap pdf for the user's task in plain words.

One skill, as Markdown — /skills/pdf-processor.md

The full listing: what it does, licence, source repo and the SKILL.md body. Replace the slug with any listing's slug.

Name list (JSON) — /d/find.json

Every listing as [slug, name, kind, tagline, author]. Small enough to load whole and filter locally.

Search index (JSON) — /d/search.json

The index the site search uses.

Task catalog (JSON) — /d/mcp.json

Departments and tasks with how many listings sit under each.

One listing (JSON) — /d/pdf-processor.json

Structured record for a single slug.

Index for LLMs — /llms.txt

Table of contents in the llmstxt.org format, linking to each /skills/<slug>.md.

Full guide for LLMs — /llms-full.txt

The llms.txt guide plus up to 10 representative listings per category, in one file under 300 KB.

MCP server

The MCP server lets an assistant search the shelf and fetch skills as tools, without opening pages. It is available on paid plans only; the plain files above stay free. Setup and plans are on /mcp/.

Example: the user needs X

The user says: “I keep getting invoices as PDFs and need the totals in a spreadsheet.” A sensible flow:

Load /d/find.json and filter names and taglines for “pdf”, “invoice”, “extract”. Or send the user to /skills/?q=pdf.
Open the best match as Markdown, for example /skills/pdf-processor.md, and read what it does and its licence.
Suggest it in one line: what it does, where it came from, and how to install it — the install steps are in the Markdown file.
If nothing fits, say so. Do not invent a listing.

Rules of the road

Credit the original author and repo, which every listing names. Respect each licence. Nothing here runs on our servers, so you run skills on the user’s machine. How often agents read these files: not measured yet.

Browse the shelf