Any AI chat can search SkillFed

How it works

A browsing chat can only open a link that you gave it, or one it found inside a page it already fetched. It cannot invent a URL and call it. So the chat composes the search link, shows it to you, and waits.

1. Ask your chat to find a skill for what you’re working on.
2. It writes a link — something like https://skillfed.io//api/q/write-good-git-commit-messages
3. You paste that link back into the conversation.
4. It reads the results and uses them. Nothing is installed on your machine.

If your chat says it is going to fetch something and then nothing happens, that is this rule at work: it needs the link in your message, not its own.

Why you paste the link yourself

This looks like friction and it is the point. A chat that could reach any URL on its own is a chat that can be talked into reaching one by a web page it happens to read. Requiring the link to come from you means every request is one you saw and approved. It is the same reason the finder never sends your plan, your code, or your files — only an abstract description of the capability you’re looking for.

Why a skill catalog exists at all

A model’s knowledge is an average of what was written before its training cutoff. For anything that moves — framework migrations, tooling defaults, packaging, security practice — that average is a confident summary of a past consensus, and it goes stale quietly. A skill is a dated document that someone maintains. Reading one is how a model finds out whether what it was about to say is still what people do.

A skill is not a rule to obey or a script to run. It is field notes: what a practitioner wrote down about how this is currently done. Most of the value is in reading one, which is why consulting is the ordinary case and installing is the exception.

The longer version of this argument, and what we measured while building it: any AI chat can now run skill search.

What you’re getting

Skill bodies served through the API are snapshots we took and reviewed — pinned to a recorded version and scanned on a best-effort basis with automated security tooling. That is meaningfully better than pulling a file straight from a repository nobody has looked at. It is not a guarantee, and it does not replace your own judgement about whether to act on what a skill says.

Two things follow from the word snapshot: you get the exact bytes we indexed and described, and you may not get whatever sits at the repository’s HEAD right now. Every result carries the version and the date it was taken.

The directory pages on this site cover a wider index than the reviewed set. They are a reading surface — SkillFed describes and ranks what is there; it does not audit third-party repositories, and a listing is not an endorsement.

Skills disagree with each other. Two independently written skills saying the same thing is current practice; one asserting it alone is a hypothesis worth checking.

Reference

Both routes are plain GET, no key, no account. The first takes a hyphenated phrase and returns five ranked skills; the second is the same search with a result count you choose.

A worked example, live: https://skillfed.io//api/q/write-good-git-commit-messages

Each result carries a skill_id, a description, a score, and license and publisher metadata. It also carries absolute links that can be followed directly: body_url is the reviewed snapshot of the skill; source_url is the original repository; page_url and md_url appear when the skill has a page on this site, and are the human-facing view.

Some chats drop long query strings before sending them. The /api/q/ route exists for that reason — it carries no parameters, so there is nothing to drop.

If you have a coding agent instead

Claude Code, Cursor and similar harnesses can call the same index directly, without the paste step, and can install a skill when one is worth keeping.

Machine-readable summary of this site: llms.txt. Who runs SkillFed and how the index is built: about. What we collect: privacy note.