Launching today
Skillier.ai
Route automatically your AI to the best skills for any task
8 followers
Route automatically your AI to the best skills for any task
8 followers
Skillier is a single skill that routes any request to the best 3 to 5 from a curated index of 4K expert skills (deduped, quality-filtered, and scanned for prompt-injection, database is growing everyday) and loads them into your AI in a click. Works with Claude, Claude Code, Cursor, Windsurf, Cline, and OpenClaw and many more. Free to start, or run it fully offline with the open-source Lite version.
Hello world 👋
I'm Mattew, one of the two makers of Skillier.
I build with AI agents every day, and use skills everyday, I got tired of either hand-picking skills one by one or stuffing my context full of stuff my agent didn't need. Figured there had to be a tool for this. There wasn't, so we cooked it.
Skillier is one skill that fixes that.
You describe what you're working on, and it:
- routes you to the best 3 to 5 skills from a curated index of 4K+ skills, you pick the ones you want
- only surfaces skills that are deduped, quality-filtered, and scanned for prompt injection
- loads them straight into your client, then gets out of the way
Two flavors:
* Lite — runs entirely locally, 1.8K skills, open source on the repo. Do whatever you want with it.
* Full — the curated 4K+ index, growing every day, with the quality and security pipeline running on top. Free to start.
Works with Claude, Claude Code, Cursor, Windsurf, Cline, and OpenClaw and a few more. (if it doesn't, tell us.)
One thing I'd love your input on:
what domain expertise would you want to see in Skillier? Tell me the field you'd want covered and I'll dig into it.
nb: to LOTR fans, there's a reference.
— Mattew
Nice idea, could be very useful. Do you keep a public list of domains covered by Skillier?
@new_user___15120260e95b718fa8b5369 Thank you. Good suggestion, we will add this on the website, and the repo alongside the sources.
Rippletide
Congrats on the launch! How do make sure there is no dangerous skill for the user?
@imbjdd Hey, thanks. Every skill goes through automated analysis for prompt injection and hidden patterns. Any executable parts are run in a sandboxed environment with no network or data access and analyzed, so we can see exactly what they do before the skill enters the index.
so confortable to not have to deal with skill management anymore , the recommandation inside the terminal session it self is really convenient, we can validate all skill recommandation
@chrysh Thank you. We wanted to indeed replaces the tedious manual research on marketplaces, and it really shines when you're stepping into a domain you don't know yet.
And you can still use your own skills alongside it.