Checked, not scraped
How we check every skill.
The skill market is flooded: millions of skills scraped off GitHub automatically, average at best — and roughly one in three carries a prompt-injection hole. A skill runs instructions inside your agent. That is why nothing gets into StackScout unchecked.
We write or read every skill ourselves
No auto-scraping GitHub. Every skill is either ours or read line by line here before it gets into the library.
No hidden behaviour
No quiet network calls, no tracking, no data leaving your machine. What a skill does is written in its SKILL.md — and that is all it does.
MIT licence — it is yours
Every skill of ours is MIT. Use it, change it, keep it. Cancelling never takes a downloaded skill away from you.
Sharp scope instead of a monster skill
Two or three sharp skills beat one do-everything skill that does nothing properly. Focused skills measurably help the agent more.
Maintained, not filed away
When tools or best practices change, we update the skills. This is not a catalogue of dead entries.
That is why the library is small and curated: quality over catalogue size.