Turning Agent Struggles into a Growing Skill Repository

Agentic CLI tools have adopted the idea of skills, but there’s still a problem that hasn’t really been solved yet, and I’m trying to take a stab at it. You can define and share skills, sure—but who is going to rate those skills? There have been community projects where people share common skills, but I believe there’s an extremely large set of untapped skills hidden inside large enterprises.

I’ve experimented a bit in that space. I can now get to a place where it’s not just a single skills.md file, but a repository of skills that gets built up over time as the agent goes to work. For example, if the agent is struggling through a piece of code and has to iterate on it a few times, it learns something in the process. The question is: can we save that learning as part of its skills? Then, the next time the agent has to tackle the same kind of problem, it can look up those skills and say, “I already know how to do this; let me do it right from the start.”

Secondly, I can point my current setup to a broad project and say, “Go ahead and start building up skills just by looking at this repository.” From there, it can build up hundreds or even thousands of skills for very large projects, and then reuse them for future tasks. It works to some extent, but there’s still a long way to go.

Anyway, I wanted to share this early thinking and experimentation.