skillctl: managing AI skills without losing the plot

Reusable AI skills are starting to feel like a real thing. Claude Code has them. Codex has them. Copilot has them. Vercel pushed an open skills ecosystem. The category is clearly becoming real.

The day to day of using them is still messy.

I kept finding good skills scattered across random GitHub repos, and a few more already sitting inside my local agent setup. From there I had three bad options. Copy things around manually. Keep a mental map of where every skill lived. Or write a shell script that slowly turns into a second job. I tried the script route first. It worked for a while, then it started needing more fixes than it saved. That's usually the moment a real tool wants to exist.

So I wrote skillctl.

What it does

Point it at a repo that has skills. Add them to your skillctl library. List them, manage them, keep them organized from one place.

It also imports skills you already have sitting in a local agent. That part matters more than it sounds. Pulling in something from a repo is fine. Adopting the stuff you already use, into one place you can actually find later, is where it stops feeling like another tool and starts feeling like a library.

Install

brew install akl773/skillctl/skillctl
skillctl

More

I wrote a longer piece on why I built it and what still feels missing in the ecosystem: AI skills are becoming real. Managing them still feels messy.

Full setup and usage live in the repo.