I’ve been building Skillshare (https://github.com/runkids/skillshare), a tool to sync AI agent skills across different environments.

While tools like Vercel’s add-skill are great for quick discovery, I’ve noticed a lack of discussion on how teams manage these shared capabilities long-term.

I built Skillshare to test a "Skills-as-Code" hypothesis:

Git-based Sync: Automatically track and pull skills from remote repositories, flattening nested structures so tools like Claude Code or Cursor can use them immediately.

Team Standards: Ensuring everyone on the team is using the same version-controlled "source of truth" for specialized scripts and prompts.

As an independent developer, I'm curious to hear your thoughts:

Is managing AI skills via Git the right mental model for production teams?

Or do you see this evolving into more of a centralized "registry" approach?

Looking forward to your feedback.