Sklm - Centralize, scope, and sync skills for every AI agent

Sklm is a CLI tool that centralizes management of skills (SKILL.md files) for your AI agents. It solves the tension between wanting skills globally available vs. per-project scoped, without polluting your agent's configuration. Multi-agent support - Global store - Registry discovery - Auto-sync

Add a comment

Replies

Best
Hello everyone! I was fed up with collecting skills across all my agents, which was cluttering up my list in my projects. So I switched to installing them on a project-by-project basis, but I also got tired of having to remember which skills I’d used previously. So I coded a tool this weekend, in the form of a CLI: Just like with skills.sh, you give the CLI the skill’s repo, and it’ll store it in a global ‘store’; then, project by project, you simply activate the skills you need – those not active in the project aren’t available in that one. Migration is available if you already have skills installed. 8 agents are already supported. The whole thing is licensed under the MIT licence I’d love to hear your feedback; it’s my first personal open-source project in production.

Nice launch. The global-store vs project-scope split is exactly the tension that shows up once people collect more than a few files.

One thing I’d be curious about is reproducibility: can a project lock the exact skill version/source commit it activated, so another machine or teammate gets the same behavior later? For agent workflows, “which skill was active” is almost part of the build environment.

 
Thanks for your feedback!

That’s an interesting question. I haven’t looked into version management for skills yet, but it’s something I'm planning to do.

Skill sprawl is a quieter problem than people admit... I've got the same skill copy-pasted across three projects and they've all drifted, so I never know which version actually fired. The "scope" part is what I'd want most. How do you decide which skills load for which agent without stuffing every context window?

 Thank you for your feedback!

Sklm does not handle the selection of skills by the agent for the moment; it is solely a project-by-project skills manager.

🚀 Sklm v0.2 is here! 💥

What's new:

✨ Unified command structure
👨🏻‍💻 A single CLI instead of multiple separate commands
📦 No more cloning from GitHub — install it directly from PyPI:

pip install sklm-cli

Two weeks ago, while organizing my projects (I'm a bit obsessed with keeping things tidy 😅), I got tired of having hundreds of AI skills scattered across different projects or cluttering up my global skills list.

So I built a small CLI to solve that problem.

The idea is simple:

  • 📚 A global store where you keep all your skills.

  • 📥 Import only the ones each project actually needs.

  • 💤 Don't need a skill anymore? Disable it in your workspace without removing it from your global store.

And since I was already building it for myself... why not open source it?

Two weeks later:

⭐ 9 GitHub stars
🏆 Top 60 on Product Hunt
🔥 265 unique visitors on the landing page

Not bad for my first open-source project. 🚀

If you're managing skills across multiple projects, I'd love to hear your feedback!