I use a bunch of different AI coding tools, and keeping all their markdown in sync was driving me nuts. CLAUDE.md here, copilot-instructions there, cursorrules somewhere else, all slowly drifting apart. Then skills came along and I was pulling in random third-party ones with no real way to version them or know what was actually inside. So I built mdm to fix both. The goal is simple: keep your agent files consistent across 45+ tools, and don't let a sketchy skill slip something nasty into your agent. mdm checks skills for prompt injection, hidden characters, and known vulnerabilities, and a lock file lets a whole team install the same skills in one command, whatever tool they're using. It started as a little Go CLI for me, but the more I dug into how people use skills, the more security became the whole point. Then I started thinking about teams instead of just myself. By v2 it had grown into a vendor-neutral way to manage agent setups, with plugins, subagents, and more. And it's not just me anymore. I've had a couple of other developers start adding features that they genuinely wanted as well.

mdm

Keep your AI agent files in sync and your skills safe

Keep your AI agent files in sync and your skills safe