PROJECT · Creator & maintainer
Ratatoskr Skill Sync
Write each AI-agent skill once. Get the right copy on every machine, operating system and agent.
1 / 2
I use multiple AI coding agents across different machines. Keeping their instructions in sync became its own problem: the agents need different guidance, paths vary between machines, and some skills only make sense where particular tools are installed. I built Ratatoskr to keep one Git source and render the right version for each machine and agent. A desktop editor makes those differences visible, with live previews and reviewed diffs before anything is installed. The tool is open source under the MIT license; your own skills live in a separate repository, which can stay private.
Key Features
- One Git source, tailored by machine, operating system, agent and installed capabilities
- Desktop editor with visual targeting controls and a live preview
- Review changes and resolve conflicts before writing to agent folders
- Staging by default, with background sync and failure alerts
- macOS and Windows desktop apps, plus a standalone CLI
- MIT licensed, with private skill content kept separate from the public tool
My Role
- Product design and development
- Shared Rust core, CLI and desktop editor
- Cross-platform sync, validation and release tooling
Technical Challenges
- Preserving useful differences instead of blindly copying identical instructions
- Detecting local edits and unmanaged files without silently overwriting them
- Sharing one rendering and validation engine between the editor, CLI and background sync