Agentic Jujutsu

This wraps Jujutsu version control with a self-learning layer that tracks AI agent operations and suggests approaches based on past successes. The ReasoningBank system records operation trajectories with success scores, then surfaces patterns when you face similar tasks later. Claims 87% auto-conflict resolution and 23x faster than Git for multi-agent work, though those numbers lack context. The quantum-resistant crypto (SHA3-512, HQC-128) feels like future-proofing overkill for most projects. Most interesting for teams running multiple AI agents that need to coordinate without stepping on each other, where the pattern learning could actually save debugging time once you've fed it enough trajectories.

Install to Claude Code

npx -y skills add ruvnet/ruflo --skill agentic-jujutsu --agent claude-code