I've had mixed results.

Most models don't have a 100% correct CLI usage and either hallucinate or use some deprecated patterns.

However `jj undo` and the jj architecture generally make it difficult for agents to screw something up in a way that cannot be recovered.

Try using https://github.com/danverbraganza/jujutsu-skill

This is enough of a command reference that with it, agents are able to work with jj pretty well.

I've gone all in on jj with a OSS framework I'm building. With just a little extra context, the agents have been amazingly adapt at slicing and dicing with jj. Gives them a place to play without stomping on normal git processes.