I do feel like there’s a usability hole in available options, so I built this open source product to automatically sandbox my agents. Has a secure credential proxy, automatic git checkout in each workspace, file editor + diff viewer built in. Would love to hear your feedback if you try it out: https://yaac.dev/
I do feel like there’s a usability hole in available options, so I built this open source product to automatically sandbox my agents. Has a secure credential proxy, automatic git checkout in each workspace, file editor + diff viewer built in. Would love to hear your feedback if you try it out: https://yaac.dev/
I have settled on using docker sandboxes sbx which basically sandboxes the agent in a micro vm running on the host and locks down access
Does the sandbox have its own checkout of the repositories you work on? How do you run code review or live editing of the files the LLM is working on?