Can you explain what exactly its doings? Is it similar to co-pilot? just looks for context? Does it do any kind of sast? Is it aware of potential CWEs? What languages does it support? is there a git repo for this? Also, what is the accuracy/precision like? False positive rate?

Created a Github bot that can review PRs to catch bugs and security issues. It has been useful as a quick first check for issues before diving into architecture changes. Also, discovered while building this that current LLM eval tools suck.