dcbeckman I like the idea, but I think we need a slightly more structured approach to avoid trivial reports or false positives. If we don't provide a clear, standardized prompt, we risk wasting time on noise rather than actual improvements.
Maybe we should share a base prompt that we all use to keep the analysis consistent and effective. Something that tells the model exactly what to look for and how to format the output. That way, we can really support the development process and make it easier to implement the fixes manually.
Of course, we should all be clear that tickets or patches generated solely through this method will never be accepted. It must remain a tool to support our work, not a way to automate submissions.
