How is Claude Code used for AWX test-writing? (research question)

,

Hi all, I’m running an academic study comparing the quality of AI-authored tests across several codebases, and AWX is one of them (found via commits carrying a Co-authored-by: Claude <model> trailer).

I reached out to a few individual contributors directly a while back but didn’t hear back, so trying here in case someone can help. Specifically:

  1. Is there a standard prompt/workflow for using AI on AWX PRs e.g. is it typically pointed at a JIRA/AAP ticket and asked to implement + write tests together, or is test-writing usually a separate pass, or something else?
  2. Do human edits, if any, tend to be just small tweaks or more major?

Thanks!