Comment by ChadMoran

I've been doing this with Claude Code and agent teams.

I have a /red-team skill that will use an agent team to criticize it's own work, grade and rank feedback, incorporate relevant feedback and then start over. It has increased the quality of output.