@devopscats Chat 3.5 seems to still struggle with this really badly, it acknowledged its mistakes when I pointed them out and then proceeded to hallucinate about cabbages and wolves that were not in the original prompt.

Claude does pretty well, my first prompt yielded something overly complicated but when I pointed that out Claude gave the solution and apologized for over complicating it. I asked it to write a more clear prompt for itself, put that into a new session and it got it in one shot.