One should not expect ChatGPT or any LLM to correctly generate Prolog code. After hundreds of Prolog code prompts of ChatGPT and having seen a lot of the different Prolog code online, as many others note about training data, if it is not in the training data then don’t expect similar for the output and that garbage in will lead to garbage out. So while I do find that ChatGPT is good at giving me a compass direction to head or a general idea of what the code should look like I always test it. It often fails the make. However if one has seen and used lots of Prolog code it does make it easier to find the problems and quickly decide if one should abandon the ChatGPT generated code or try to fix the issues.
As I don’t understand what you seek by reading the prompt I can’t judge the code.
Have you tried converting the Bratko code to Haskell, then having ChatGPT extend it and get it working using your knowledge of Haskell and then convert that back to Prolog?
You really need to take the time to work through the first several chapters of the book because when you get stuck you then have to learn how to debug the Prolog and you can come back to your problem. You keep trying to skip the basics and we are not here to teach them to you.