Implement error handling for flawed prompt output

[0x4D5352]

When generating drones, sometimes the GPT4ALL model will spit out responses that include ### PROMPT: or special characters, which can cause problems down the road for formatting of things like directory structure and stable diffusion prompts.

This could be fixed with better prompts and/or better models, but user-implemented error handling will still be useful. Can probably be handled by passing the output of the model into a more sophisticated parser.