I couldn't find any information on the set up of PhilosopherAI. It doesn't appear to pass prompts through verbatim, because it rejects some; I assume the user-submitted prompt is prepended by another one, which sets it up to expect a "topic".