In LayerX's proof-of-concept, it's not too malicious. The hidden prompt tells the chatbot to create a Python-based script ...