The recently observed phenomenon of ‘doomprompting’ with LLM and AI agent results can lead to poor outcomes and huge costs.