Prompted Policy Search uses LLMs for RL policy search, outperforming common methods across diverse tasks.