Synthetic Instincts: Echoing Reinforcement Learning Agents for Behavior Tree Generation