rl-anything
Automatic skill/rule optimization using direct LLM patches with fitness evaluation and human approval workflow