learning-quality
Universal evaluation layer for standard RL environments. Measures what an agent learned - not just how much reward it accumulated.