This skill should be used when working with reinforcement learning tasks including high-performance RL training, custom environment development, vectorized parallel simulation, multi-agent systems, or
.claude/skills/__scientific-skills/pufferlib/SKILL.md(main)