The Agent Skills Directory

[SAFE]: The skill provides legitimate templates and documentation for the PufferLib reinforcement learning framework, which is a high-performance library for speed-optimized training.
[SAFE]: External dependencies and references target standard, well-known libraries and services in the machine learning ecosystem, such as PyTorch, NumPy, Weights & Biases, and Neptune.
[SAFE]: Training and environment scripts follow established best practices for reinforcement learning development, including proper neural network initialization and efficient data handling via shared memory buffers.
[SAFE]: Network operations are confined to official logging integrations that require user-supplied credentials or project names, posing no risk of unauthorized data exfiltration.

alterlab-pufferlib