The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill provides legitimate documentation and templates for the PufferLib reinforcement learning library. All scripts use standard scientific computing libraries (NumPy, PyTorch, Gymnasium) in a manner consistent with their intended purpose.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of environment observations. While this constitutes an ingestion point for external data, the data is processed as numerical tensors for model training rather than natural language instructions, presenting no identified path for injection attacks. Standard reinforcement learning practices are followed.
[EXTERNAL_DOWNLOADS]: The skill references the official PufferLib GitHub repository and documentation site for installation and reference purposes. These are well-known resources in the reinforcement learning community.

alterlab-pufferlib