pufferlib
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: The skill provides documentation and implementation templates for PufferLib, a performance-oriented reinforcement learning framework. All scripts and references align with legitimate machine learning development workflows.- [COMMAND_EXECUTION]: The skill includes instructions for running training tasks via CLI tools such as 'puffer' and 'torchrun', which are standard utilities for distributing and managing reinforcement learning workloads.- [EXTERNAL_DOWNLOADS]: The documentation references the installation of the 'pufferlib' package and its dependencies (such as PyTorch and Gymnasium) from standard public registries.- [SAFE]: The training templates include support for experiment tracking and telemetry services (Weights & Biases, Neptune) using standard configuration patterns that avoid hardcoding sensitive credentials.
Audit Metadata