Skills
skills/k-dense-ai/claude-scientific-skills/pufferlib/Gen Agent Trust Hub

pufferlib

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill documentation and templates (e.g., SKILL.md, scripts/train_template.py, references/training.md) reference the use of command-line tools such as torchrun and the puffer CLI. These tools are used for standard operations like distributed training and environment management in reinforcement learning workflows.- [EXTERNAL_DOWNLOADS]: The SKILL.md file provides instructions to install the pufferlib package via uv pip install pufferlib. This is a legitimate installation step for the core library the skill is built around.- [DATA_EXFILTRATION]: The training templates and documentation (scripts/train_template.py, references/training.md) facilitate optional integration with experiment tracking services like Weights & Biases (wandb) and Neptune. While this involves sending training metrics to external servers, it is a well-known and standard practice in machine learning for monitoring experiments.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 08:47 PM