pytorch-lightning

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the user to install well-known machine learning libraries such as lightning, wandb, mlflow, and deepspeed from official package registries. These are established tools in the deep learning ecosystem.
  • [COMMAND_EXECUTION]: Shell commands are used appropriately for package management (uv pip install) and training orchestration. No suspicious or arbitrary command execution was detected.
  • [DATA_EXFILTRATION]: While the skill documents integration with remote logging and experiment tracking platforms (Weights & Biases, MLflow, Comet.ml), these are intended primary functions of the framework for monitoring training progress and are standard industry practice.
  • [PROMPT_INJECTION]: The instructions focus purely on technical documentation and code organization for PyTorch Lightning. There are no attempts to override agent behavior or bypass safety filters.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 11:09 AM