The Agent Skills Directory

[SAFE]: The skill provides standard documentation and Python/CLI code snippets for training anomaly detection models using the Anomalib library.
[INDIRECT_PROMPT_INJECTION]: The skill handles user-provided dataset paths and configuration files. While these represent ingestion points for external data, the skill uses them for their intended purpose in a machine learning training workflow and does not contain patterns that suggest exploitation of these surfaces.
[DYNAMIC_EXECUTION]: The skill documents the use of jsonargparse for CLI configurations and keyword argument forwarding to the underlying PyTorch Lightning Trainer. These are standard implementation patterns within the Anomalib framework and do not constitute an unsafe dynamic execution risk.

anomalib-training