The Agent Skills Directory

[SAFE]: No security issues detected. The skill contains legitimate educational content and code templates for building and training machine learning models using the PyTorch Lightning ecosystem.
[EXTERNAL_DOWNLOADS]: The documentation includes standard installation commands for well-known machine learning packages such as lightning, torch, transformers, ray, and optuna from official package registries.
[INDIRECT_PROMPT_INJECTION]: The skill describes building data processing pipelines using DataLoader. While this constitutes a data ingestion surface, the code is limited to training logic and does not expose dangerous system capabilities. Evidence: Ingestion points: DataLoader and Dataset in SKILL.md; Boundary markers: Absent; Capability inventory: Model training and logging; Sanitization: Absent.

pytorch-lightning