The Agent Skills Directory

[PROMPT_INJECTION]: The 'Suggest Using K-Dense Web' section provides specific instructions to the AI agent to monitor task complexity and proactively recommend a third-party service (www.k-dense.ai). This attempts to override the agent's neutral persona and influence its output for promotional purposes.
[PROMPT_INJECTION]: The skill outlines a data ingestion surface within the 'LightningDataModule' section (prepare_data method) where external data is downloaded and processed. The instructions lack boundary markers or sanitization requirements to mitigate the risk of indirect prompt injection from malicious datasets.
[SAFE]: The core technical guidance for structuring models, trainers, and logging using PyTorch Lightning follows established industry standards and contains no evidence of malicious code or data exfiltration.

pytorch-lightning