The Agent Skills Directory

[SAFE] (SAFE): No security issues were detected in the provided markdown files. The content consists entirely of technical documentation and code examples for a legitimate machine learning library.
[DATA_EXPOSURE] (SAFE): Code snippets use standard local directory paths (e.g., ~/kg-datasets/) for dataset storage, which is conventional and does not expose sensitive system secrets or credentials.
[EXTERNAL_DOWNLOADS] (SAFE): The documentation refers to well-known scientific datasets and pre-trained models (like ESM). While the library may download these during use, no suspicious or untrusted download patterns (like piped shell executions) are present.
[PROMPT_INJECTION] (SAFE): The instructions are descriptive and technical; they do not contain any attempts to override agent constraints or extract system prompts.

torchdrug