The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides Python code snippets intended for the agent to execute to fetch run history from experiment tracking APIs and to process local TensorBoard event files. This is a functional requirement for its core purpose of data comparison.
[PROMPT_INJECTION]: The skill demonstrates a surface for indirect prompt injection as it ingests run metadata and metric history from external services and local JSON files. However, the risk is minimal as these sources typically provide structured numeric or log data.
[SAFE]: No malicious patterns were identified. The skill interacts with well-known industry services using standard authentication methods and does not exhibit signs of exfiltration, persistence, or obfuscation.

compare