huggingface-trackio

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFE
Full Analysis
  • Command Line Interface: The skill facilitates the use of the trackio CLI to discover projects, inspect runs, and retrieve metric values. This is an expected pattern for local experiment management and automation.\n- Network Communication: The tool supports synchronizing metrics with Hugging Face Spaces and sending notifications via user-configured webhooks for services like Slack or Discord. These operations are part of the core functionality for real-time monitoring.\n- Data Ingestion Considerations: As the agent interprets metrics and alert descriptions generated during training, it is a recommended practice to ensure these are processed as data to prevent potential confusion with instructions.\n
  • Ingestion points: Experiment metrics and alert text are retrieved using the trackio list alerts and trackio get metric CLI commands.\n
  • Boundary markers: There are no explicit delimiters or specific 'ignore instructions' warnings defined for the external metric data in the provided references.\n
  • Capability inventory: The skill allows the agent to execute training scripts and perform local data queries via the CLI tool.\n
  • Sanitization: No specific sanitization or validation rules for the text content of logged alerts are described in the documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:19 PM