tao-train-action-recognition
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references Docker images from the vendor's official container registry (nvcr.io). The benchmark metadata also refers to a vendor-owned internal GitLab container registry. These are standard and trusted resources for the TAO Toolkit.
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute shell commands using Docker to perform model training, evaluation, and inference. These are required for the skill's primary machine learning purpose.
- [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill processes user-provided video datasets and YAML configuration files.
- Boundary markers: No specific delimiters or safety instructions are defined to separate instruction from data in the input files.
- Capability inventory: The skill has permissions to execute Docker commands and interact with the local file system for results management.
- Sanitization: No input validation or sanitization routines for the configuration files were identified.
Audit Metadata