tao-generate-image-embeddings

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions to execute Docker commands with configurations optimized for GPU performance, such as sharing the host network and IPC namespaces.
  • [EXTERNAL_DOWNLOADS]: The skill downloads official TAO Docker images from the NVIDIA Container Registry (nvcr.io) and retrieves pre-trained model weights from HuggingFace.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: Processes Parquet files provided by the user that contain lists of image filepaths.
  • Boundary markers: No specific boundary markers or delimiters are used for the content within the Parquet files.
  • Capability inventory: The skill possesses capabilities to execute shell commands via Docker and write Parquet files to the local file system.
  • Sanitization: Employs a verification script (verify_image_embeddings_spec.py) to validate that input files exist and output directories are writable before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 02:56 AM
Security Audit — agent-trust-hub — tao-generate-image-embeddings