tao-train-pose-classification

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, creating a vulnerability surface for indirect prompt injection.\n
  • Ingestion points: Ingests external skeleton data files (.npy and .pkl) and DeepStream JSON files during training, evaluation, and inference as specified in SKILL.md and references/skill_info.yaml.\n
  • Boundary markers: The instructions do not define delimiters or explicit boundary markers to isolate the processed data from agent instructions.\n
  • Capability inventory: The skill utilizes the Bash tool to execute docker run commands and TAO CLI operations.\n
  • Sanitization: No specific sanitization or content validation procedures for the ingested dataset files are described in the agent instructions.\n- [EXTERNAL_DOWNLOADS]: The skill fetches a Docker container image from a remote registry to execute its core functions.\n
  • Evidence: The skill pulls the TAO toolkit image nvcr.io/nvstaging/tao/tao-toolkit-pyt:7.2.0-rc-36-multiarch from NVIDIA's registry as part of its standard workflow defined in SKILL.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 02:56 AM
Security Audit — agent-trust-hub — tao-train-pose-classification