skills/nvidia/skills/nv-reason-cxr/Gen Agent Trust Hub

nv-reason-cxr

Pass

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns or security risks were identified. The skill correctly implements security best practices, such as providing explicit medical disclaimers and limitations.
  • [EXTERNAL_DOWNLOADS]: The skill connects to Hugging Face (huggingface.co, *.hf.space) to download model weights or perform remote API inference. These are standard operations for an AI model wrapper and target well-known, trusted services.
  • [COMMAND_EXECUTION]: Static analysis flagged the use of subprocess.run() in tests/test_nv_reason_cxr.py; however, this is used exclusively for unit testing the runner script and does not pose a runtime risk to users.
  • [DYNAMIC_EXECUTION]: Static analysis flagged eval() and __import__() in scripts/run_nv_reason_cxr.py. These are false positives or benign in context:
  • The .eval() calls are PyTorch model methods used to switch the neural network into evaluation mode.
  • The __import__() call is used within a setup utility to check if required packages are properly installed before running inference.
  • [REMOTE_CODE_EXECUTION]: The skill manages dependencies using standard package managers (pip) targeting official registries. No unverified or suspicious remote script execution patterns were found.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 8, 2026, 07:15 PM
Security Audit — agent-trust-hub — nv-reason-cxr