nv-generate-mr

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The runtime path scripts/run_mr.py ingests outsider-provided JSON from the user-supplied model_config_override file (CLI argument model_config) via _load_config_override()/json.loads(), then passes rendered config into the upstream scripts.diff_model_infer subprocess; thus attacker-authored free text can reach runtime logic without selecting a specific trusted item first.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill and its runtime/setup explicitly fetch upstream code and model weights (which include executable inference code) from https://github.com/NVIDIA-Medtech/NV-Generate-CTMR.git and https://huggingface.co/nvidia/NV-Generate-MR, and the wrapper requires those external assets to run the upstream scripts that are executed at runtime.