nv-generate-mr-brain

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

• Third-party content exposure detected (high risk: 0.85). The runtime reads the user-supplied model_config_override JSON file (positional PATH_TO_MR_BRAIN_CONFIG.json) via _load_config_override() and then includes the parsed override content in the emitted JSON payload (input.model_config_override) that is returned to the agent/LLM context.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill requires and may fetch/execute upstream code and weights at runtime from https://github.com/NVIDIA-Medtech/NV-Generate-CTMR (git clone / upstream scripts.diff_model_infer executed) and from https://huggingface.co/nvidia/NV-Generate-MR-Brain (downloaded model weights via scripts.download_model_data), so remote content is fetched and executed and is a required dependency.