nv-generate-mr-brain-finetune
Pass
Audited by Gen Agent Trust Hub on Jul 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill's primary script, scripts/run_mr_brain_finetune.py, orchestrates a multi-step machine learning workflow by executing Python modules (embedding creation, training, and inference) using the subprocess module. Commands are passed as lists of arguments with shell=False, preventing shell-injection vulnerabilities.
- [EXTERNAL_DOWNLOADS]: The skill manifest explicitly references and downloads resources from official vendor-controlled infrastructure, including cloning the NV-Generate-CTMR repository from NVIDIA's GitHub organization and fetching model weights from the nvidia/NV-Generate-MR-Brain repository on Hugging Face.
- [SAFE]: The skill performs appropriate data validation on user-supplied JSON datalists, checking file types and ensuring image paths are relative to the provided base directory. No evidence of credential exposure, malicious persistence, or unauthorized data exfiltration was found during the analysis.
Audit Metadata