tao-finetune-anomalygennext
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a bash wrapper to execute
torchrunand Python-based validation scripts. All inputs to these commands are validated using regex patterns and canonical path resolution. - [EXTERNAL_DOWNLOADS]: The skill utilizes an official NVIDIA container image (
nvcr.io/nvidia/paidf-anomalygen:1.1.0) and requires assets from Hugging Face. These are verified as legitimate vendor-supported resources. - [DYNAMIC_EXECUTION]: The test suite employs
subprocess.runto verify skill behavior. This is limited to local script execution for testing purposes and does not impact the production runtime. - [SAFE]: Extensive validation is performed on user datasets, including file extension white-listing and binary mask verification using the Pillow library, preventing the processing of malformed or malicious data.
Audit Metadata