living-backdrop
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches machine learning models from well-known repositories, specifically the Real-ESRGAN upscaler from GitHub and Depth-Anything from Hugging Face. These are legitimate dependencies required for the skill's primary image processing functionality.
- [DYNAMIC_EXECUTION]: Static analysis tools identified an
eval()call inscripts/upscale.py. Manual inspection confirms this is the PyTorch.eval()method used to switch a neural network into inference mode, which is a standard and safe operation in machine learning development. - [COMMAND_EXECUTION]: The skill uses a bash script (
scripts/build.sh) to orchestrate a pipeline of image processing tools including ImageMagick, AVIF, and WebP encoders. The script's operations are transparent and consistent with the task of preparing assets for web deployment.
Audit Metadata