comfyui-asset-generation
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: Several scripts (
assemble_segments.py,video_motion.py,prepare_h3_motion_reference.py, etc.) utilizesubprocess.runto executeffmpegandffprobe. These operations are limited to media processing and do not use a shell (shell=True), reducing the risk of command injection. - [EXTERNAL_DOWNLOADS]: The script
check_instrumental.pydownloads pre-trained audio classification models from HuggingFace (a well-known and trusted platform) to detect voices in background music. - [DYNAMIC_EXECUTION]: The script
part_views.pyusesimportlib.import_moduleto load backend-specific logic. This is restricted to a hardcoded whitelist of modules (ph3_paint,qwen_paint) and is not exploitable for arbitrary code execution. - [STATIC_ANALYSIS_FALSE_POSITIVE]: The static analyzer flagged a high-severity
eval()call inscripts/check_instrumental.py. Review of the code confirms this is actually a call to the.eval()method of a PyTorch/Transformers model object (e.g.,ASTForAudioClassification.from_pretrained(TAGGER).eval()), which switches the model to evaluation mode. It is NOT the dangerous Python built-ineval()function.
Audit Metadata