see-through-anime-layer-decomposition

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches core machine learning libraries (PyTorch, TorchVision) from official distribution points and downloads pre-trained models from HuggingFace repositories (layerdifforg, 24yearsold).- [COMMAND_EXECUTION]: Orchestrates the image processing pipeline by executing local Python scripts (inference_psd.py, heuristic_partseg.py) through the subprocess.run interface.- [DYNAMIC_EXECUTION]: Implements dynamic module loading using importlib.util to access local body tag definitions stored within the common/live2d/scrap_model.py file.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted image data for decomposition (Ingestion: --srcp parameter in inference_psd.py). Boundary markers are absent. The skill possesses capabilities for file system writes to workspace/ and subprocess execution. Input validation is focused on file paths and image formats, typical for this domain.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 10:36 AM