model-characters

Pass

Audited by Gen Agent Trust Hub on Nov 18, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The Python scripts (read_character_ref.py, check_character.py) list numpy and pillow as dependencies. These are well-known, standard libraries from official registries and are used for legitimate image and data processing.
  • [COMMAND_EXECUTION]: The skill instructions use uv run --no-project to execute internal scripts for character reference analysis and model checking. These operations are local, performed on project files, and do not involve network activity or privileged commands.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data, including 2D images and 3D model files (GLB and JSON).
  • Ingestion points: read_character_ref.py processes image files; check_character.py parses GLB binary data and JSON scene recipes.
  • Boundary markers: Not explicitly defined in the scripts, but the output is structured JSON or generated image guides.
  • Capability inventory: The scripts perform mathematical and geometric analysis using NumPy and PIL; they do not execute shell commands or make network requests based on input content.
  • Sanitization: Content is processed as raw bytes/pixels for geometric validation; no Natural Language instructions are extracted or executed from the data.
  • [SAFE]: No malicious patterns such as obfuscation, credential theft, or persistence mechanisms were detected. The skill follows established best practices for 3D modeling workflows.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 18, 2026, 06:32 AM