blender-character-workflow
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external 3D model files (FBX, GLB, Blend) using scripts/fit_vrchat_hair.py, which creates a surface where malicious instructions could be embedded in model metadata or object names to influence the agent's behavior.
- Ingestion points: scripts/fit_vrchat_hair.py (processes assets provided via --head and --hair arguments).
- Boundary markers: The instructions do not specify delimiters or warnings to isolate untrusted model data from the prompt context.
- Capability inventory: The skill can execute Blender API commands, perform geometry calculations, and write files to the local system (Blend, GLB, and JSON exports).
- Sanitization: Mesh names and properties are used directly in the fitting and export logic without explicit sanitization or validation.
Audit Metadata