blender-preview
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests visual data from Blender renders to perform visual diagnosis, which represents a potential attack surface for instructions embedded within the generated images. Ingestion points: Blender camera, front, side, and top view PNGs (SKILL.md). Boundary markers: None defined for the description of visible defects. Capability inventory: File existence verification and SHA-256 hashing (SKILL.md). Sanitization: Not applicable for visual preview descriptions.
- [SAFE]: The skill enforces file integrity by requiring SHA-256 validation for all captured previews before they are displayed.
- [SAFE]: No unauthorized network communication or external code execution patterns were identified.
Audit Metadata