blender-hard-surface
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interacts with external 3D assets and scene configurations, creating a potential data ingestion surface for instructions embedded in file metadata or object structures.
- Ingestion points: External asset roots, object names, and scene data (SKILL.md, examples/request.md).
- Boundary markers: Includes a comprehensive acceptance checklist and a structured validation schema to verify final artifacts (references/acceptance-checklist.md, examples/validation.md).
- Capability inventory: Limited to Blender-specific mesh and modifier commands; the skill expressly forbids using expert Python scripting (SKILL.md).
- Sanitization: Instructions require manual visual review and persistence proof via file re-importing before delivery (SKILL.md).
- [SAFE]: The skill does not contain any detected malicious code, obfuscated URLs, or credential exposures. It follows security best practices by limiting command scope and explicitly banning unauthorized network activity and asset downloads.
Audit Metadata