alpha-scene-gen
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill constructs and executes Python scripts within a Blender environment using the
execute_blender_codetool. It provides templates inreferences/blender_helpers.mdfor downloading, sanitizing, and positioning 3D models. - [EXTERNAL_DOWNLOADS]: The skill uses
urllib.request.urlopento fetch GLB assets from remote URLs provided by the configured 3D generation services (Alpha3D, Meshy, or Tripo). - [INDIRECT_PROMPT_INJECTION]: The skill processes natural language scene descriptions to generate planning data and Blender code. 1. Ingestion points: User-provided scene and asset descriptions processed in
SKILL.md. 2. Boundary markers: The skill includes a mandatory 'Show the plan, then STOP for explicit confirmation' step inSKILL.mdto mitigate unintended execution. 3. Capability inventory: The skill has the ability to execute arbitrary Python in Blender and perform network operations viareferences/blender_helpers.md. 4. Sanitization: No programmatic sanitization of the natural language input is performed prior to plan generation.
Audit Metadata