scenario-blender-geometry-nodes
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits a risk surface for indirect prompt injection as it processes external Blender scene data and user-supplied parameters to generate geometry and file-based outputs.\n
- Ingestion points: The skill ingests data through modifier inputs in the G.attach and G.set_inputs functions and reads attribute data from existing Blender objects via G.stats and G.attribute.\n
- Boundary markers: There are no specific delimiters or instructional guardrails implemented to prevent the agent from following directions potentially embedded in ingested string-based attributes or scene metadata.\n
- Capability inventory: The skill has the capability to write to the local file system via the G.snapshot function (which creates PNG images and legend text files) and to modify the internal Blender database through the Builder class.\n
- Sanitization: The skill relies on the standard Blender API for type validation but does not perform additional sanitization or escaping of ingested content before it influences subsequent agent actions or file writes.
Audit Metadata