blender-pipeline
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest, store, and manage metadata within Blender scenes and JSON manifest files, creating a potential surface for indirect prompt injection if that metadata is later processed by an agent without sanitization.
- Ingestion points: Tool parameters in
tag_asset_metadata.py,set_project_context.py, andcreate_publish_manifest.pyallow for arbitrary metadata ingestion. - Boundary markers: There are no specific delimiters or markers provided to isolate ingested metadata from agent instructions.
- Capability inventory: The skill can modify Blender scene properties and write local JSON manifests.
- Sanitization: Metadata is stored as provided without explicit sanitization, which is expected for the tool's primary purpose but identifies the ingestion surface.
Audit Metadata