asset-optimization
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill functions by querying and modifying Blender scene data through the Model Context Protocol (MCP). This constitutes a data ingestion surface where external content (object names, metadata, or scene properties) is brought into the agent's context. While no specific attack content was found, the interaction with untrusted scene data presents a theoretical surface for indirect prompt injection.
- Ingestion points: MCP scene queries (polycount, material count, naming conventions).
- Boundary markers: None explicitly defined in the markdown instructions.
- Capability inventory: The skill uses MCP tools to query data and perform cleanup operations within the Blender environment.
- Sanitization: No specific sanitization or filtering of scene-derived data is described in the workflow.
- [NO_CODE]: The skill consists entirely of Markdown instructions and documentation. It does not include any executable scripts, shell commands, or source code files, which significantly reduces the direct attack surface.
Audit Metadata