blender-mcp-setup
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill instructions require the agent to generate and execute Python code snippets within the user's Blender environment using the
mcp__blender__execute_blender_codetool. While this is the intended functionality for scene management, runtime execution of dynamically generated code is a high-capability operation. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external Blender scene data and
.blendfiles which may originate from untrusted sources. - Ingestion points: Blender scene data, objects (
bpy.data.objects), and file metadata (bpy.data.filepath) are read into the agent's context. - Boundary markers: The instructions explicitly advise following a safety protocol (
blender-content-safety-and-privacy) when handling untrusted files, including using isolated profiles. - Capability inventory: The skill possesses code execution capabilities (
mcp__blender__execute_blender_code) and visualization capabilities (mcp__blender__get_viewport_screenshot). - Sanitization: The instructions recommend disabling automatic script execution in Blender for untrusted scenes.
- [COMMAND_EXECUTION]: The skill uses specific tool calls to execute logic and inspect the system state within the host Blender application, accessing environment details like version strings and active render engines.
Audit Metadata