The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the blender command-line interface in background mode to perform scene exports, which is a recommended practice to avoid MCP server timeouts during long-running operations.
[EXTERNAL_DOWNLOADS]: The instructions recommend the use of @gltf-transform/cli for optimizing 3D models. This package is managed by the official npm registry, which is a well-known and trusted service in the development ecosystem.
[DYNAMIC_EXECUTION]: Programmatic control of Blender is achieved through the execute_python tool and the blender --python-expr flag. These patterns are central to the skill's primary purpose and are implemented using standard Python and Blender API conventions.
[INDIRECT_PROMPT_INJECTION]: The skill extracts metadata from Blender scenes, such as object names, material properties, and animation track data. While these represent ingestion points for potentially untrusted data, the risk is assessed as low due to the use of standard formatting and the constraint of the execution environment.
Ingestion points: Scene hierarchy data, material node configurations, and animation track metadata extracted via the Blender API.
Boundary markers: Not explicitly implemented in the example script templates.
Capability inventory: Python API access within Blender, CLI execution for export tasks, and use of npx for external optimization tools.
Sanitization: Example scripts utilize json.dumps() to ensure that extracted scene data is properly escaped before being presented to the agent.

blender-mcp