The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates remote code execution by instructing the agent to run uvx --python 3.11 blender-mcp. This command fetches and executes the blender-mcp package from a third-party repository (ahujasid/blender-mcp) at runtime.
[EXTERNAL_DOWNLOADS]: The instructions require the agent to download and enable a remote script (addon.py) within the Blender environment to establish a socket connection.
[COMMAND_EXECUTION]: The skill directs the agent to install system-level prerequisites including Blender 3+, Python 3.10+, and the uv tool. It also involves the execution of arbitrary Python scripts within Blender for asset rebuilding and exporting, which modifies the local environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection vulnerabilities.
Ingestion points: The agent processes external 3D models, textures, and reference images as part of the asset creation workflow.
Boundary markers: No delimiters or instructions to ignore embedded commands are specified for handling these external data sources.
Capability inventory: The skill utilizes subprocess calls via uvx, shell commands for software installation, and dynamic script execution within Blender.
Sanitization: No sanitization or validation of the external content is implemented before it is processed by the agent or the Blender environment.

blender-mcp