The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to download a Python script using curl from a third-party GitHub repository (https://raw.githubusercontent.com/ahujasid/blender-mcp/main/addon.py) and install it as a Blender addon. This source is not verified or trusted.
[REMOTE_CODE_EXECUTION]: The core functionality of the skill includes an execute_code command designed to run arbitrary Python (bpy) code. Since Blender's Python environment typically has access to the full Python standard library (including os and subprocess), this allows for arbitrary command execution on the host machine.
[DYNAMIC_EXECUTION]: The skill provides a Python helper function (blender_exec) that programmatically constructs JSON payloads containing arbitrary code and transmits them over a TCP socket to be executed by the Blender instance.
[INDIRECT_PROMPT_INJECTION]: The skill creates a vulnerability surface by processing user-provided instructions and converting them into executable Python code without sanitization.
Ingestion points: The skill receives natural language instructions from the agent to perform Blender actions.
Boundary markers: No delimiters or safety instructions are provided to prevent the agent from being coerced into generating malicious Python payloads.
Capability inventory: The skill possesses the capability to execute any Python code, write files (e.g., render outputs), and perform network operations via sockets.
Sanitization: There is no evidence of validation, escaping, or filtering for the code strings passed to the execute_code tool.

blender-mcp