The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements direct dynamic code execution via the exec() and compile() functions. The execute_python tool (in scripts/execute_python.py) and the execute_script_file tool (in scripts/execute_script_file.py) allow the execution of arbitrary Python logic provided as a string or read from a file. This bypasses standard tool boundaries and gives the agent full access to the Blender Python API (bpy).
[COMMAND_EXECUTION]: By design, the skill allows the agent to execute arbitrary commands within the host's Python interpreter. This grants the capability to perform any action the Blender process is permitted to do, including modifying the filesystem, spawning subprocesses, or altering the host environment.
[DATA_EXFILTRATION]: The execute_script_file tool can be used to read the contents of any file the process has access to (scripts/execute_script_file.py lines 39-41). Additionally, the get_blender_info tool performs reconnaissance by exposing sensitive environment metadata, including the binary_path, python_executable path, and platform details (scripts/get_blender_info.py lines 21-30).

blender-scripting