The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages Blender production stages using a Python manifest script (scripts/run_manifest.py). This script includes a security function _inside() that validates all file paths against a root directory using os.path.realpath and os.path.commonpath. This effectively prevents path traversal attacks by ensuring the agent cannot write to or read from arbitrary locations on the host system.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data, such as reference images and asset descriptions. It mitigates potential injection risks by enforcing a rigid, multi-stage workflow that requires the agent to generate a structured "Production Brief" and "Reference Analysis" before any operations are executed, providing clear boundaries for user-supplied data.
[DATA_EXPOSURE]: Analysis of the included scripts and instructions confirms that the skill does not attempt to access sensitive files (e.g., SSH keys, AWS credentials), environment variables, or other private user data. All operations are confined to specific project directories.

blender-director