The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/batch_process.py contains a function _module that uses importlib to dynamically load and execute Python modules from a file path provided via the --script CLI argument.
Evidence: The _module function uses importlib.util.spec_from_file_location and spec.loader.exec_module to load code from a path that is not statically defined.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it processes external files (e.g., .fbx, .blend) and can execute scripts on their data.
Ingestion points: File and directory paths passed to the --in argument in scripts/batch_process.py.
Boundary markers: None identified in the script or instructions to warn the agent about embedded instructions in asset metadata.
Capability inventory: File system access (read/write) and execution of Python code within the Blender environment.
Sanitization: No content validation or sanitization of input file data is performed before processing.
[EXTERNAL_DOWNLOADS]: The SKILL.md file provides instructions for CI/CD pipelines to download the official Blender Linux distribution from a well-known service.
Evidence: Reference to https://download.blender.org/release/Blender5.2/blender-5.2.2-linux-x64.tar.xz in the CI recipe section.

blender-pipeline-automation