The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external 3D assets (CAD files), which constitutes a potential attack surface for indirect injection via file metadata or structures.
Ingestion points: The --input command-line parameter defined in the argparse block of all reference scripts (e.g., examples/import-tessellate-export-glb.py).
Boundary markers: The skill mandates using argparse as the exclusive input channel to ensure data is treated as parameters rather than executable instructions.
Capability inventory: The scripts utilize the pxz SDK for 3D processing and the standard os module for directory management (os.makedirs).
Sanitization: Input paths are normalized using os.path.abspath before being passed to the SDK operations.
[COMMAND_EXECUTION]: The example scripts perform necessary file system operations, such as creating output directories via os.makedirs. These operations are well-scoped to the /workspace convention described in the documentation and are required for the skill's primary function of asset transformation.

at-upa-scripting