The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of ZIP archives downloaded from the Meshy animation service.
Ingestion points: Animation ZIP files are downloaded to the user's local directory and subsequently unzipped and processed by the packager (WORKFLOW.md).
Boundary markers: No explicit markers or instructions to disregard embedded commands are defined for the data extracted from the external archives.
Capability inventory: The skill uses Python and Blender to process these files, write new files to the local project directories, and integrate assets into a Unity project (SKILL.md, WORKFLOW.md).
Sanitization: There is no evidence of content sanitization or safety validation for the metadata or structure within the downloaded ZIP content beyond the technical requirements of the packaging scripts.
[COMMAND_EXECUTION]: The skill invokes multiple local scripts using Blender and Python 3 to perform rigging, animation packaging, and deployment tasks.
Evidence: Commands are executed for humanoid_rig_input.py, humanoid_package.py, pipeline.py, and unity_install.py within the tools/3d/animation_quality/creatures/ directory (SKILL.md).
[EXTERNAL_DOWNLOADS]: The skill interaction model includes uploading project assets to and downloading generated content from a third-party service (meshy.ai).
Evidence: The workflow involves using a browser interaction tool to upload .glb files and download .zip archives from the Meshy platform (WORKFLOW.md).

humanoid-animation