The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external media files which serve as ingestion points for untrusted data, though the risk is inherent to the skill's primary purpose.
Ingestion points: Processes user-provided image files (e.g., raw.png) and 3D models (e.g., unit.glb) in SKILL.md.
Capability inventory: The skill uses the um CLI tool to perform file system operations and network requests to remote services.
Boundary markers: The instructions do not specify delimiters or safety warnings for the agent when handling potentially malicious file content or metadata.
Sanitization: No explicit sanitization or validation of input files is mentioned prior to processing by the toolchain.
[EXTERNAL_DOWNLOADS]: The skill documents dependencies on external third-party software that must be present on the execution environment.
Software requirements: References the need for Blender, ImageMagick (magick), and DirectXTex (texconv) for rendering and conversion tasks.
[COMMAND_EXECUTION]: The skill relies on shell command execution to perform asset processing.
Command usage: Extensively uses the um utility with various subcommands (sprite, render3d, fal) to manipulate local files and interact with remote APIs for 3D remeshing.

asset-pipeline