The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documentation describes how to load external 3D models and textures from external sources, which is a standard surface for processing untrusted data.
Ingestion points: Demonstrated through the use of GLTF, FBX, OBJ, and SVG loaders in references/02-loaders.md and references/15-specialized-loaders.md.
Boundary markers: Code snippets focus on functional implementation and do not include specific boundary delimiters for untrusted asset content.
Capability inventory: Loaded assets are processed by the Three.js rendering engine and animation mixer for display in the browser.
Sanitization: No custom sanitization logic is shown for model data, relying on standard Three.js loader parsing behavior.
[DYNAMIC_EXECUTION]: The skill covers the creation and compilation of custom GLSL and WGSL shader programs at runtime, which is core to 3D graphics development.
Evidence: references/11-materials-advanced.md and references/13-node-materials.md provide examples of defining shader strings and node-based logic that are compiled by the GPU via Three.js during application execution.

threejs