The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes code snippets that fetch decoder and transcoder libraries from well-known services including Google GStatic and JSDelivr. These are standard external resources required for handling Draco and KTX2 compressed assets in Three.js applications.
[INDIRECT_PROMPT_INJECTION]: The skill defines a vulnerability surface by instructing the agent on how to ingest and process external data such as GLTF models and textures.
Ingestion points: External assets are loaded via GLTFLoader.load(), TextureLoader.load(), and fetch() in SKILL.md.
Boundary markers: The instructions do not define delimiters or specific safety instructions to ignore embedded content within the processed assets.
Capability inventory: The skill allows for 3D scene manipulation and network reads but does not include scripts with shell execution, file system writes, or sensitive data access across its files.
Sanitization: There is no mention of sanitizing or validating asset metadata before processing.
[METADATA_POISONING]: The skill's frontmatter contains a risk: critical field which is deceptive and inconsistent with the actual content of the skill, which is a benign technical documentation resource.

threejs-loaders