The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied text ("world briefs") which it transforms into a structured world_spec intermediate.
Ingestion points: User input in the form of a world brief is used to define the scene parameters (SKILL.md).
Boundary markers: The instructions do not specify the use of delimiters or "ignore" instructions to prevent the agent from obeying commands embedded within a brief.
Capability inventory: The skill utilizes the ThreeJSWorld tool to perform file writes and environment builds (SKILL.md).
Sanitization: There is no evidence of sanitization or filtering logic applied to the user input before it is serialized into the world_spec JSON.
[DYNAMIC_EXECUTION]: The skill involves the runtime generation of executable JavaScript and HTML files to render the 3D world.
Evidence: The build operation writes files including index.html and world-runtime.js to the project directory (references/world-spec.md).
Context: These generated files contain the logic for Three.js scene construction. While this is the intended purpose of the skill, generating executable scripts from data influenced by user input creates a dynamic execution surface.

threejs-world-generation