threejs-world-generation
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text briefs which are used to populate
explicit_constraints. This represents a surface for indirect prompt injection where untrusted input could attempt to influence the agent's behavior. - Ingestion points: User-provided world briefs (SKILL.md Step 1).
- Boundary markers: The instructions require explicit separation of user-supplied facts (
explicit_constraints) from agent-inferred parameters (inferred_details). - Capability inventory: The agent can execute the
threejs_worldtool and write structured specification files (world.json,index.html) to the project workspace. - Sanitization: The skill does not define specific sanitization routines for input text, but the data is constrained by a strict JSON schema for 3D world parameters.
- [REMOTE_CODE_EXECUTION]: The skill uses the
threejs_asset_catalogtool to download 3D assets (GLTF/GLB models). These are data assets for rendering and do not constitute executable code. The skill explicitly prohibits remote model fetching at runtime to ensure determinism. - [DATA_EXPOSURE_AND_EXFILTRATION]: No patterns for accessing sensitive system files or exfiltrating data to external domains were found. The skill operates within a defined
projects/<id>/workspace.
Audit Metadata