three-game-starter
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill interviews the user and uses the answers to generate code and documentation, creating a potential surface for indirect injection if malicious input is provided.
- Ingestion points: User responses to the six-round interview defined in SKILL.md.
- Boundary markers: None explicitly defined in the instructions; responses are interpolated into the scaffold.
- Capability inventory: Extensive file-writing capabilities used to generate the project files in src/ and config directories.
- Sanitization: No explicit sanitization or validation of user-provided strings is mentioned before code generation.
- [EXTERNAL_DOWNLOADS]: The skill references and suggests installation of standard industry dependencies and CLI tools.
- Evidence: References well-known packages for Three.js development including three, @dimforge/rapier3d-compat, howler, tweakpane, postprocessing, gsap, sharp, and gltf-transform.
- Mitigation: The skill explicitly instructs the user to alias external CDN URLs (like those for opentype.js) to local bundled dependencies to prevent unauthorized external network requests.
Audit Metadata