optimize-3d-assets
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and processing of external 3D asset files (models, textures, and AI-generated geometry), which constitutes an untrusted data surface.
- Ingestion points: 3D asset files (GLTF, OBJ) and AI-generated models are accessed and audited via the
read_fileandsearch_codetools. - Boundary markers: The skill includes explicit instructions for "Security Guardrails" (ASI05, ASI07) which direct the agent to use strict parameterization and avoid loading from untrusted domains.
- Capability inventory: The agent is authorized to use
write_file,run_build, andrun_dev_server, providing a significant execution context for any successful exploit. - Sanitization: Instructions mandate schema validation for tool dependencies and loader configurations to prevent malicious configuration injection.
Audit Metadata