world-labs-marble
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructions create a potential surface for indirect prompt injection by detailing the ingestion and processing of multimodal user data to generate 3D assets.
- Ingestion points: The skill describes workflows for processing untrusted text prompts, images, panoramas, and video clips, as defined in the 'Input modes and their constraints' section of SKILL.md.
- Boundary markers: The provided instructions lack specific guidance on using delimiters or system-level constraints to isolate untrusted data from agent commands.
- Capability inventory: The skill defines capabilities for network-based world generation and modification via the World API (api.worldlabs.ai).
- Sanitization: No input validation, filtering, or sanitization steps are included in the documented workflows for handling user-supplied content.
Audit Metadata