vision-understanding
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security vulnerabilities were identified. The skill is instructional, providing architectural guidance for routing media tasks to appropriate models (e.g., Qwen, Gemini) and interacting with the vendor's API endpoint (
api.runware.ai). - [PROMPT_INJECTION]: The skill facilitates processing untrusted external media (images, video, audio) which represents a surface for indirect prompt injection. The skill includes specific mitigation instructions for the agent to 'Stay grounded' and only report what is directly visible or audible, reducing the risk of the model following instructions embedded within the processed media.
Audit Metadata