yao-image
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides tools for analyzing image content, which creates a surface for indirect prompt injection via text or instructions embedded within images or hosted at external URLs.\n
- Ingestion points: Data enters the agent context through the image_path and mask parameters in the image_read and image_edit tools (SKILL.md).\n
- Boundary markers: The instructions do not specify any markers or negative constraints to distinguish between tool instructions and content found within images.\n
- Capability inventory: The skill facilitates calls to external vision and generation models via the tai tool command, which involves sending data to external providers (SKILL.md).\n
- Sanitization: There are no instructions for validating or sanitizing the content of the images or the domains of the URLs provided.
Audit Metadata