alt-text
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local image data which may contain embedded text instructions (OCR). An adversary could provide an image with text that attempts to override the skill's constraints or manipulate the output.
- Ingestion points: Local image files specified by the path in
$0or image attachments. - Boundary markers: None. The instructions do not explicitly tell the agent to ignore instructions found within the image itself.
- Capability inventory: Relies on the agent's native vision and text generation capabilities.
- Sanitization: None. The skill assumes the image content is safe to describe.
Audit Metadata