vision-framework
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data sources such as images and live camera feeds for OCR and barcode detection, which represents an indirect prompt injection surface.
- Ingestion points: The skill utilizes
RecognizeTextRequestandDetectBarcodesRequestinreferences/vision-requests.mdto ingest and process text and barcode data from user-provided images or video buffers. - Capability inventory: Extracted data (text, URLs) is returned for use in the application, which could influence downstream agent behavior if the source images contain malicious instructions.
- Sanitization: No specific filtering or sanitization of the recognized text or barcode payloads is documented in the provided patterns.
- Boundary markers: There are no explicit instructions or delimiters shown to differentiate between intended data and potentially adversarial instructions embedded in processed images.
Audit Metadata