axiom-vision-ref

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly ingests untrusted user-provided/third-party visual and textual content — e.g., VNRecognizeTextRequest, VNDetectBarcodesRequest, DataScannerViewController, and the Visual Intelligence SemanticContentDescriptor/IntentValueQuery flows described in SKILL.md — and the guidance shows the agent/app must read/interpret those inputs and act on decoded text/links/labels, so external content could influence behavior.