fal-vision

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

• Third-party content exposure detected (high risk: 0.90). This skill's analyze.sh explicitly accepts arbitrary --image-url values and sends those public images to fal.run model endpoints (scripts/analyze.sh builds a payload with "image_url" and posts it), so untrusted, user-provided web content (images/embedded text) is ingested and can materially influence model outputs and subsequent actions.