generate-design

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

• Third-party content exposure detected (high risk: 0.70). The skill explicitly downloads and reads/generated variant images returned by the third‑party Sivi API (https://connect.sivi.ai) and requires the agent to "read" and visually analyze those local image files (see SKILL.md Steps 5 and 7), and it also accepts arbitrary user-provided public image URLs to include in the API payload—so untrusted third‑party content is ingested and can influence agent behavior.