ralph

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

• Third-party content exposure detected (high risk: 0.90). The workflow explicitly handles "URL-based visual cloning tasks" (Step 5) — e.g., "clone https://example.com" — and mandates routing such targets through Visual Ralph, which requires fetching and interpreting live public web pages/screenshots as part of the verification/edit loop, so untrusted third-party content can influence agent decisions and actions.