ui-agent-browser

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

• Third-party content exposure detected (medium risk: 0.30). The UI Agent Browser runtime workflow explicitly uses agent-browser/Playwright to visit “the affected routes,” then reads rendered DOM/accessibility-tree and “interact with” and “inspect the data contract” (i.e., LLM ingests free text from pages provided by the target app/routes, which can include outsider-authored content).