Manus AI Agent Integration

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly lets the Manus agent browse and scrape the open web and ingest arbitrary URLs (Create Task attachments with "type":"url"), and its playbooks/examples mention scraping App Store reviews and Reddit sentiment analysis—i.e., untrusted public/user-generated content is fetched and interpreted as part of the agent's workflow.