manus

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly performs web browsing and "gathers and combines information from multiple websites" (SKILL.md) and accepts arbitrary URL attachments/connectors ("File Attachments" and "connectors"/"createShareableLink" in SKILL.md and references/api.md), so it fetches untrusted public content that the agent reads and uses to drive its research and actions.