addsearch

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly directs the agent to interact with AddSearch data (a hosted search over external websites) via Membrane actions (e.g., "membrane action run ... --connectionId=CONNECTION_ID"), which will fetch and return arbitrary public website content that the agent is expected to read and could materially influence subsequent actions.