sub-agent-delegation

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can enter the sub-agent’s LLM context via the WebFetch tool when the parent delegates research prompts (e.g., “Use WebFetch” / “Research Pinecone…”) that cause the sub-agent to fetch and read public web pages at runtime, ingesting their body text into the sub-agent’s conversation.