research

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md Step 2 explicitly instructs the agent to "Conduct research using sub-agent(s) ... using web search or any other relevant tools," meaning it will fetch and interpret public/untrusted web content (web pages, citations) as part of its workflow and those sources can influence its actions and outputs.