research-session

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

• Third-party content exposure detected (high risk: 0.80). The skill's Phase 3 routing explicitly tells the agent to "Invoke the /paper-read skill. Pass through any paper reference the user provides" and to start a /lit-search workspace for a topic, which clearly allows ingestion/processing of arbitrary external paper URLs or open-web literature search results (untrusted third‑party content) as part of the workflow.