session-evidence

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow runs sessions analyze ... --include-turns --extract-only --turn-query ... which ingests and returns transcript snippets/artifacts from cached session transcripts (i.e., free-form text authored by other participants in those sessions, not chosen by the operating user at query time), and that text is then placed into the agent’s LLM context via the tool output.