dag-task-runner

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

• Third-party content exposure detected (medium risk: 0.30). The runner ingests the user-provided task text indirectly by having the parent agent generate subtask_prompt and then passes that free-text into each subagent via agent.send(stitched) in scripts/run_dag.ts, so outsider-authored text can reach the LLM without selecting a specific item first.