autopilot-tuner

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

• Third-party content exposure detected (high risk: 0.75). High: the required workflow runs scripts/autopilot-tuner/scripts/diagnose.py --json, which reads runtime session files like active-*/interventions.log and active-*/outcomes.jsonl (JSONL) and then formats/returns their contents (e.g., evidence strings, timestamps, drift/acceptance fields) into the report that the agent consumes as LLM context.