codex-goal-mining

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

• Third-party content exposure detected (medium risk: 0.30). In scripts/codex-goal-report.py, the workflow reads outsider-authored free text from user/session-generated JSONL logs (candidate_goal_files() → collect_local_goals() → json.loads(line) and extracting goal.objective), which can be posted/poisoned by an external operator into Codex run history that the script then ingests.