finetuning

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

• Third-party content exposure detected (low risk: 0.10). Outsider-authored free text is ingested only from user-provided/local JSONL files into scripts like scripts/calibrate_grader.py (it reads args.data JSONL lines and feeds messages[-1]["content"] plus model outputs to the grader), but there is no external “publish/queue/feed” path that the workflow automatically fetches outsider content from at runtime.