review-training-data-quality

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

• Third-party content exposure detected (medium risk: 0.30). The runtime path scripts/summarize_jsonl_fields.py reads and JSON-parses the provided corpus.jsonl/input file at args.input.read_text(...).splitlines(), so outsider-authored free text inside that corpus can be ingested without selecting a specific item first.