sglang-humanize-review

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

• Third-party content exposure detected (medium risk: 0.30). At runtime, the workflow runs summarize_sglang_review_corpus.py/query_sglang_review_corpus.py, which read full human-authored GitHub PR review bodies from references/sglang-review-corpus.jsonl.gz (compiled from GitHub issue/PR comment and review text) and then incorporate that free text into the synthesized review response.