retro

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

• Third-party content exposure detected (high risk: 0.75). Retro’s workflow ingests outsider-authored free text when it runs t3 <overlay> retro review-findings <pr-url>: it fetches PR review comments from the forge client at runtime and then reads/classifies that comment text to generate findings, which is then used to drive LLM context/output.