pr-loop

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

• Third-party content exposure detected (high risk: 0.85). The pr-loop runtime workflow ingests PR reviewer content by calling pr_status.dart, which fetches PR GraphQL data (including graphData.comments and graphData.reviewThreads) from GitHub; those comments/threads are authored by outsiders (e.g., gemini-code-assist/gemini-code-review and other reviewers) and are read as free-text comment.body into the agent’s LLM context for decision-making.