pr-merge-train

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md:28-31 and 97-125 describe runtime reads of outsider-authored GitHub PR content (PR titles/bodies, author, review threads/comments) via GitHub CLI/API (e.g., gh pr list and GraphQL discovery), which can introduce free text from other people into the agent’s LLM context.