babysit-prs

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime loop fetches and reads full PR comment bodies authored by external users (e.g., via fetch-all-pr-comments.sh), then performs untrusted “finding extraction” and per-finding classification on that text without requiring the agent to first select a specific user-posted item.