reproduce-and-fix-issues

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

• Third-party content exposure detected (high risk: 0.85). The workflow reads the full source thread and tracker issue (outsider-authored Slack messages and GitHub issue text) and uses that readable prose in the agent’s context during “Study the report” (steps 6 and 9), creating an indirect prompt-injection risk.