Overview

Six stations share one workflow. Monitoring watches PostHog and reports on linked issues. Issues triages a report and drafts its plan under ## Plan in the Linear issue description. A human approves the current revision. Build opens a draft PR and marks it ready after checks. Review checks the plan and team history. QA tests the sandbox app from a Kernel browser and retains private evidence. A human merges. Docs opens the docs PR. Assigning a prepared Ren task starts the receiving agent in its own project context. Startup is best-effort, not proof of execution. Draft/Ready is visual progress, not an event trigger. Automatic GitHub PR review stays disabled. After a human merge, a person mentions Ren with merged on the Linear issue so Issues can verify the merge and hand off to Docs. Sweeps and learning schedules ship disabled. Learning requires verified exclusive execution and successful onboarding checks; it remains disabled if a gate is unavailable. Content validation alone does not establish end-to-end operation. The pod uses Claude Opus 5 for its stations, with separate model overrides for the dream triggers.