Long Task Coordinator

If your work spans multiple sessions, gets handed off to other agents, or needs to survive interruptions, this keeps a single durable state file as the source of truth instead of relying on chat history. It enforces a clear cycle: read state, recover if needed, decide next action, persist changes, then report. The key insight is treating "awaiting result" as a legitimate status rather than a failure, and separating recovery logic from actual domain work. Skip it for quick single-turn tasks, but reach for it when losing context mid-stream would actually hurt. Works with Read, Write, Edit, Bash, and the usual file tools.