maintaining-long-running-task-continuity

What problem does it solve? Long-running tasks often break when they cross response, tool, context, or execution boundaries, losing progress and forcing work to restart. This Skill structures large tasks into verifiable outcome units with resumable checkpoints so interrupted work can resume from evidence instead of re-reading entire conversation history. ## Core Features & Use Cases - Checkpoint Contract: Records objective, completed outcomes, changed artifacts, validation evidence, blockers, protected decisions, and the next exact action in a structured YAML format. - Mode-Based Workflow: Follows an initialize → checkpoint → resume → partial-delivery → close lifecycle to manage task state explicitly. - Partial Delivery: When execution is blocked, delivers completed results first and separates unfinished work, causes, and resume points. - Use Case: During a multi-hour repository migration, the AI checkpoints progress every few tool batches; if the session ends, a new session reads only the latest checkpoint and authoritative sources to resume exactly where work stopped. ## Quick Start Break this large task into verifiable outcome units and maintain a resumable checkpoint after each meaningful step so I can resume if the session is interrupted.