handoff

To persist project state across multiple agent sessions, you need a durable external store that coordinates checkpoints. This skill enforces a state machine and append-only logging to keep your goal, next action, and evidence intact across restarts.

You need a state machine for long-running agent tasks when work spans multiple sessions and requires persistent progress tracking. It enforces structured phases like orient, work, checkpoint, done, and blocked to guide autonomous or human-guided operations.

To prevent state drift when resuming work from an external store, apply drift checks and startup smoke tests. This validates that the anchored goal and append-only logs remain consistent with the actual project state before continuing.

Yes, you can use GitHub issues as an external store for agent checkpoints. The system supports anchoring progress to external stores like GitHub issues or repo docs to maintain durable project state across different harnesses.

The best way to handle blocked states in multi-session workflows is through a structured state machine that explicitly logs the blockage. Append-only logs and structured checkpoints ensure the dead-end list and context survive session restarts.

You can set up durable project state by symlinking the skill folder into your agent's skills directory and starting a session. This initializes the storage ladder, supporting multiple storage modes to suit different harnesses without requiring external dependencies.