autopilot-recovery

To recover an automation session after a crash, scan existing checkpoints to evaluate validity and select the optimal resume point. The workflow validates state-snapshot integrity and reconstructs the phase context to minimize lost progress.

State-snapshot validation works by checking the state-snapshot.json file and verifying the anchor SHA during recovery. This deterministic process identifies gaps or git risks before safely restoring the session to a valid resume point.

The best way to resume an autopilot workflow from a corrupted state is to use fallback phase-context snapshots. If the primary state-snapshot fails validation, the recovery logic reconstructs the session context from these phase-specific snapshots.

Yes, you can automate cleanup of recovery artifacts when restoring a session. The mode-aware restoration process includes optional cleanups and user prompts to safely remove obsolete checkpoints and maintain workflow integrity.

You should use manual recovery decisions instead of automatic restoration when complex git risks or checkpoint gaps are detected. The deterministic recovery logic provides guidance for evaluating these edge cases before safely proceeding.

Yes, crash-safe orchestration requires git anchors for state snapshots. The recovery-decision logic validates state-snapshot data against an anchor SHA to ensure checkpoint integrity and prevent replaying corrupted workflow states.