checkpoint

You can automate long-running project workflows by executing tasks end-to-end within a checkpoint, requiring human approval only at the checkpoint completion. This reduces manual coordination and drift by enforcing build, lint, and test guardrails before generating a final commit.

State-driven progress tracking resumes interrupted automation tasks by loading saved states from .planning/CHECKPOINTS.md and .planning/PRD.md files. This allows the workflow to pick up exactly where it left off across sessions without losing completed task history.

Yes, checkpoint automation requires a repository with .planning state files, specifically CHECKPOINTS.md and PRD.md. These files provide the necessary project state and milestone definitions to validate completion, generate diffs, and commit upon approval.

The best way to enforce build and lint guardrails is using a checkpoint workflow that automatically validates completion criteria before final commit. It enforces build, lint, and tests, ensuring code quality standards are met before allowing progression to the next checkpoint.

Yes, you can track checkpoint status across multiple sessions using the provided status dashboard. It offers visibility into active milestones and supports resuming interrupted checkpoints, ensuring long-running automation flows remain coordinated and recoverable.

Start the first checkpoint by issuing the /checkpoint start CP1 command. The system auto-runs the tasks to completion with built-in checks, and you approve the final commit when prompted, allowing automatic progression to the next checkpoint afterward.