Explicit checkpointing protocol for AI coding agents with test verification gates, dependency tracking, and rollback protocols. Prevents compounding errors.
We're going to build [PROJECT] incrementally. After each step:
CHECKPOINT PROTOCOL:
1. Summarize what was just implemented
2. Confirm tests pass (run: [test command])
3. List any new dependencies added
4. Note any deviations from the original plan
5. State what you recommend implementing next
ROLLBACK TRIGGER:
If at any point tests fail or behavior breaks:
COMMIT POINTS:
After each successful checkpoint, I'll tell you to "commit" - this is your signal that we've saved this state and can proceed.
Current checkpoint: [0/N]
Ready to begin?