Checkpoint
Checkpoint
A saved point in a task's progress that work can safely resume from
✓ Runtime & SessionWhat is a Checkpoint
A checkpoint is a saved point of known-good progress that work can resume from, rather than starting over from the beginning if something interrupts it. In practice this is often nothing more elaborate than a git commit after a working step, or a saved session state before attempting something risky, a deliberate point marked as safe to return to.
Checkpointing matters most on long or risky tasks: a multi-hour build, a large refactor, anything where an interruption partway through would otherwise mean redoing completed work. Committing at each stable step, rather than only once at the very end, is what makes an interrupted session recoverable instead of wasted.
A checkpoint is narrower than a handoff, a handoff carries context and reasoning to a different agent or session, while a checkpoint is mainly about progress: how far did the work get, and what state was it in when it got there. The two are often used together, a checkpoint marking where things stand, a handoff explaining why.
Build your own agent skills, not just prompts.
We design, build, and ship the Claude Code skill libraries production teams actually run on.
Talk to AY Automate ↗