resume-from-checkpoint

To resume an interrupted pipeline, the skill reads the latest saved checkpoint from .claude/checkpoints, reconstructs the required state, reports the next phase, and prompts for user confirmation before re-spawning agents.

Yes, you can restore pipeline state after a session restart or /compact by loading the latest checkpoint, which identifies the next phase to run and retrieves key context like modified files and decisions for the next agent.

The checkpoint mechanism resolves the project root directory by reading the PIPELINE_CONFIG.md file when available, using it to locate the checkpoint storage path and reconstruct the execution state.

When resuming an interrupted automation workflow, the recovered context includes the next phase to run, files modified, prior decisions, prerequisites, and any unresolved alerts scanned before prompting for user confirmation.

Yes, user confirmation is required before re-spawning agents after a pipeline crash or unexpected session end, ensuring a safe resume by reviewing the reconstructed state and unresolved alerts before execution continues.

When loading a pipeline checkpoint, the skill scans for unresolved alerts and reports them alongside the next phase and key context, requiring user confirmation before proceeding to re-spawn the next agent.