What problem does it solve? Commands, lifecycle steps, and processing loops run in environments where crashes, restarts, and retries are normal. When partial state from a failed run changes the next run's outcome, every restart becomes a debugging session. This Skill provides a design principle for making operations safe to re-run. ## Core Features & Use Cases - Convergent Startup: Scan for existing state, clean stale artifacts, and adopt live sessions instead of assuming a clean slate. - Content-Based Cleanup: Compare artifacts by content equivalence rather than creation order to avoid duplicate or orphaned state. - Self-Healing Locks: Use PID-based stale lock detection so crashed processes do not permanently block future runs. - Idempotent Scheduling: Respawn failed work cleanly and regenerate fresh input after each cycle. - Use Case: When writing a CLI command that provisions resources or a worker loop that processes a queue, apply the three-question test (runs twice, crashes at any point, converges to the same end state) to identify where a reconciliation step is needed. ## Quick Start Ask the AI to review your command, lifecycle step, or processing loop for idempotency and identify where partial state could change the outcome of a re-run.
