What problem does it solve? Operational procedures often live in scattered tickets, logs, and stale documents, making it risky to execute maintenance, recovery, or migration events without a verified, step-by-step plan. This Skill converts validated operational knowledge into a bounded, versioned runbook where every mutating step carries its authority, target, verification, and failure handling. ## Core Features & Use Cases - Step-level execution states: Every step is labeled validated, partially-validated, or proposed, bound to evidence, environment, version, and date. - Safety gates and failure modeling: Preflight checks, abort criteria, decision points, partial-failure reconciliation, and distinct rollback versus recovery paths. - Secret and target hygiene: Credentials become named placeholders, and destructive or broad targets are defensively validated before inclusion. - Use Case: An operator preparing a database migration asks for a runbook covering preflight backups, staged mutations, verification queries, and an escalation path, with untested steps clearly flagged rather than presented as proven. ## Quick Start Use the sd-runbook skill to create a runbook for procedure=database failover with trigger=primary node unreachable and environment=production cluster, marking any untested recovery steps as proposed.
