automationlisted

# Purpose You convert manual operational procedures (runbooks) into parameterized, idempotent scripts with safety gates, and optionally graduate them to self-healing controllers. You also design and produce chaos experiments against the project's cloud infrastructure (see `.claude/project.json` -> `cloud.*`) with mandatory environment guards. All output is executable shell scripts or Python automation code with dry-run, confirmation, and rollback steps. The examples below use Kubernetes/kubectl -- adapt commands to the project's runtime (`cloud.runtime`). Related skills: `code-standards` (for script quality), `api-integration` (if the runbook touches the main app's APIs). # When to use - Trigger A -- A manual runbook exists (documented or ad-hoc) that is executed more than twice per week - Trigger B -- An operational task involves restarting pods, flushing caches, scaling deployments, or rotating secrets - Trigger C -- A chaos experiment is needed to validate system resilience (pod kill, network partition, resource exhaustion) - Trigger D -- A toil reduction analysis is requested for a specific operational area # When NOT to use - Writing or modifying CI/CD pipelines (GitLab CI, GitHub Actions) -- use `deployment` - Authoring or updating deployment manifests or infrastructure templates -- use `deployment` - Writing application code (TypeScript, Python) for the main app or the device/edge repo -- use appropriate application skills - Ad-hoc debugging of a production incide