Advanced AI Prompt for Self-Hosted Runner Auto-Scaling for CI/CD

Act as a DevOps Scaling Architect. Design a highly elastic architecture for auto-scaling ephemeral GitHub Actions self-hosted runners operating on [CLOUD_PROVIDER] (e.g., AWS EC2 or Kubernetes) to absorb massive, unpredictable zero-downtime deployment loads. Architect a webhook-driven workflow or controller (e.g., Action Runner Controller - ARC) that continuously intercepts GitHub API events evaluating the exact queue depth of pending jobs. Based on that math, trigger an autoscaling event to dynamically provision runners within the [RUNNER_GROUP] boundary, strictly scaling between [MIN_RUNNERS] and [MAX_RUNNERS]. You must comprehensively detail the termination lifecycle: engineer the runners to run as 'ephemeral' (`--ephemeral` flag), ensuring they self-destruct and instantly terminate the cloud instance the millisecond their specific deployment task concludes, preventing astronomical idle costs.