What problem does it solve? Sandbox workloads stall when warm pools are undersized, autoscaling bounds are misconfigured, or quota ceilings block growth, leaving sandboxes queued or failing to start. This Skill guides you through sizing pools, tuning autoscaling groups, reading quota, and recovering pools that refuse to grow. ## Core Features & Use Cases - Pool Sizing: Set replicas directly on fixed pools or edit autoscaling group bounds when the autoscaler owns the size, using abx scale and abx update commands. - Stuck Pool Recovery: Diagnose why a pool will not grow by checking pool status, quota, and cluster events, then apply the counter-intuitive fix of reducing the replica target before climbing again. - Capacity Planning: Plan for N concurrent sandboxes by inspecting instance types, committed quota, and peak concurrency rather than total work. - Use Case: A reinforcement learning rollout needs 50 concurrent sandboxes but they keep queuing. Use this Skill to check the pool's desired versus actual replicas, confirm quota headroom, and set the group's maxReplicas to the true peak. ## Quick Start Ask the AI to check why the demo-1c2gi pool in the demo environment is stuck below its target size and fix the capacity.
