What problem does it solve? Large ICFP Contest searches—layout exploration, module synthesis, batch public-case checks—are too slow or memory-heavy for one local machine. This Skill snapshots the repository working tree to S3 and runs the work as parallel AWS Batch EC2 Spot array jobs, then collects every per-index artifact. ## Core Features & Use Cases - Environment health checks: Verifies the CloudFormation stack, compute environment, job queue, and job definition are healthy before submitting work. - Cost and sharding guidance: Estimates vCPU-hours and USD cost, and decides whether work should run locally or as an array job with per-shard vCPU/memory sizing. - Job lifecycle management: Submits array jobs, monitors status, waits for completion, downloads complete per-index results, and cancels only your own jobs. - Use Case: You have 128 independent layout-search seeds that would take hours locally. The Skill bundles them into one array job of 128 single-vCPU children, runs them on Spot capacity, and downloads all 128 result directories. ## Quick Start Use the icfpc-aws-compute skill to run this layout search as an AWS Batch Spot array job and download all results.
