This is a parallel execution orchestrator that spins up 12 concurrent Sparky subagents to work through plan files simultaneously. It uses a rolling pool scheduler that keeps slots filled as tasks complete, which means you're not waiting on sequential execution for independent work. The skill enforces strict canonical file naming across all subagents to prevent the chaos of 12 agents inventing their own filenames, then does a final integration pass to reconcile conflicts and run tests. It's overkill for small plans but genuinely useful when you have a 20-task plan file and want to compress hours into minutes. The tradeoff is complexity: you need well-structured plans and you'll spend time in the integration phase cleaning up parallel execution artifacts.
npx skills add https://github.com/am-will/codex-skills --skill super-swarm-spark