convert-to-fanout

A fan-out architecture optimizes AI agent performance by splitting a monolithic agent into an orchestrator and workers. The orchestrator delegates mechanical tasks to cheaper models while retaining the high-cost model for critical judgment, reducing overall operational costs.

Converting a raw API agent to a fan-out architecture involves assessing current tasks, selecting appropriate fan-out or long-horizon patterns, and applying detailed steps to distribute work to cheaper models. The skill guides this transformation for both raw API and Claude Code agents.

Select a fan-out pattern for tasks that can be distributed concurrently to workers, and a long-horizon pattern for sequential tasks. The skill diagnoses your agent's specific tasks to guide the appropriate architecture pattern selection.

Yes, the fan-out conversion supports Claude Code agents by providing specific application steps for this agent type. It transforms the setup into an orchestrator and workers while ensuring the new configuration maintains quality and cost efficiency.

You verify quality after optimizing your AI agent architecture by using the skill's built-in verification and measurement features. These ensure the new orchestrator and worker setup maintains quality standards and achieves cost efficiency.