convert-to-fanout

What problem does it solve?

This Skill addresses the high cost of running monolithic AI agents by converting them into a fan-out architecture, which utilizes cheaper models for mechanical tasks while retaining the high-cost model for critical judgment.

Core Features & Use Cases

• Fan-Out Conversion: Transforms a monolithic agent into an orchestrator and workers, distributing tasks to cheaper models.
• Task Diagnosis: Assesses the current agent setup, identifying mechanical and judgment tasks.
• Pattern Selection: Guides the selection of the appropriate fan-out pattern based on task shape (fan-out or long-horizon).
• Conversion Application: Provides detailed steps for applying the conversion to different agent types (raw API or Claude Code).
• Verification & Measurement: Ensures the new setup maintains quality and cost efficiency.
• Use Case: Convert a web research agent that reads and processes multiple sources to utilize an orchestrator and Sonnet 5 workers for reading, while the main model (e.g., Fable 5) handles synthesis and judgment.

Quick Start

Run the convert-to-fanout skill on your current agent to optimize its performance.