refactor

To plan and execute a structured code refactor, you provide a module path to trigger a multi-agent workflow that analyzes code, generates implementation plans, and applies gradual architecture improvements with traceable roll-back plans.

Yes, you can improve type safety and architecture in parallel by distributing refactoring tasks across agents, enabling parallel execution and verification while preserving safety across independent modules.

The refactor workflow handles documentation by automatically generating ANALYSIS.md, PLAN.md, and METRICS.md files under a dedicated directory, ensuring changes are documented and providing traceable roll-back plans.

To delegate refactoring tasks to an advisor, you need an environment supporting subagent delegation where you can specify the advisor type, model, prompt, and description to initiate the structured refactoring workflow.

The best way to ensure safe gradual architecture improvements is using a guided multi-agent workflow that evaluates alternatives, creates structured implementation plans, and verifies changes in parallel before execution.

You should avoid an automated multi-agent workflow for code refactoring when the scope is unclear, as the system requires a clear module path and defined scope to generate accurate analysis and implementation plans.