refactor surgeon persona

The problem

Large refactors often break production because structural changes get mixed with logic updates. Manual cleanups frequently lack a reversal path or a safety net, leading to unverified code changes and "heroic rewrites" that fail.

What it does

• Establishes behavioral safety nets using characterization tests before any code is touched.
• Decomposes complex structural changes into small, individually shippable steps that keep the build green.
• Enforces a strict separation between refactoring commits and behavior-changing commits.
• Justifies every change by quantifying the reduction in friction, bug rate, or onboarding time.
• Identifies specific refactoring patterns like Strangler Fig or Extract Method with surgical precision.

Why this beats prompting it yourself

General LLMs often try to rewrite entire files, introducing subtle logic bugs. This skill enforces a disciplined, incremental methodology that prioritizes evidence over aesthetic taste, ensuring every step is provably behavior-preserving.

Use cases

• Untangling legacy payment systems or critical business logic where downtime is not an option.
• Breaking down a monolithic function into testable modules without changing its output.
• Preparing a codebase for a new feature by reducing existing change friction.
• Implementing the Strangler Fig pattern to migrate logic away from a mess of technical debt.

Known limitations

Explicitly excludes feature development, performance tuning, and product-level decision making. Requires user confirmation for all file modifications and repository changes.