How does this skill prevent regressions?

It utilizes baseline characterization tests to capture the current system's exact behavior before any changes are made, ensuring those tests remain passing throughout the refactor.

Can I use this for any programming language?

Yes, while the helper scripts are in Python, the methodology is language-agnostic and designed to work with any codebase that supports automated testing.

Is it safe to use on my main branch?

No, the workflow strictly requires working on a dedicated refactor branch to maintain the safety and integrity of your production-ready code.

What is a Refactor IR?

The Refactor Intermediate Representation is a formal catalog documenting all architectural unknowns (holes), their dependencies, required constraints, and resolution status.

What is Design by Typed Holes?

It is a methodology where you treat missing or changing parts of an architecture as 'holes' with specific types and constraints, resolving them one by one through iterative testing and validation.

Typed Holes Refactor

Name: Typed Holes Refactor
Author: rand

byrand

•

Developer Tools

Systematically refactor complex codebases by treating architectural unknowns as typed holes and resolving them through iterative, test-driven validation.

The Typed Holes Refactor skill implements a formal meta-framework for large-scale code transformations, allowing developers to manage architectural uncertainty with precision. By categorizing unknowns as 'holes' in a Refactor Intermediate Representation (IR), it guides Claude through a structured workflow of characterization testing, constraint propagation, and incremental resolution. This approach ensures that even massive refactors remain safe and verifiable, preserving existing behaviors while systematically migrating toward a target architecture with full documentation of every change.

Key Features

01Baseline characterization testing to prevent behavioral regressions during migration

02Comprehensive delta reporting on metrics like LOC, complexity, and performance

03Automated hole discovery and dependency mapping via Refactor IR

04TDD-driven resolution loop ensuring all changes pass strict validation gates

0557 GitHub stars

06Automated constraint propagation to maintain architectural integrity across modules

Use Cases

01Migrating between architectural patterns or framework versions with verified equivalence

02Consolidating technical debt in large, legacy monorepos through systematic decoupling

03Decoupling tightly integrated modules into clean, interface-driven layers

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rand/cc-polymath typed-holes-refactor

For use in Claude.ai and ChatGPT

Download Skill