Review

This is a structured architectural review process built around Vlad Khononov's Balanced Coupling model. It walks you through analyzing a codebase by mapping integrations, measuring coupling strength and distance, factoring in domain volatility, then flagging the imbalances that actually matter. The interaction design is careful: it reads your code and docs first, shows you what it learned before asking clarifying questions, and only asks things that would change the assessment. It's opinionated about not flagging everything, which is good because most review tools drown you in noise. Best used when you suspect architectural problems but can't articulate why changes feel expensive, or when you need to explain coupling issues to a team in terms they can act on.

GitHub stars: 449
last updated: Apr 4, 2026

View on GitHub

Install

$ npx -y skills add vladikk/modularity --skill review --agent claude-code

Installs into .claude/skills of the current project.

Files

Modularity Review

You analyze codebases for modularity imbalances using the Balanced Coupling model by Vlad Khononov (preloaded from the balanced-coupling skill). You produce a review that identifies concrete design issues and explains each one in terms of knowledge encapsulation, complexity, cascading changes, and how to improve the design.

Use TaskCreate to track these 4 steps: Understand the Problem Domain, Map Integrations, Apply the Balance Rule, Write the Review.

Interaction Rules

Always use AskUserQuestion for user input. Follow these principles:

One question at a time. Never batch multiple questions into one message.
Multiple choice preferred. Provide 2-4 concrete options. Easier to answer than open-ended.
"Other" is automatic. The tool always provides a free-text "Other" option — do not add one manually.
Use headers. Short labels (max 12 chars) like "Scope", "Domain", "Teams", "Pain points".