How does the branch-review skill handle large code changes?

It avoids token limits by only passing the file names to agents, who then use 'read' tools to analyze the final state of specific files instead of processing massive diff text.

In what format is the review report generated?

The skill generates a comprehensive, integrated report in Korean Markdown, typically saved in the docs directory under the relevant domain.

Can I run the review manually at any time?

Yes, you can trigger the review whenever you need by typing the /branch-review command in your Claude Code interface.

Does this skill require specific project documentation?

While it can run on any branch, it is optimized for projects with a CLAUDE.md file for architecture rules and standardized feature plan documents.

Which specialized agents are used in the review?

The skill uses four agents: code-reviewer (quality/naming), test-engineer (coverage), backend-architect (API/performance), and architect-reviewer (consistency/SOLID).

Branch Architecture Reviewer

Name: Branch Architecture Reviewer
Author: invigoworks

byinvigoworks

0•

コラボレーションツール

Conducts multi-perspective code reviews using four specialized AI agents to ensure architectural alignment and implementation quality.

The branch-review skill automates the pre-PR quality assurance process by deploying four parallel agents—code-reviewer, test-engineer, backend-architect, and architect-reviewer. It systematically evaluates changes against the project's CLAUDE.md architecture rules and original feature plans. By reading final file states instead of just diffs, it provides high-precision feedback on Hexagonal Architecture compliance, CQS patterns, and test coverage, culminating in a detailed Korean Markdown report that highlights critical issues and suggested improvements.

主な機能

010 GitHub stars

02Consolidated Korean Markdown report with severity-ranked action items

03Automated CLAUDE.md architecture rule validation

04Plan-to-implementation consistency checking

05Full-file context analysis to prevent token-limit diff errors

06Parallel 4-agent specialized review (Code, Test, Backend, Architecture)

ユースケース

01Ensuring strict adherence to Hexagonal Architecture and Zero-DTO policies

02Performing a final quality gate check before submitting a Pull Request

03Verifying that implementation matches the original feature design documents

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add invigoworks/invigo-marketplace branch-review

For use in Claude.ai and ChatGPT

Download Skill