code-review

Analyze git diffs to generate severity-classified code review reports.

Updated May 4, 2026
One-click install
npx skills add https://github.com/mhert/claude-code-helpers --skill code-review-mhert
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: code-review
Source: https://github.com/mhert/claude-code-helpers/tree/main/skills/code-review
Command: npx skills add https://github.com/mhert/claude-code-helpers --skill code-review-mhert

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires git, and includes references (resource) components.

What problem does it solve?

It reduces the time and missed issues involved in reviewing a feature branch by generating a structured, severity-ranked review across correctness, security, architecture, testing, performance, and style.

Core Features & Use Cases

  • Six-dimension multi-agent review: Runs focused checklists in parallel and merges results into one deduplicated report.
  • Commit-attributed findings: Connects reported issues back to the commit that introduced the lines using git blame so teams can act confidently.
  • Scope flexibility: Reviews a PR, an explicit ref range, a single commit, or pending diffs (staged/working), while avoiding empty diffs and default-branch self-reviews.
  • Change-focused by design: Only evaluates what the branch actually changed and drops pre-existing issues outside the diff.

Quick Start

Tell the AI: "Review my branch and return a severity-classified report with a clear verdict."

Frequently Asked Questions about code-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get an automated code review for a pull request that covers security and performance?

A multi-perspective code review evaluates your pull request diff across six dimensions—correctness, security, architecture, testing, performance, and style—then merges and deduplicates results into one severity-classified report.

Can I review a single git commit or staged changes instead of a full pull request?

Yes, code review supports scope flexibility by analyzing explicit ref ranges, single commits, and pending staged or working diffs using git, while automatically avoiding empty diffs and default-branch self-reviews.

How does commit-attributed code review connect findings back to the commits that introduced them?

Commit-attributed findings use git blame to connect each reported issue back to the specific commit that introduced the changed lines, allowing teams to act confidently on severity-classified review results.

Do I need git installed to run a multi-perspective code review on my branch?

Yes, git is a required dependency because the review process uses git commands to compute cumulative diffs, analyze ref ranges, and attribute findings to originating commits via git blame.

What is the best way to review only the changes my feature branch introduced without flagging pre-existing issues?

A change-focused code review evaluates only what your branch actually changed by analyzing the cumulative diff versus a chosen base scope, dropping pre-existing issues that fall outside the diff.

How does a severity-classified code review determine its final verdict?

The review applies a mechanical verdict rule based on the presence of Critical or Warning severity findings in the deduplicated report, producing a clear pass-or-fail judgment for the branch.