Skip to main content
skill CX Codex

review · buiducnhat — Claude Code skill: what it does, how to install

Reviews uncommitted git diffs using full file context and produces structured findings with severity levels and an approval verdict.

MIT 279 installs ★ 48 updated 2026-07-06

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add buiducnhat/cobrew --skill review -a claude-code
Open on GitHub

What review does

The skill inspects current workspace changes using git diff and git diff --cached. Instead of looking only at diff hunks, it reads modified files in full to evaluate surrounding architecture. It also checks docs/SUMMARY.md and project coding standards, running relevant linters or test suites for affected areas when available.

Every issue receives a severity grade from S0 (critical breakages or data loss) down to S3 (minor polish). The final report provides a change summary, prioritized findings tied to file lines, positive observations, and a must-fix checklist. If any S0 or S1 issue is present, the final verdict strictly defaults to Request Changes.

What it helps with

  • catch bugs and security flaws in uncommitted git changes
  • verify full file context instead of reviewing isolated diff hunks
  • generate a prioritized must-fix checklist before committing code
  • enforce project coding standards and test coverage on modified files

Example prompts for the agent

  • Review current uncommitted git changes and give an approval verdict
  • Check staged changes against project code standards and list blocking issues
  • Inspect modified files for reliability gaps and generate a review report

How to install review

1Skills CLI (Vercel)

npx skills add buiducnhat/cobrew --skill review -a claude-code

2Manually: put the files in place

~/.claude/skills/review/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/review/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about review

How to install review in Claude Code?

Run npx skills add buiducnhat/cobrew --skill review -a claude-code in your terminal.

What does the review skill do?

It reviews uncommitted git changes with full file context, classifies issues from S0 to S3, and delivers a structured report with actionable suggestions.

When does the skill issue a Request Changes verdict?

The skill marks the review as Request Changes whenever it identifies at least one blocking issue labeled S0 (critical defect) or S1 (high-priority flaw).

Source text

SKILL.md · MIT
---
name: review
description: Review current uncommitted git changes with full file context and produce a structured report with severity levels, actionable fixes, and an approval verdict.
license: MIT
---

# Review

## Overview

Use this skill to review uncommitted changes in the current git workspace.  
Focus on correctness, safety, maintainability, and alignment with project standards.

The review must produce:

1. A concise summary of what changed
2. Prioritized findings with severity labels
3. Actionable fix suggestions
4. A final verdict (`Approve` or `Request Changes`)

## Workflow

### Step 1: Gather Context

1. Inspect current changes:
   - `git diff`
   - `git diff --cached` (if staged changes exist)
2. Read the full modified files (not only the diff hunks) to understand surrounding logic and architecture.
3. Load only the project context relevant to the changed files before judging style/patterns:
   - If `docs/SUMMARY.md` exists, read it first.
   - Load only task-relevant detail docs.
   - Prioritize `Code Standard` docs for implementation conventions.
   - If docs conflict with code or user intent, use the available input/question tool before broad changes.
4. Run relevant quality checks for touched areas (lint/type/tests when practical).

### Step 2: Analyze Changes

Evaluate each change across these dimensions:

1. **Correctness** - Does behavior match intended requirements?
2. **Safety & Security** - Are there vulnerabilities, data leaks, auth gaps, or unsafe assumptions?
3. **Reliability** - Are edge cases, null states, retries, and error paths handled?
4. **Style & Consistency** - Does code follow project conventions and established patterns?
5. **Performance** - Any unnecessary expensive operations or regressions?
6. **Testing** - Are critical paths covered with appropriate tests?
7. **Maintainability** - Is the code clear, modular, and easy to evolve?

### Step 3: Classify Findings by Severity

Assign one severity per finding using this rubric:

- **S0 - Critical**
  - Production-breaking issue, severe security risk, data corruption/loss, or irreversible side effects
  - Must be fixed before merge

- **S1 - High**
  - Likely bug, correctness flaw, significant reliability issue, or major missing validation
  - Should be fixed before merge

- **S2 - Medium**
  - Non-blocking but meaningful issue affecting maintainability, performance, or clarity
  - Fix recommended soon

- **S3 - Low**
  - Minor improvement, polish, or style-level suggestion
  - Optional unless team standards require it

### Step 4: Produce Structured Review Report

Use the report format below in this exact section order.

## Output Format

### 1) Summary

- What changed (2-6 bullets)
- Risk profile (low/medium/high)
- Areas reviewed (files/modules)

### 2) Findings

For each finding, use this template:

- **ID**: `R-001` (incrementing)
- **Severity**: `S0 | S1 | S2 | S3`
- **Category**: `Correctness | Security | Reliability | Style | Performance | Testing | Maintainability`
- **Location**: `path/to/file.ext#Lx-Ly` (or function/class name)
- **Issue**: Clear statement of the problem
- **Why it matters**: User/system impact
- **Suggestion**: Concrete fix guidance
- **Confidence**: `High | Medium | Low`

If no findings exist, explicitly write: **No actionable findings**.

### 3) Positive Notes

List good practices observed, such as:

- Strong test coverage additions
- Clean separation of concerns
- Thoughtful error handling
- Consistent style and naming

### 4) Must-Fix Checklist

Include only `S0` and `S1` findings:

- [ ] `R-00X` short fix description
- [ ] `R-00Y` short fix description

If none, state: **No must-fix items**.

### 5) Verdict

Use one of:

- **Approve** - No blocking issues (`S0/S1`) remain.
- **Request Changes** - One or more blocking issues (`S0/S1`) found.

Optionally include:

- **Re-review focus**: exact files/areas to re-check after fixes.

## Rules

- Be specific and cite exact locations whenever possible.
- Do not judge code in isolation; always consider surrounding context.
- Prefer actionable suggestions over vague criticism.
- Distinguish clearly between blocking and non-blocking feedback.
- Avoid speculative claims; if uncertain, lower confidence and explain why.
- Align feedback with project documentation and coding standards.

## Review Quality Checklist

Before finalizing, confirm:

- All modified files were reviewed with context
- Each finding has severity, impact, and fix suggestion
- Blocking issues are separated into a must-fix checklist
- Final verdict matches the finding severities
- Feedback is concise, precise, and implementable

© buiducnhat · MIT · buiducnhat/cobrew

Similar skills

All skills for Code review and refactoring →

review by other authors comes first.

review garrytan review is a Claude Code skill that checks local changes, a branch, or a PR before merge. It examines the diff and writes a review report. CX Codex +1 514installs review official anthropics review is a Claude Code skill that checks incoming agreements against a company playbook and routes them to the right review skills. It runs when a contract is submitted for review. CX Codex +1 778installs review JuliusBrussee Runs an adversarial senior review of the spec before any code is written. Builds a skeptic and tries to refute the plan instead of rubber-stamping it. CX Codex +1 811installs review parcadei review runs parallel specialized subagents to evaluate code quality, architecture, and impact, then synthesizes a final merge verdict. CX Codex +1 523installs review jsmastery-pro review audits completed features against your original plan, system architecture, and production readiness. Flags discrepancies in a structured report without making code edits. CX Codex +1 1,000installs code-review justxor The code-review skill checks changes in the current branch against main and produces a report with notes on correctness, security, tests, and readability. CX Codex +1 ★ 353repo stars independent-review Ronny (MiniMax Code team) independent-review is a Claude Code skill that reviews changes from a clean context before merge. It compares the diff with the requirement and test output. CX Codex +1 ★ 741repo stars open-code-review-delegate alibaba open-code-review-delegate is a Claude Code skill that reviews commits, PRs, and workspace diffs using the host model. The ocr CLI selects files and rules while the agent model inspects code. CX Codex +1 5,000installs hunk-review modem-dev hunk-review — a skill for Claude Code connects to live Hunk diff review sessions via CLI. Triggers during interactive code reviews when Hunk runs in the terminal or browser. CX Codex +1 2,700installs pr-review-canvas cursor pr-review-canvas generates a standalone interactive HTML walkthrough for any GitHub pull request. It helps reviewers focus on core architecture by folding boilerplate and rendering smart diffs. CX Codex +1 2,400installs difit-review yoshiko-pg difit-review launches the local difit viewer for git diffs and attaches review notes to exact code lines. Runs only upon explicit request. CX Codex +1 2,000installs

From the same set7

Project sponsors

Companies that keep this analytics open