code-review-expert skill (sanyuan0704) for Claude Code: review your git changes

code-review-expert makes your agent review your uncommitted git changes like a senior engineer and rank each finding P0 to P3.

Installs: 13.6K; on skills.sh
Stars: 4K; on GitHub
Files: 6; besides SKILL.md
Updated: May 2026; last repo commit

code-review-expert is a skill from sanyuan0704's sanyuan-skills repo. Run /code-review-expert and the agent reads your git diff, checks it for SOLID violations, dead code, security risks and code quality problems, then lists findings by severity. It reviews only by default: it does not change code until you pick an option.

It is built around seven steps, from preflight to a next-steps prompt, and four checklists in references/. It handles an empty diff by asking whether to review staged changes or a commit range, and reviews large diffs (over 500 lines) file by file in batches. It bundles no scripts; SKILL.md has it use git status, git diff, rg or grep on your repo.

Good for

Developers who want a self-review before opening a pull request
Reviewers who want a consistent severity-ranked checklist
Teams checking security and race conditions in a diff

Install code-review-expert

Claude Code

Copy this prompt.

Install the code-review-expert skill in Claude Code. Run this command:

npx skills add https://github.com/sanyuan0704/sanyuan-skills --skill code-review-expert

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

Expert code review of current git changes with a senior engineer lens. Detects SOLID violations, security risks, and proposes actionable improvements.

Sections in SKILL.md

Overview
Severity Levels
Workflow
1) Preflight context
2) SOLID + architecture smells
3) Removal candidates + iteration plan
4) Security and reliability scan
5) Code quality scan
6) Output format
Code Review Summary
Findings
P0 - Critical
P1 - High
P2 - Medium
P3 - Low
Removal/Iteration Plan

Key facts

Author: sanyuan0704 (GitHub account that owns the repo)
Source repo: sanyuan0704/sanyuan-skills
Installs on skills.sh: 13,648 as of 2026-11-02
GitHub stars: 3,969 (whole repo) as of 2026-11-02
License: MIT
Last repo commit: 2026-05-11
Files: 7 files including SKILL.md (README.md, agents/agent.yaml and 4 checklists in references/)
skills.sh audits: Gen Agent Trust Hub pass, Socket pass, Snyk pass

What it includes

Scopes the change with git status and git diff; uses rg or grep to find usages.
Runs code · 1. Preflight context
Checks SRP, OCP, LSP, ISP and DIP using references/solid-checklist.md.
Read · 2. SOLID and architecture smells
Finds unused or flagged-off code; safe delete now vs defer, with a follow-up plan.
Read · 3. Removal candidates
XSS, injection, SSRF, auth gaps, secret leakage, race conditions and more.
Read · 4. Security and reliability scan
Error handling, performance (N+1, hot paths) and boundary conditions.
Read · 5. Code quality scan
Summary, findings by P0 to P3, optional removal plan and inline code-comment lines.
6. Output format
Asks: fix all, fix P0/P1 only, fix specific items, or no changes. Waits for your answer.
7. Next steps confirmation

From the official docs

Permissions

Low risk

The skill makes your agent run git and search commands on your repo and read your diff, which goes to your model provider. It does not edit files until you confirm.

What it can do

Read your data
Create and change things
Delete things
Read-only modeHow to turn it on

Watch out for

Fix options edit your files
If you choose "Fix all" or similar, the agent implements changes in your working tree. Review them with git diff before committing. github.com
Runs git and search commands
Preflight uses git status -sb, git diff --stat, git diff, and rg or grep. Your diff becomes part of the conversation with your model. github.com

Technical details

Turning on read-only	SKILL.md defaults to review-only output and says not to implement changes until you explicitly confirm.
Scopes	Your git working tree

Questions

How do I install the code-review-expert skill?

Use the install box above. Then run /code-review-expert in a repo with uncommitted changes. The skill's README also gives an npx skills command with a path option.

Will it change my code?

Not until you confirm. SKILL.md says to default to review-only output, and the last step asks how to proceed with options including "No changes". It says not to implement anything until you explicitly confirm.

What does it review: staged, unstaged or a commit?

It starts from your current git changes using git status, git diff --stat and git diff. If the diff is empty, it asks whether to review staged changes or a specific commit range.

What do P0, P1, P2 and P3 mean?

P0 is critical and must block merge. P1 is high and should be fixed before merge. P2 is medium: fix now or open a follow-up. P3 is low and optional.

Does it need an API key?

The skill does not mention one. It bundles no scripts. It reads your repo with git and rg or grep, so those commands need to be available.

Does it work in Claude Code, Codex or the Claude app?

The skill's agents/agent.yaml says it is agent-agnostic with no provider-specific configuration. The install box lists Claude Code, the Claude app upload and Codex routes.

Is code-review-expert safe to install?

It bundles no scripts. It makes the agent run git status, git diff and rg or grep on your repo, and the diff goes to your model provider. SKILL.md says not to implement changes until you explicitly confirm, but if you choose a fix option the agent edits your working tree, so review it with git diff.

Information

updated Nov 2, 2026

Repository: sanyuan0704/sanyuan-skills
Folder: skills/code-review-expert

sanyuan0704/sanyuan-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 5

code-review-expert SKILL.md · github.com
sanyuan0704/sanyuan-skills README · github.com
skill README · github.com
code-review-expert on skills.sh · skills.sh
agent.yaml · github.com