Guide · Agent skill
How to use the Code Reviewer agent skill for AI code reviews
Updated · written with AI from the project's own documentation and checked against our catalogue data
The code-reviewer skill is a set of instructions and checklists that an AI agent loads when you ask it to review code. It covers pull requests, diffs, commits, and pasted functions or files in 14 languages, from TypeScript and Python to Rust and Dart/Flutter. This guide explains what the skill changes, how to install it, how to trigger it with plain requests, and where its limits are, so you can use it for a consistent first-pass review.
Code Revieweralirezarezvani/claude-skills · MIT Download the skill Quick start
- 1Download the ZIP. Download the code-reviewer ZIP from this page.
- 2Unzip it into the right folder. Unzip into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder to limit it to that project.
- 3Or upload it in the Claude apps. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
- 4Check the prerequisites. Make sure Python is installed to run the bundled scripts, and Git is available if you will analyse branches or commits.
- 5Ask for a review in plain words. Say something like 'review this PR' or 'is this safe to merge' and point the agent at the diff, branch or pasted code.
- 6Read the verdict and findings. Work through blocking findings first, then review non-blocking ones before deciding on the change.
What the Code Reviewer skill makes the agent do
Without a skill, an AI agent reviews code from general knowledge, and its output can vary from one request to the next. This skill gives the agent a fixed structure: general rules that apply to every language, plus one rule file for the language in question.
The checks cover security problems such as SQL injection and missing authentication, resource and error-handling mistakes, async handling, and performance. The result is a review with blocking and non-blocking findings and a merge verdict.
- Universal rules: security, async, resources, exceptions and performance, applied to all languages
- Per-language rule files with idioms and language-specific risk signals
- Bundled Python scripts that estimate a pull request's risk and complexity and check code quality
- A verdict of approve, approve with suggestions, request changes, or block
Install the Code Reviewer skill
Download the ZIP from this page. You can install the skill in one of three ways. Pick the one that matches how you work with the agent.
For a personal setup, unzip the download into ~/.claude/skills/. To limit the skill to one project, unzip it into that project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.
- Personal: unzip into ~/.claude/skills/
- Single project: unzip into the project's .claude/skills/ folder
- Claude apps: upload the ZIP under Settings → Capabilities → Skills
Trigger the skill with natural requests
You do not need a special command. The skill is meant to load when you ask for a review in ordinary words. The skill description names phrases such as 'code review', 'review this PR', 'review this diff' and 'is this safe to merge', and also covers asking what is wrong with a snippet, even a single function.
Be specific about what you want reviewed. Pointing the agent at a diff, a commit, a branch, or a pasted function gives it a clear scope.
- Review this PR and tell me what blocks merging.
- Review this diff for security problems such as SQL injection or missing auth.
- Is this safe to merge? Here is the diff.
- What is wrong with this function? [paste the function]
- Look over this handler for error-handling mistakes.
What the bundled scripts do and what they need
The skill's documentation describes three Python scripts: a PR analyzer that compares git branches, a code quality checker, and a review report generator. The PR analyzer reports a complexity score from 1 to 10, a risk category (critical, high, medium or low), file prioritisation, and commit message validation.
The skill needs Python to run these scripts and Git if you analyse branches or commits in a repository. The skill does not use the network. Check that the scripts folder is present in your download, since the documentation refers to it.
- Requires: Python (for the scripts) and Git (for branch or commit analysis)
- Runs scripts: yes
- Uses network: no
python scripts/pr_analyzer.py . --base main --head feature-branchRead the verdict and the findings
The review report maps a quality score to a verdict. A score of 90 or more with no high issues gives Approve. A score of 75 or more with two or fewer high issues gives Approve with suggestions. Scores from 50 to 74 give Request changes. A score below 50, or any critical issue, gives Block.
Treat the findings as a list to check, not a final decision. Blocking findings are the ones to fix before merging; non-blocking findings are worth reading but may not need action.
- 90+ and no high issues: Approve
- 75+ with 2 or fewer high issues: Approve with suggestions
- 50 to 74: Request changes
- Below 50, or any critical issue: Block
Safety notes and limits
The checks are heuristic pattern rules. They can miss real issues and can flag code that is actually fine. The skill is not a replacement for a human security audit or a compliance review.
The scoring thresholds are fixed and have not been validated against your codebase. Do not auto-merge or gate production deployments on the verdict alone. Use the skill as a consistent first pass before a human reviews the change.
- Not a replacement for a human security audit or compliance review
- Pattern-based checks can produce false positives and false negatives
- Do not auto-merge or gate deployments on the verdict alone
What people use it for
- First-pass review of a pull request before a human reviewer looks at it
- Checking a single function or handler for obvious security or error-handling problems
- Applying the same review standards across a codebase that mixes several languages
- Getting a complexity and risk estimate for a branch with the PR analyzer script
Frequently asked questions
Is the code-reviewer skill free?
The skill is published under the MIT licence, as stated on this page. Check the licence file in the repository for the full terms before you use or redistribute it.
What do I need to run the code-reviewer skill?
You need Python to run the bundled scripts in the scripts folder, and Git if you analyse branches or commits in a repository. The skill does not use the network.
Where do I install the Code Reviewer skill?
Unzip the ZIP into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project. In the Claude apps, upload it under Settings → Capabilities → Skills.
Why isn't the skill triggering when I ask for a review?
Use clear review wording such as 'code review', 'review this PR' or 'review this diff', and paste or point to the code. Confirm the folder is in the right location, since the skill is loaded from there.
Can I rely on the skill to approve code for production?
No. The checks are heuristic pattern rules and the scoring thresholds are fixed and not validated against your codebase. Use the verdict as a first pass and keep a human reviewer in the loop.
How does it compare with a human code review or a security audit?
The skill gives a consistent first-pass checklist, but it is not a human security audit or compliance review. It can miss issues and flag false positives, so it complements human review rather than replacing it.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you want a consistent, structured first pass on pull requests, diffs or single functions, especially across several languages. Pick something else, or a human reviewer, if you need a security audit, a compliance review, or an automated merge gate, because the checks are heuristic and the thresholds are fixed.
Download the skill