AI code review skill (SKILL.md) for Claude Code, Codex and Cursor

Run it in a fresh session so the reviewer isn't the agent that wrote the code.

Preview: SKILL.md

---
name: ai-code-review
description: Reviews the current branch diff for AI-written code risks, in priority order (access control, secrets, dependencies, input handling, errors, scope) and reports findings by severity without changing code. Use when the user asks for a code review, a security review, a PR review, or whether a branch is safe to merge.
metadata:
  author: codema
  version: "1.0"
---

# AI code review

Review the changes on the current branch against the main branch. Report first. Do not edit, commit, or push anything unless the user explicitly asks after reading the report.

## 0. Scope the diff

1. Find the base branch (usually `main`). Run `git diff --stat <base>...HEAD` and `git diff <base>...HEAD`.
2. List every changed file with a one-line summary of what changed.
3. Flag any changed file that does not relate to the stated task as UNEXPECTED.
4. If the diff is over roughly 800 changed lines, say so and suggest splitting it before a full review.

## 1. Access control (highest priority)

What's inside

• One validated SKILL.md named ai-code-review, so it doesn't clash with Claude Code's built-in /code-review
• Findings by severity, each with file, line and a suggested fix

How to use it

1. Unzip into your skills folder.
2. Open a new session.
3. Ask: "review this branch before I merge".
4. Check each finding yourself before acting.

Who it's for

Builders who merge AI-written code and want a second pass.

Related

• How to review AI-generated code before production
• Git and GitHub basics for vibe coders
• Starter skills pack