code-review: Claude Code Skill & Agent Tool
An open-source skill by mattpocock (opens in a new tab), published in the mattpocock/skills (opens in a new tab) repository on GitHub. Reviews every change since a commit, branch or tag along two axes — your repository's documented coding standards and the original issue or spec — running both reviews in parallel and reporting them side by side.
Install code-review
Requires Node.jsnpx skills add mattpocock/skills --skill code-reviewRun this command in your project terminal to install this skill for Claude Code.
What this skill does
code-review is Matt Pocock's review workflow for AI coding agents. You give it a fixed point — a commit, a branch, a tag or the merge-base with main — and it reviews every change made since then along two separate axes. The Standards review checks the diff against the coding standards your repository documents, such as CODING_STANDARDS.md or CONTRIBUTING.md, plus a baseline of classic code smells that applies even when nothing is written down. The Spec review compares the same diff with what the originating issue, ticket or specification actually asked for.
Both reviews run in parallel sub-agents and are reported side by side under their own headings. The skill deliberately never merges or re-ranks the findings: a change can follow every convention and still miss the point of the ticket, or solve the ticket while breaking house rules, and mixing the two lists hides exactly that. It ends with a one-line summary of the findings per axis and the worst issue within each.
Use it with Claude Code before opening a pull request, or as a second pair of eyes on a teammate's branch.
Key Capabilities
- Pins the review to a fixed point: a commit, branch, tag or merge-base
- Standards axis: your documented coding standards plus a baseline of classic code smells
- Spec axis: checks the diff against the originating issue or specification
- Runs both reviews in parallel sub-agents and reports them side by side, never merged or re-ranked
- Closes with a one-line summary: findings per axis and the worst issue in each
How to use with Claude Code & AI Agents
Install the skill
Run the command above at the root of your project. The CLI asks which agents to set up — Claude Code, Cursor and other supported agents — and adds code-review to the project, so teammates who pull the repository get it too. Add-gto install it for your user account across every project.Open a new agent session
Start a fresh Claude Code session in that project so the agent loads the new skill. This skill needs no extra tool or API key.Ask in plain words
Describe the task, for example:“Review everything on this branch since main against our standards and issue #42.”The agent matches your request to the skill's description and follows its SKILL.md instructions step by step. Read them on GitHub before using the skill on a system that matters.