Code review skill: Matt Pocock's two-axis review with parallel sub-agents

Source: mattpocock/skills · Maintainer: community: Matt Pocock · License: MIT · Checked: 23 October 2026

What is the code review skill?

The code review skill has your coding agent review a branch or pull request on two separate questions: does the code follow this repo's own documented standards, and does it do what the original issue or spec asked for? It runs each question in its own sub-agent, at the same time, and reports the answers side by side. It is for developers who use agents to write code and want a review that knows their conventions and their requirements, not a generic one.

It lives in mattpocock/skills, Matt Pocock's collection of "agent skills that I use every day to do real engineering". The skill's description reads: "Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes: Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/spec asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X"."

As of 23 October 2026, its skills.sh page listed about 787,000 installs. The repo had about 301,000 GitHub stars.

What does the code review skill do?

You type /code-review, or the agent picks it up when you ask to review a branch, a PR or anything "since X". The SKILL.md then runs five steps:

1. Pin the fixed point. You name a commit, branch, tag, main or HEAD~5; if you don't, it asks. It captures git diff <fixed-point>...HEAD and the commit list, and stops early if the ref doesn't resolve or the diff is empty.
2. Find the spec. In order: issue references in commit messages (#123, Closes #45), a path you pass, a matching file under docs/, specs/ or .scratch/, then asking you. With no spec, the Spec axis is skipped and the report says so.
3. Find the standards. Every file in the repo that documents how code should be written, always including CODING_STANDARDS.md or CONTRIBUTING.md if present. On top sits a "smell baseline" of twelve smells from chapter 3 of Martin Fowler's Refactoring: Mysterious Name, Duplicated Code, Feature Envy, Data Clumps, Primitive Obsession, Repeated Switches, Shotgun Surgery, Divergent Change, Speculative Generality, Message Chains, Middle Man and Refused Bequest. Your repo's documented rules always override the baseline, and smells are always labelled as judgement calls.
4. Spawn two sub-agents in parallel. The Standards sub-agent cites the rule (file and rule) or names the smell and quotes the hunk. The Spec sub-agent reports missing requirements, scope creep and wrong implementations, quoting the spec line for each. Both briefs ask for under 400 words and skip anything tooling already enforces.
5. Aggregate without merging. The two reports appear under ## Standards and ## Spec, with a one-line summary of the worst issue in each axis. The skill deliberately does not rank findings across axes, because code can pass one and fail the other.

Running the axes in separate sub-agents keeps each one's context clean, so a spec problem can't be drowned out by style notes.

What's in the folder:

• SKILL.md: the five-step process, the smell baseline and the sub-agent briefs.
• agents/openai.yaml: Codex metadata, display name "Code Review".

The repo's longer code-review docs explain the design and answer common questions.

How do you install the code review skill?

The README offers a plugin or editable files and says to pick one per agent, since installing both gives you every skill twice.

Claude Code plugin (whole set, updates itself). From the README, via Anthropic's official plugin marketplace:

claude plugin install mattpocock-skills@claude-plugins-official

Plugin skills are namespaced, so it appears as /mattpocock-skills:code-review.

Codex plugin. From the README:

codex plugin marketplace add mattpocock/skills
codex plugin add mattpocock-skills@mattpocock

Just this skill, as editable files. With the skills CLI, take code-review plus setup-matt-pocock-skills, which the skill tells you to run if no issue tracker is configured:

npx skills add mattpocock/skills --skill code-review --skill setup-matt-pocock-skills

Add -g for every project. For Claude Code the folder lands in .claude/skills/code-review/, or ~/.claude/skills/code-review/ with -g, the standard Claude Code skill locations. Codex and Cursor read the same SKILL.md format; see Codex skills and Cursor skills.

Then set up the repo. Run /setup-matt-pocock-skills once per repo. It records where your issues live (GitHub, GitLab, local Markdown or anything you describe) in docs/agents/issue-tracker.md, which the Spec axis reads to fetch issues. Without it, the Spec axis still works if you pass a spec path.

How do you use it?

Start a fresh session, name the fixed point, and point it at the spec if it isn't linked from a commit. Example prompts (suggestions, not quotes from the repo):

• /code-review main
• /code-review since v2.3.0, against docs/specs/team-billing.md
• Review this branch since HEAD~4. The issue is #212.
• /code-review origin/main (after committing your work in progress)

Requirements. Git, and an agent that can run sub-agents in parallel. Fetching issues needs your tracker's CLI, such as gh for GitHub or glab for GitLab, per the setup docs. No API keys.

Fresh session, on purpose. The docs recommend reviewing in a new session rather than the one that wrote the code, quoting a reader: "Same context reviewing itself isn't review, it's confirmation bias with a slash command."

Tips and limits

From the repo's docs and Claude Code's docs, not from testing:

• The name clash. Claude Code ships a bundled /code-review that checks the diff for correctness bugs (commands reference). Per the skills docs, a personal or project skill named code-review replaces it, but the bundled alias /review still runs the built-in. With the plugin install, use the namespaced /mattpocock-skills:code-review. The repo's docs call this the most reported problem.
• Runaway sub-agents. The docs list a known open bug: a sub-agent can find /code-review again and spawn more agents (one report passed 50). The fix people use on forks is adding "Do not invoke /code-review or spawn additional agents: perform this review directly." to both sub-agent briefs. Watch the agent count if you run it unattended.
• Committed work only. The three-dot diff ignores staged and unstaged changes.
• Findings are leads. The skill does not re-verify sub-agent claims, so check each citation. The docs also warn it finds something new every run; act on findings with a cited rule, then stop.
• Good specs make good reviews. A vague spec makes the Spec axis vague. The repo's to-spec and to-tickets skills, and the grill me skill, help produce one.

Security audits. On 23 October 2026, the skill's skills.sh page showed Pass from Gen Agent Trust Hub and Socket, and Warn from Snyk, with no further detail on the page. The skill runs only git commands and sub-agents, but read it before installing.

Other code review skills

Two other popular review skills take different approaches:

• requesting-code-review in obra/superpowers sends a general-purpose sub-agent a template (code-reviewer.md) filled with a description, the plan or requirements and a base and head commit. You fix Critical issues at once, Important ones before moving on, and note Minor ones. It is built to run after each task in the Superpowers workflow.
• code-review-and-quality in addyosmani/agent-skills is one review across five axes: correctness, readability, architecture, security and performance. Its rule is to approve a change "when it definitely improves overall code health, even if it isn't perfect."

Both are MIT licensed. If you'd rather not install a skill, the code review prompts page has copy-ready prompts for security, performance and test-gap reviews.

Save it in your Kipt library

In Kipt, open Skills and paste the skill's GitHub folder link into the import box. Kipt reads the SKILL.md and its files; for a repo with many skills, like this one, it shows a list to pick from, so you can take code-review with setup-matt-pocock-skills. You can also drop the folder or a .zip. Your skills live in your own private GitHub repo and install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

A private copy is also a simple way to rename the skill if the clash with the built-in /code-review bothers you. Kipt is not affiliated with Matt Pocock or mattpocock/skills.

Similar skills

• Grill me skill: settle the plan before code exists, so the Spec axis has something to check.
• TDD skill: Matt Pocock's test-first loop for the build step.
• Superpowers skills: the workflow that includes requesting-code-review.
• Claude Code subagents: how the parallel reviewers work under the hood.