bmad-code-review skill (BMad Method): parallel code review with triage

bmad-code-review has several independent reviewers examine a code change, then triages the findings and offers to fix them.

Installs: 13.6K; on skills.sh
Stars: 54.9K; on GitHub
Files: 12; besides SKILL.md
Updated: Nov 2026; last repo commit

bmad-code-review is a skill in the BMad Method. Say "run code review" or "review this code" and it finds your change (a PR, branch, commit range or uncommitted work), writes the diff to a temp file, and launches review lenses as subagents. It then verifies each finding, drops false ones, groups duplicates by root cause and sorts them as patch, decision needed or defer.

It does not run on its own. SKILL.md starts a uv run command against _bmad/scripts/render_skill.py, so BMad must be set up in your project first (the install box covers this) and uv must be installed. By default it asks before each major step, and its fix step can apply patches to your code if you choose that.

Good for

Developers who want several independent reviews of a PR or branch
Teams already using the BMad Method
People who want a triaged, de-duplicated findings list instead of raw review output

Install bmad-code-review

Claude Code

Copy this prompt.

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

/plugin marketplace add bmad-code-org/bmad-plugins

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 uv, Node.js 20.12 or newer, Python 3.10 or newer.

The BMAD-METHOD README calls the V7 line a preview alpha with breaking changes between updates. It also lists a stable v6.12 line installed with npx bmad-method install, which this skill page does not cover. The README names the standalone plugin bmad-core-tools, while the bmad-plugins marketplace lists it as bmad-toolbox.

From the project README. Source.

What's in the skill

When it loads

Review code changes with several independent reviewers in parallel, then triage and present the findings. Use when the user says "run code review" or "review this code"

Key facts

Author: bmad-code-org (GitHub account that owns the repo)
Source repo: bmad-code-org/bmad-method
Installs on skills.sh: 13,636 as of 2026-11-02
GitHub stars: 54,859 (whole repo) as of 2026-11-02
Last repo commit: 2026-11-01
Files: 13 files including SKILL.md (5 step files, workflow.md, 2 toml files, 2 reference files, 2 review prompts)
skills.sh audits: Gen Agent Trust Hub pass, Socket pass, Snyk warn
Part of: BMad Method, plugin bmad-method (21 skills in the bmad-plugins marketplace)
Needs: uv, a BMad-initialised project (_bmad/ folder), and subagents if your tool has them
License: MIT, stated in the BMAD-METHOD README (GitHub does not detect a license for the repo)

What it includes

Quick (one lens) or thorough (default, four lenses); name one in your request.
Review modes
Review by inspection only; patches are handed off as a Markdown block, not applied.
Read · No-code review
For bots and CI: never asks, and applies every patch finding unless it is no-code.
Write · Non-interactive run
Finds the target (PR, branch, commits, uncommitted), writes diff and claims temp files.
Runs code · Step 1: Gather context
Runs the lenses as parallel subagents, from blind-hunter to intent-alignment.
Read · Step 2: Review
Verifies each finding, grades severity, rejects false ones, groups by root cause.
Files · Step 3: Triage
Sorts findings into patch, decision_needed or defer.
Step 4: Route
Shows results, then offers: apply every patch, leave as action items, or walk through.
Write · Step 5: Present

From the official docs

Permissions

Medium risk

The skill runs a uv script from your project, launches subagents and, if you approve, edits your code. Start with a no-code review and read each prompt before you approve fixes.

What it can do

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

Watch out for

It can apply patches to your code
In interactive mode the fix step offers "Apply every patch" with no per-finding confirmation. Choose "Walk through each patch" or a no-code review if you want control. github.com
Non-interactive runs apply patches without asking
If the request says no questions, or a bot or CI job runs it, SKILL.md appends workflow.interactive=false. Step 5 then applies every patch finding without asking, unless you asked for a no-code review. github.com
It runs a uv script from your project
SKILL.md runs uv run --no-cache against _bmad/scripts/render_skill.py. That script is created by BMad setup, so review what setup installs. github.com
It writes findings into project files
With a plan file, step 5 appends a Code Review section to it. Deferred findings are appended to deferred-work.md in the initiative folder. Step 3 writes a triage file to a temp location outside the repository. github.com
It writes temp files and may call gh
Step 1 writes the diff and a claims file to uniquely named temp files outside the repository, may resolve PRs with gh pr view, and runs tickets.py from the BMad folder. github.com
The plugin route adds 21 skills
The bmad-method plugin in the bmad-plugins marketplace holds 21 skills. The tree shows no hooks or MCP config in it, only plugin metadata. github.com

Technical details

Turning on read-only	Ask for a no-code review: SKILL.md appends --set workflow.fixes=handoff, and the workflow then says to review by inspection only and not modify code, run builds or tests.
Scopes	Your git repo and _bmad/ folder GitHub CLI (gh) for pull requests

Questions

How do I install bmad-code-review?

Use the install box above. The skill needs more than its own folder: it calls scripts in your project's _bmad/ folder, so install the BMad Method and ask its bmad skill to run bmad setup. You also need uv.

What happens if BMad is not set up?

SKILL.md says that if the render script does not exist, BMad is not set up in the project. The skill offers to set it up with the bmad skill, or to install that skill first, then runs the command again.

Will it change my code?

In the default interactive mode, only if you choose it: the fix step asks how to handle patch findings, with an option to apply every patch without per-finding confirmation. A no-code review never edits code and hands the findings to a coding agent instead. A non-interactive run (a bot or CI job) applies every patch finding without asking, unless it is no-code.

What are the review lenses?

The thorough review runs four: blind-hunter, edge-case-hunter, verification-gap and intent-alignment. The quick review runs one. They run as subagents, and the workflow asks once for permission to use them. If none are available it writes each prompt to a file for you to run elsewhere.

Does bmad-code-review need an API key?

The skill does not mention one. It runs a local uv script and git commands, and can call gh pr view to resolve a pull request, so GitHub CLI access helps for PR reviews.

bmad-code-review vs a single-prompt review skill?

This one is a multi-step workflow: parallel reviewers, then triage and routing. For a lighter one-pass review, see code-review-expert, which reads your git diff in one pass and waits for you before fixing.

Is bmad-code-review safe to install?

It is a Markdown workflow plus config, but it runs a uv script from your project, writes temp files outside the repo and can edit your code if you choose to apply patches. Read the safety notes below, and start with a no-code review.

Information

updated Nov 2, 2026

Repository: bmad-code-org/bmad-method
Folder: skills/bmad-code-review

bmad-code-org/bmad-method skills.sh

Sources

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

Show all 10

BMAD-METHOD README · github.com
bmad-plugins README · github.com
bmad-code-review SKILL.md · github.com
bmad-code-review workflow.md · github.com
bmad-code-review customize.toml · github.com
bmad-code-review on skills.sh · skills.sh
code-review-expert SKILL.md · github.com
bmad-code-review step 5 · github.com
bmad-code-review step 1 · github.com
step-03-triage.md · github.com