bmad-code-review

Review code changes adversarially and categorize findings into decision, patch, defer, or dismissal.

36|16|Updated May 1, 2026

Search Tags:#code review #edge cases #spec compliance #triage #git diff #adversarial analysis #actionable fixes

Ready to run

npx skills add https://github.com/tronghieu/lumina-wiki --skill bmad-code-review-tronghieu

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: bmad-code-review
Source: https://github.com/tronghieu/lumina-wiki/tree/main/.agents/skills/bmad-code-review
Command: npx skills add https://github.com/tronghieu/lumina-wiki --skill bmad-code-review-tronghieu

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

It helps teams catch defects, edge cases, and spec mismatches in code changes before they ship, by running multiple adversarial review perspectives and organizing the results into clear action categories.

Core Features & Use Cases

Blind Hunter: Reviews the diff adversarially (diff-only) to surface likely issues and overlooked failure modes.
Edge Case Hunter: Reviews the diff with project read access to find boundary-condition bugs and integration problems.
Acceptance Auditor: When a spec is provided, validates changes against acceptance criteria and loaded context docs, highlighting missing or contradictory behavior.

Use it when preparing PRs or changes for sprint stories that need rigorous verification against requirements, especially when you want findings grouped into decision-needed, patchable, deferred, and dismissed categories.

Quick Start

Ask the agent to run code review for the code changes you want reviewed (for example, “run code review on this diff”).

Frequently Asked Questions about bmad-code-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I run an adversarial code review on a git diff before merging?▼

Run adversarial code review on a git diff by triggering a multi-layer workflow that analyzes changes through Blind Hunter, Edge Case Hunter, and optional Acceptance Auditor perspectives. You get structured, triaged findings categorized for decision, patch, defer, or dismissal.

What is adversarial code review and how does it find edge cases in branch diffs?▼

Adversarial code review applies multiple analytical perspectives to branch diffs to actively hunt for defects and boundary-condition bugs. It uses a Blind Hunter for diff-only issues and an Edge Case Hunter with project read access to find integration problems and edge cases.

Can I validate staged or uncommitted changes against acceptance criteria from a spec?▼

Yes, validate staged or uncommitted changes against acceptance criteria using the Acceptance Auditor. When you provide a spec, it loads context docs, checks changes against requirements, and highlights missing or contradictory behavior in the diff.

How do I triage code review findings into actionable categories for a pull request?▼

Triage code review findings by normalizing adversarial analysis results into actionable categories. The workflow groups issues into decision-needed, patchable, deferred, and dismissed buckets so you can manage pull request changes clearly.

Does adversarial code review work with commit ranges and provided file lists?▼

Adversarial code review works with commit ranges, provided diffs, and file lists. You can review changes with optional spec context, applying the multi-layer review workflow to surface defects, edge cases, and spec mismatches across any provided diff or commit range.

What is the best way to catch spec mismatches in code changes before they ship?▼

The best way to catch spec mismatches is running the Acceptance Auditor workflow with a provided spec. It validates code changes against loaded context docs and acceptance criteria, highlighting any missing or contradictory behavior before shipping.