bmad-code-review

Reviews code changes with parallel independent reviewers, then triages and presents findings.

Updated Mar 1, 2026
One-click install
npx skills add https://github.com/Hexalith/Hexalith.Parties --skill bmad-code-review-hexalith
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: bmad-code-review
Source: https://github.com/Hexalith/Hexalith.Parties/tree/main/.agents/skills/bmad-code-review
Command: npx skills add https://github.com/Hexalith/Hexalith.Parties --skill bmad-code-review-hexalith

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Code reviews often miss edge cases, verification gaps, and spec deviations because a single reviewer cannot cover every angle. This Skill runs several independent review layers in parallel against a diff, then verifies, triages, and routes every finding so only real, actionable issues reach the developer. ## Core Features & Use Cases - Parallel adversarial review layers: Launches independent reviewers (Blind Hunter, Edge Case Hunter, Verification Gap, Acceptance Auditor) against a unified diff file, each with its own methodology. - Evidence-based triage: Verifies every finding against the actual code, assigns verdicts (high/medium/low/false/maybe-false), groups by root cause, and routes results into decision-needed, patch, or defer buckets. - Story and sprint integration: Writes findings into story files, updates story status, and syncs sprint-status.yaml when a story key is available. - Use Case: After finishing a feature branch, ask for a code review; the Skill builds the branch diff, runs all review layers, rejects false positives, and offers to apply the surviving patches automatically. ## Quick Start Ask the assistant to run a code review on your current branch diff against main and apply the confirmed patches.

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 automated code review on a git branch?

Invoke the review workflow and choose a diff source such as a branch diff against main, a commit range, staged changes, or uncommitted changes. The Skill writes the unified diff to a temp file and runs parallel review layers against it.

What does a multi-layer AI code review check for?

It checks for missing edge-case handling, regressions in deleted code, falsified claims in commit messages, verification gaps where no test would catch a breakage, and deviations from the specification's acceptance criteria.

Can I customize the review layers or add my own checks?

Yes. The customize.toml file defines each review layer's instruction, and team or user TOML overrides can replace instructions, disable layers, or add persistent facts and activation steps using structural merge rules.

Does the code review work without a specification file?

Yes. You can explicitly run in no-spec mode, which skips the Acceptance Auditor layer and reclassifies decision-needed findings as patch or defer. Spec-based reviews additionally validate against acceptance criteria.

What happens if subagents are not available in my environment?

The workflow writes each reviewer prompt with the full diff content into the implementation artifacts directory and pauses, asking you to run each prompt in a separate session and paste the findings back to resume triage.

Why did the review reject some of the findings it reported?

Triage verifies every finding against the actual code before accepting it. Findings whose claimed bad outcome does not occur are rejected as false, and low-value findings with complex fixes are rejected, with refutations listed in the report.