Dual Review

Sends your work to a second AI agent (via the claude-mcp-bridge or codex MCP server) for independent structured review. Iterates on NEEDS_REVISION feedback up to 3 rounds before implementing.

Workflow

1. Explore — Analyze the codebase to understand the area being changed
2. Plan — Draft an implementation plan
3. Pre-review — Send the plan to the reviewer MCP for structured review
4. Iterate — If NEEDS_REVISION, address the issues and re-submit (up to 3 rounds)
5. Implement — After APPROVED, implement the changes
6. Post-review — Send the final diff for post-implementation review
7. Summary — Present a summary with the full review history

Usage

/dual-review <description of what you want to build or change>

How It Works

The skill uses the review tool from the Claude MCP bridge (or Codex MCP) to get an independent second opinion. The reviewer runs in a separate Claude instance with read-only access to the codebase.

Review Response Schema

The reviewer returns structured JSON:

{
  "verdict": "APPROVED | NEEDS_REVISION",
  "issues": [
    {
      "severity": "critical | major | minor",
      "description": "What the issue is",
      "recommendation": "How to fix it"
    }
  ],
  "suggestions": ["Optional improvement ideas"]
}

Iteration Rules

• critical issues must be addressed before re-submitting
• major issues should be addressed
• minor issues are at your discretion
• After 3 NEEDS_REVISION rounds, present the issues to the user for a decision

Configuration

The reviewer MCP server must be running. Default: http://localhost:8940/mcp

Configure in .mcp.json:

{
  "mcpServers": {
    "claude_reviewer": {
      "type": "streamable-http",
      "url": "http://localhost:8940/mcp"
    }
  }
}

Environment Variables

• CLAUDE_REVIEW_MODEL — Model for the reviewer (default: opus)
• CLAUDE_REVIEW_MAX_TURNS — Max turns per review (default: 8)
• CLAUDE_REVIEW_CWD — Working directory for the reviewer
• CLAUDE_REVIEW_ALLOWED_TOOLS — Comma-separated tools the reviewer can use