Council Code Review Skill

Get multiple AI perspectives on code changes with structured, actionable feedback.

When to Use

• Review pull requests before merging
• Get code quality feedback on implementations
• Identify potential issues across multiple dimensions
• Validate changes against coding standards

Workflow

1. Prepare Input: Provide file paths or git diff
2. Invoke Review: Call mcp:llm-council/verify with code-review rubric
3. Process Feedback: Receive structured scores and issue list
4. Address Issues: Fix blocking issues before proceeding

Input Formats

Supports both:

• file_paths: List of files to review (full file analysis)
• git_diff: Unified diff format for change-focused review
• snapshot_id: Git commit SHA (required for reproducibility)

Rubric (ADR-016)

Output Schema

{
  "verdict": "pass|fail|unclear",
  "confidence": 0.82,
  "rubric_scores": {
    "accuracy": 7.5,
    "completeness": 8.0,
    "clarity": 9.0,
    "conciseness": 8.5,
    "relevance": 9.0
  },
  "blocking_issues": [
    {
      "severity": "major",
      "file": "src/api.py",
      "line": 42,
      "message": "Missing input validation"
    }
  ],
  "suggestions": [...],
  "rationale": "Overall, the code is well-structured..."
}

Example Usage

# Review specific files
council-review --file-paths "src/main.py,src/utils.py" --snapshot abc123

# Review git diff
council-review --git-diff "$(git diff HEAD~1)" --snapshot $(git rev-parse HEAD)

# Review with custom focus
council-review --rubric-focus Security --file-paths "src/auth.py"

Progressive Disclosure

• Level 1: This metadata (~200 tokens)
• Level 2: Full instructions above (~800 tokens)
• Level 3: See references/code-review-rubric.md for detailed scoring anchors

Related Skills

• council-verify: General verification
• council-gate: CI/CD quality gate