What makes a 'Council' review different from a standard AI review?

Unlike single-model reviews, the Council approach uses multiple AI models to evaluate code independently. This peer-review system reduces bias, increases accuracy, and provides a more balanced verdict on complex logic.

Can I use this skill to review specific git commits?

Yes, the skill supports git diff inputs and requires a snapshot ID (commit SHA) to ensure reviews are reproducible and specific to the exact state of your code.

Is it possible to focus the review on specific areas like security?

Yes, you can use the --rubric-focus flag (e.g., '--rubric-focus Security') to direct the council's attention to specific priorities during the evaluation.

What dimensions does the skill use to score my code?

It uses five primary dimensions: Accuracy (35%), Completeness (20%), Clarity (20%), Conciseness (15%), and Relevance (10%).

Council Code Review

Name: Council Code Review
Author: amiable-dev

byamiable-dev

•

コラボレーションツール

Performs multi-model, peer-evaluated code reviews to provide structured feedback and automated quality gates for pull requests.

概要

The Council Code Review skill leverages a multi-model 'LLM Council' to provide objective, high-confidence evaluations of your source code. By aggregating perspectives from multiple AI models, it analyzes code changes against a weighted rubric covering accuracy, completeness, clarity, conciseness, and relevance. This skill is ideal for teams looking to automate the first pass of PR reviews, enforce coding standards, and catch blocking logic errors before they reach human reviewers, providing detailed JSON reports with line-specific issue tracking.

主な機能

Identification of line-specific blocking issues and improvement suggestions
Weighted scoring system based on ADR-016 standards
Structured JSON output including pass/fail verdicts and confidence levels
Multi-model peer evaluation via LLM Council integration
1 GitHub stars
Support for both full-file analysis and focused git diff reviews

ユースケース

Automating the initial review of pull requests to identify bugs and logic errors
Generating objective code quality metrics for CI/CD gates
Validating code changes against project-specific quality rubrics and ADRs

概要

主な機能

Identification of line-specific blocking issues and improvement suggestions
Weighted scoring system based on ADR-016 standards
Structured JSON output including pass/fail verdicts and confidence levels
Multi-model peer evaluation via LLM Council integration
1 GitHub stars
Support for both full-file analysis and focused git diff reviews

ユースケース

Automating the initial review of pull requests to identify bugs and logic errors
Generating objective code quality metrics for CI/CD gates
Validating code changes against project-specific quality rubrics and ADRs