Ultimate Code Review
bradwindy/ultimate-code-review
23 agentes Opus analisam código em paralelo para revisão técnica profunda
Ver no GitHubAgent de revisão profunda de código que mobiliza um time de 23 especialistas Opus para analisar PRs e MRs em paralelo. Cada agente cobre um aspecto específico (bugs, segurança, performance, arquitetura, testes) e depois um revisor adversário desafia as conclusões. Integra com GitHub e GitLab.
Instalação
npx convez-ai bradwindy/ultimate-code-review1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Revisão técnica profunda de pull requests antes de merge
- Análise de segurança, performance e arquitetura de mudanças
- Identificação de bugs, side effects e riscos de concorrência
- Auditoria de cobertura de testes e conformidade com guidelines
Veredito

Solução robusta e completa para revisão técnica automatizada, ideal para times que precisam de análise profunda. Custo elevado (~$5-50 por review), mas compensado pela profundidade multi-especialista e síntese adversarial que reduz pontos cegos.
README
A tradução para o português deste README ainda está na fila.
Ultimate Code Review
Deep code review plugin for Claude Code. Spawns a 23-agent team to review PRs, MRs, or branch comparisons.
Quick Start
/plugin marketplace add bradwindy/ultimate-code-review
/plugin install ultimate-code-review@ultimate-code-review
/ultimate-code-review https://github.com/org/repo/pull/123
Requirements
- Claude Code v1.0.33+
- Opus model access (Pro plan or higher)
- Agent teams enabled:
CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 ghCLI (for GitHub PRs)glabCLI (for GitLab MRs)git(always required)
Installation
From Marketplace (recommended)
# Add the marketplace
/plugin marketplace add bradwindy/ultimate-code-review
# Install the plugin
/plugin install ultimate-code-review@ultimate-code-review
Local Development
git clone git@github.com:bradwindy/ultimate-code-review.git
claude --plugin-dir ./ultimate-code-review
Usage
# Review current branch vs upstream
/ultimate-code-review
# Review a specific PR
/ultimate-code-review https://github.com/org/repo/pull/123
# Review a specific MR
/ultimate-code-review https://gitlab.com/org/repo/-/merge_requests/456
# Review local branch comparison
/ultimate-code-review feature..main
# Also post summary to PR/MR
/ultimate-code-review --post https://github.com/org/repo/pull/123
Architecture
Three-phase review:
- Context Gathering: Team lead collects diff, builds change manifest
- Deep Analysis: 23 specialist agents review in parallel (all Opus, max effort)
- Synthesis: Findings merged, then adversarially challenged by devil's advocate
All agents verify technical claims against the web. No reliance on internal knowledge.
Agents
23 specialist agents with non-overlapping scopes:
Bug-Focused (1-7): Deep Bug Scanner, Side Effects Analyzer, Concurrency Reviewer, Silent Failure Hunter, Data Flow Analyzer, Memory & Resource Analyzer, Performance Analyzer
Security & Types (8-10): Security Auditor, Type Design Reviewer, API Contract Reviewer
Context & Quality (11-23): Git History Analyzer, Cross-PR Learning Agent, Guidelines Compliance, Comment Compliance Checker, Comment Quality Reviewer, Dependency Analyzer, Code Simplification, Style Consistency, Test Coverage Analyzer, Architecture Boundary, Logging & Observability, Migration & Deployment Risk, Scope Relevance Reviewer
Synthesis (24-25): Synthesizer, Devil's Advocate
Cost
This plugin uses 23+ Opus agents at max effort. Expect significant token usage ($5-50+ per review depending on diff size). This is by design -- depth over economy.
License
MIT -- see LICENSE for details.