Custom Agents Orchestrator
jayfoeelite/custom-agents-orchestrator
Orquestrador de 46 agentes IA para projetos complexos com verificação triangulada
Ver no GitHubSistema sofisticado de orquestração de agentes IA implementando a metodologia SPARC com 46 agentes especializados coordenados para design, implementação e validação de projetos complexos. Utiliza Triangulação Cognitiva e avaliação RULER para garantir alinhamento com a intenção do usuário em cada fase do workflow.
Instalação
npx convez-ai jayfoeelite/custom-agents-orchestrator1 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
- Desenvolvimento de sistemas de software com requisitos complexos e multidisciplinares
- Implementação TDD com validação de qualidade automatizada em cada fase
- Verificação de alinhamento entre intenção do usuário e implementação final
- Geração de múltiplas trajetórias de implementação com ranking comparativo
- Auditorias de segurança e concorrência em arquiteturas distribuídas
Veredito

Projeto ambicioso e bem-estruturado que implementa metodologia rigorosa (SPARC) com múltiplos agentes especializados. Ideal para equipes que precisam de orquestração completa de desenvolvimento, mas exige curva de aprendizado significativa e integração cuidadosa com fluxos existentes.
README
A tradução para o português deste README ainda está na fila.
Custom Agents Orchestrator
A sophisticated AI Agent Orchestration System implementing the SPARC (Specification, Pseudocode, Architecture, Refinement, Completion) methodology with advanced Cognitive Triangulation, RULER-based quality evaluation, and specialized domain agents.
🌟 Overview
This project coordinates multiple specialized AI agents to collaboratively design, implement, test, and validate complex software projects through a structured, verifiable workflow that ensures alignment with user intent at every stage.
✨ Key Features
🏗️ SPARC Methodology Framework
- Specification Phase: Comprehensive requirements gathering and acceptance criteria
- Pseudocode Phase: Language-agnostic logical blueprints
- Architecture Phase: High-level system design with resilience patterns
- Refinement Phase: Iterative TDD implementation with quality gates
- Completion Phase: Documentation, maintenance, and final verification
🔀 Cognitive Triangulation
Multi-stage verification ensuring alignment between:
- User's Core Intent → User Stories → Specifications → Pseudocode → Architecture → Implementation → Tests
- Enforced by the Devil's Advocate agent at critical checkpoints
- Prevents requirement drift and ensures implementation fidelity
📊 RULER Quality Evaluation
- LLM-as-Judge methodology for comparative quality assessment
- Generates multiple implementation trajectories
- Ranks solutions based on efficiency, clarity, and maintainability
- Ensures optimal implementation selection
🤖 46 Specialized Agent Modes
Orchestrators (11 agents)
- uber-orchestrator: Master conductor managing overall project flow
- orchestrator-goal-clarification: Intent validation and synthesis
- orchestrator-sparc-*-phase: Phase-specific coordinators (Specification, Pseudocode, Architecture)
- orchestrator-sparc-refinement-*: Testing and implementation managers
- orchestrator-sparc-completion-*: Documentation and maintenance coordinators
- orchestrator-simulation-synthesis: Multi-method verification orchestrator
- orchestrator-state-scribe: Intelligent state interpreter and recorder
Workers (9 agents)
- spec-writer-comprehensive: Modular specification creation
- spec-writer-from-examples: User story extraction from examples
- pseudocode-writer: Detailed logic blueprints
- coder-test-driven: TDD implementation specialist
- coder-framework-boilerplate: Project scaffolding generator
- tester-tdd-master: Test implementation expert
- tester-acceptance-plan-writer: High-level test strategy
- docs-writer-feature: Feature documentation specialist
- research-planner-strategic: Adaptive multi-arc research strategist
Validators & Auditors (5 agents)
- devils-advocate-critical-evaluator: Cognitive Triangulation enforcer
- auditor-concurrency-safety: Race condition detector
- auditor-financial-logic: Capital & risk validator
- validator-api-integration: External dependency verifier
- validator-performance-constraint: Sub-100ms latency enforcer
Quality & Analysis (4 agents)
- ruler-quality-evaluator: LLM-as-judge quality arbiter
- bmo-system-model-synthesizer: As-built system documentation
- bmo-holistic-intent-verifier: Final triangulation verifier
- optimizer-module: Code quality and performance enhancement
Domain Specialists (17 agents)
Financial trading and specialized vertical agents including performance fee calculators, client portal generators, audit trail recorders, and more.
🚀 Getting Started
Prerequisites
- Python 3.7+ (for validation tooling)
- Git
- RooCode or compatible AI coding assistant
Installation
- Clone the repository:
git clone https://github.com/YOUR_USERNAME/custom-agents-orchestrator.git
cd custom-agents-orchestrator
- Install Python dependencies for tooling:
pip install -r tools/requirements.txt
- (Optional) Validate agent definitions:
python tools/validate-agents.py
📁 Project Structure
custom-agents-orchestrator/
├── agents/ # 46 agent YAML definitions
├── docs/ # Documentation
│ ├── research/ # Research findings and decisions
│ ├── TUTORIAL.md # Interactive tutorial
│ ├── UBER_ORCHESTRATOR_MODE_DELEGATION_GUIDE.md
│ └── project_plan.md # Implementation roadmap
├── memory-bank/ # Project context and state
│ ├── productContext.md
│ ├── activeContext.md
│ ├── systemPatterns.md
│ ├── decisionLog.md
│ └── progress.md
├── schemas/ # JSON Schema for validation
│ └── agent-mode-schema.json
├── tools/ # Validation and analysis tools
│ ├── validate-agents.py
│ ├── generate-dependency-graph.py
│ └── merge-agents.py
├── SETUP_GUIDE.md # Detailed setup instructions
└── validation_report.md # Agent validation results
🛠️ Usage
Working with Agent Modes
All 46 agent modes are defined in the agents/ directory as YAML files. Each agent has:
- Unique slug identifier
- Role definition
- Custom instructions
- Group permissions
- Communication protocols
Validation Tools
Validate all agents:
python tools/validate-agents.py
Generate dependency graph:
python tools/generate-dependency-graph.py
Merge agents for RooCode:
python tools/merge-agents.py
Agent Communication Protocol
All agents use standardized routing headers:
To: [recipient-agent-slug]
From: [sender-agent-slug]
📚 Documentation
- Tutorial: Interactive guide to creating and validating agents
- Uber Orchestrator Guide: Complete workflow documentation
- Setup Guide: Detailed installation and configuration
- Project Plan: V2 implementation roadmap
- Agent Dependency Graph: Visual workflow representation
🔄 Workflow
- Goal Clarification: uber-orchestrator → orchestrator-goal-clarification
- Research & Planning: → research-planner-strategic
- Triangulation Check #0: → devils-advocate-critical-evaluator
- Specification Phase: → orchestrator-sparc-specification-phase
- Pseudocode Phase: → orchestrator-sparc-pseudocode-phase
- Architecture Phase: → orchestrator-sparc-architecture-phase
- Refinement Loop (per feature):
- orchestrator-sparc-refinement-testing
- orchestrator-sparc-refinement-implementation
- Ultimate Triangulation Audit:
- bmo-system-model-synthesizer
- bmo-holistic-intent-verifier
- Final Verification: → orchestrator-simulation-synthesis
🧪 Testing Strategy
- State-Based Classical TDD: Focus on observable outcomes
- Multi-Methodology: Unit, Integration, Property-Based, Chaos, Metamorphic
- No Mock Internal Collaborators: Maximize refactoring flexibility
- Comprehensive Coverage: Edge cases, error scenarios, performance constraints
🤝 Contributing
- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Validate your changes (
python tools/validate-agents.py) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
📊 Project Status
- ✅ Phase 1: Agent Definition & Validation (Complete)
- ✅ Validation Tooling (Complete)
- ✅ Documentation & Tutorial (Complete)
- ✅ RooCode Integration (Complete)
- 🔄 Phase 2: Full Automation via Anthropic API (In Planning)
See V2_IMPLEMENTATION_STATUS.md for detailed progress.
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
🙏 Acknowledgments
- Built on the SPARC methodology
- Implements Cognitive Triangulation principles
- Uses RULER (LLM-as-Judge) quality evaluation
- Powered by RooCode AI coding assistant
📧 Contact
For questions, issues, or contributions, please open an issue on GitHub.
Note: This is an intelligent orchestration system for AI-assisted software development. It requires understanding of AI agent workflows and the SPARC methodology for effective use.