convez.ai
AgenteDevOps & InfraestruturaAvançado
tomas-rampas

Claude Code Agentic Framework

tomas-rampas/claude-agentic-framework

19 agentes especializados com validação multi-fase para toda a cadeia de desenvolvimento

Ver no GitHub
InicialPluginScore · de 0 a 100

Framework de configuração para Claude Code CLI que estende a plataforma com 19 agentes especializados, 44 gates de qualidade e sistema de aprendizado automático. Integra 5 servidores MCP para inteligência de código, operações de arquivo, documentação e execução de shell.

Instalação

$npx convez-ai tomas-rampas/claude-agentic-framework

1 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

Estrelas
13
Forks
3
Acompanhando
0
Saúde
28%
Último commit
há 3 semanas
Versão
v5.2.0
Issues abertas
0
Contribuidores
4

Casos de uso

  • Roteamento inteligente de tarefas de desenvolvimento entre agentes especializados
  • Validação de qualidade em múltiplas fases com 44 quality gates automatizados
  • Captura automática de padrões e memória institucional para projetos
  • Orquestração de workflows complexos de arquitetura de microserviços e DevOps
  • Integração com MCP servers para code intelligence, .NET tooling e automação shell

Veredito

Framework robusto para orquestração de desenvolvimento agentic no Claude Code CLI, com arquitetura ambiciosa de validação multi-fase e memória institucional. Requer investimento significativo em configuração MCP e compreensão profunda do ecossistema Claude para máximo aproveitamento.

README

A tradução para o português deste README ainda está na fila.

Claude Code CLI - Agent Execution Framework

A configuration framework for Claude Code CLI that adds 19 specialized agents, automated pattern capture, institutional memory, and intelligent task routing.

This is NOT a standalone tool — it requires Claude Code CLI as the underlying platform.


Overview

This framework extends Claude Code CLI with:

  • 19 Specialized Agents covering the full development lifecycle
  • 44 Quality Gates with multi-phase validation per agent
  • Self-Learning System — automated pattern capture, institutional memory, and semantic search
  • MCP Integration — 5 MCP servers for code intelligence, file ops, documentation, .NET tooling, and shell execution

Prerequisites

RequirementPurpose
Claude Code CLIAgent execution platform (required)
Node.js/npmRunning MCP servers via npx
GitVersion control

Install Claude Code CLI

# macOS
brew install --cask claude-code

# Linux / WSL
curl -fsSL https://anthropic.com/install-claude.sh | sh

# Verify
claude --version

Configure MCP Servers

Create or update the project-level .mcp.json (or ~/.claude/mcp.json):

{
  "mcpServers": {
    "serena": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-serena"]
    },
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/home/your-username"]
    },
    "context7": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-context7"]
    },
    "dotnet": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-dotnet"]
    },
    "bash": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-bash"]
    }
  }
}

Replace /home/your-username with the directory you want the filesystem server scoped to.

MCP ServerPurpose
serenaSemantic code intelligence, symbol operations + institutional memory
filesystemEnhanced file operations for large files and atomic updates
context7External documentation and best practices
dotnet.NET/C# build, test, and project tooling
bashShell command execution for automation tasks

Installation

# Clone into Claude Code CLI config directory
git clone <your-framework-repo> ~/.claude
cd ~/.claude

# Validate framework integrity
./scripts/validate-framework.sh

Quick Start

# Start Claude Code CLI
claude

# Or start with a specific task
claude "Create a REST API in Rust with JWT authentication"

Tasks are automatically routed to the appropriate agent. Examples:

"Create user stories for authentication feature"        → product-owner
"Design microservices architecture"                      → system-architect
"Implement JWT authentication in Rust"                   → rust-expert
"Create ASP.NET Core REST API with Entity Framework"     → csharp-expert
"Build gRPC microservice in Go"                          → go-expert
"Build a Fisher Transform indicator for MetaTrader 5"    → mql-trading-dev
"Review pull request for code quality"                   → code-review-gatekeeper
"Set up Kubernetes deployment with Helm"                 → devops-orchestrator
"Write API documentation for REST endpoints"             → technical-docs-writer
"Analyze codebase security vulnerabilities"              → comprehensive-analyst

Agents

Planning & Requirements

AgentFocus
product-ownerUser stories, backlog management, acceptance criteria

Architecture & Analysis

AgentFocus
system-architectSystem design, technology selection, SOLID principles
comprehensive-analystSecurity audits, performance profiling, investigation
code-review-gatekeeperCode review, quality gates, standards compliance

Language Experts

AgentFocus
rust-expertSystems programming, memory safety, async/await
csharp-expertASP.NET Core, Entity Framework, Azure
go-expertMicroservices, gRPC, Kubernetes operators
java-expertSpring Boot, Maven/Gradle, enterprise apps
python-expertDjango/Flask, data science, automation
typescript-expertReact/Next.js, Node.js, frontend/backend
mql-trading-devMQL4/MQL5, C/C++ DLLs, MetaTrader trading systems

Scripting & Automation

AgentFocus
bash-expertShell scripting, Linux automation, CLI tools
powershell-expertWindows administration, Azure/AWS automation

Specialized Domains

AgentFocus
database-specialistSchema design, query optimization, SQL/NoSQL
frontend-specialistReact/Vue/Angular, responsive design, accessibility
security-specialistVulnerability assessment, compliance, auth
uiux-specialistUser flows, design systems, wireframing

Infrastructure & Documentation

AgentFocus
devops-orchestratorCI/CD, containers, cloud deployment, monitoring
technical-docs-writerAPI docs, user guides, architecture docs

Project Structure

~/.claude/
├── CLAUDE.md                # Agent execution rules and task routing
├── claude.json              # Agent configuration (v3.0.0)
├── .mcp.json                # MCP server definitions (serena, filesystem, context7, dotnet, bash)
├── agents/                  # 19 agent definitions (.md with YAML frontmatter)
├── commands/                # 6 commands (delegate, analyze-framework, list-agents, etc.)
├── hooks/                   # 44 quality gates (agent-specific + framework-wide)
├── shared/                  # Shared configs (base-config, mcp-config, agent-patterns, memory-categories)
├── skills/                  # 14 operational skills (validation, debugging, analytics, scaffolding)
├── scripts/                 # 3 validation scripts
└── security-check.sh        # Security validation

Management Commands

CommandPurpose
/delegateIntelligent task routing to specialized agents
/analyze-frameworkFramework health checking and validation
/list-agentsAgent catalog with filtering and multiple output formats
/validate-hooksHook coverage and consistency verification
/agent-statusReal-time agent monitoring and performance metrics
/quality-reportQuality metrics, trend analysis, and reporting

Skills

The framework ships 14 operational skills in skills/ that support validation, debugging, and analytics:

SkillPurpose
framework-validatorComprehensive framework health, consistency, and integrity checks
config-validatorValidate framework configuration files for correctness
dependency-checkerCheck framework dependencies and tool requirements
agent-debuggerDebug agent routing, execution, and integration issues
agent-routing-advisorIntelligent task-to-agent mapping recommendations
hook-auditorAudit hook coverage, consistency, and effectiveness
hook-config-generatorGenerate new validation hooks from templates
code-scaffolderGenerate idiomatic project scaffolding for supported languages
refactoring-advisorIdentify refactoring opportunities and improvement patterns
git-workflow-assistantGuide branching, commit conventions, and PR workflows
migration-assistantAssist with framework version migrations
performance-analyticsAnalyze agent and hook performance metrics
quality-reporterGenerate quality metrics reports and trend analysis
workflow-visualizerVisualize multi-agent workflows and coordination patterns

Self-Learning System

The framework learns from every interaction through 8 memory categories:

  • Workflow Patterns — successful agent combinations and sequences
  • Solution Patterns — proven solutions for common problems
  • Optimization Insights — performance improvements and efficiency gains
  • Lessons Learned — knowledge from successes and failures
  • Code Patterns — reusable architectural and implementation patterns
  • Project Context — project-specific standards and decisions
  • Quality Insights — quality-related discoveries and improvements
  • Integration Knowledge — system integrations and API knowledge

Pattern capture, lesson extraction, and optimization tracking are handled automatically via hooks in the hooks/ directory.


Troubleshooting

Claude Code CLI not found:

which claude  # If missing, reinstall per the Prerequisites section

MCP servers not available:

cat ~/.claude/.mcp.json          # Check configuration
npx -y @modelcontextprotocol/server-serena --version  # Test server

Agents not found:

ls -1 agents/*.md | wc -l  # Should show 19
./scripts/validate-agents.sh

Delegation not working:

cat commands/delegate.md | head -20  # Verify command exists

Resources


Built for Claude Code CLI • 19 Specialized Agents • 44 Quality Gates • 14 Skills • Self-Learning • v3.0.0

Ver o original no GitHub

Bom para

Relacionados