Claude Agents Pro

February 9, 2026 · View on GitHub

License: Apache-2.0 Agents Skills Commands Hooks Quality

Personal collection of production-ready Claude Code agents with automated quality validation and multi-tier orchestration.

System Architecture

Core Components

  • 61 Specialized Agents: Tiered collection from orchestration to domain experts
  • 120 Progressive-Disclosure Skills: Modular knowledge packages across 17 domains (~70% token savings)
  • 35 Slash Commands: Quick access workflows using latest Claude models (Opus 4.6, Sonnet 4.5)
  • 10 Event-Driven Hooks: Security, quality, git, and workflow automation
  • Automation Scripts: Install, verify, and score agent quality
  • MCP Integration: Memory persistence and advanced reasoning capabilities
  • Quality Framework: Automated validation, scoring rubric (70+ minimum, 85+ production)

📊 Agent Tiers (61 Total)

  • Tier 0 (Meta): Multi-agent orchestration and workflow coordination
  • Tier 1 (Foundation): Core engineering (API, domain modeling, testing, review, debugging, performance, system design)
  • Tier 2 (Development): Language/platform specialists (frontend, mobile, Python, TypeScript, Angular, Django, Rails, Laravel, Java, Kotlin, C#)
  • Tier 3 (Specialists): Domain experts (AWS/Azure/GCP cloud, backend, database, DevOps, observability, SRE, data, full-stack)
  • Tier 4 (Experts): Machine learning, MLOps, LLM architecture, and prompt engineering
  • Tier 5 (Platform): Developer experience optimization, Git workflow automation
  • Tier 6 (Integration): Research, technical documentation, and knowledge management
  • Tier 7 (Quality): Security architecture, compliance, and audit
  • Tier 8 (Finance): Trading, risk management, compliance, quantitative analysis, portfolio management

Quick Start

Installation

# Clone the repository
git clone https://github.com/ubehera/claude-agents-pro.git
cd claude-agents-pro

# Install agents (user-scoped, available across all projects)
./scripts/install-agents.sh --user

# Or install project-scoped (only in current directory)
./scripts/install-agents.sh --project

# Restart Claude Code to load agents

Validation & Quality

# Validate agent structure and frontmatter
./scripts/verify-agents.sh

# Score agent quality (minimum 70/100, production 85/100)
python3 scripts/quality-scorer.py --agents-dir agents
python3 scripts/quality-scorer.py --agent agents/01-foundation/api-platform-engineer.md

Agent Discovery

Once installed, agents are automatically selected based on task context:

  • Direct Invocation: "Use the API platform engineer to design a REST API"
  • Implicit Routing: "Help me debug this performance issue" → performance-optimization-specialist
  • Orchestration: Complex tasks routed through orchestration-coordinator → specialists

See agents/README.md for complete trigger patterns and invocation examples.

Workflow Patterns

🔄 Agent Orchestration

  1. Discovery: orchestration-coordinator decomposes problems and selects specialists
  2. Implementation: Foundation/specialist tiers execute domain work
  3. Quality Gates: Review agents validate readiness (test, security, performance)
  4. Operations: DevOps agents finalize deployment and monitoring

🎯 Common Workflows

  • Feature Development: /workflows:feature-development → DDD workflow with quality gates
  • API Design: /01-foundation:api → api-platform-engineer with OpenAPI/GraphQL
  • Security Review: /quality:security-audit → security-architect assessment
  • Performance: /quality:performance → performance-optimization-specialist
  • Debugging: /01-foundation:debug → error-diagnostician intelligent debugging

Key Documentation

  • agents/README.md: Complete agent catalog with invocation triggers
  • agents/AGENT_CHECKLIST.md: Pre-flight checklist for agent updates
  • agents/TESTING.md: Comprehensive testing procedures
  • skills/README.md: Skills catalog (120 skills across 17 domains)
  • hooks/README.md: Hook development guide (10 event-driven hooks)
  • mcp-servers/README.md: MCP server registry (20+ verified servers)
  • commands/README.md: Slash command catalog and usage
  • patterns/orchestration/: Multi-agent coordination patterns
  • prompts/CLAUDE.md: Operating instructions for Claude Code
  • SYSTEM_OVERVIEW.md: Architecture, tier system, and coordination patterns

User-Level Configuration Template

meta-CLAUDE.md - Template for user-level Claude Code configuration (deploys to ~/.claude/CLAUDE.md)

This file provides:

  • Persona and communication style settings
  • Memory management and session continuity patterns
  • Tool usage optimization and MCP integration
  • DDD workflow patterns and quality gates
  • Agent orchestration and delegation examples

Important: This is NOT project-level documentation. When deployed to ~/.claude/, it applies to all Claude Code sessions across all projects. The "meta-" prefix is removed during installation.

MCP Configuration

Location: ./.mcp.json (project-level). Claude Code merges this with your user-level ~/.mcp.json.

Included by default:

  • memory: Persistent knowledge graph for session continuity
  • sequential-thinking: Complex problem decomposition

Enable additional servers by editing ./.mcp.json:

{
  "mcpServers": {
    "aws-docs": {
      "command": "uvx",
      "args": ["awslabs.aws-documentation-mcp-server@latest"],
      "env": { "AWS_REGION": "us-east-1" },
      "disabled": false
    }
  }
}

Tips:

  • Do not commit secrets; use environment variables
  • Toggle servers with "disabled": true|false
  • All agents now inherit MCP tools automatically (no explicit tools: field)

License

Apache License 2.0. See LICENSE for details.