AI Agent Framework + Task Master for Claude Code

Enterprise AI assistance with built-in ISO/CMMI compliance and integrated work management

What Is This?

A clone-and-use framework providing your development teams with:

1. 9 Specialist AI Agents - Organized by department for Claude Code, Claude Desktop, and Claude.ai 🌟
2. 7 Agent Shortcuts - Slash commands with auto prompt enhancement (/as-dev-backend, /as-dev-frontend, etc.)
3. Task Master AI - Integrated work breakdown and progress tracking via MCP
4. Compliance Built-In - ISO/IEC 27001/27701/27002 + CMMI Level 3 automatic validation
5. Claude Skills - Reusable specialized capabilities (Jira, Confluence, Database, Figma, Document Parser)
6. Workflow Commands - Slash commands for Git operations (/push, /undo) and project setup (/activate)
7. MCP Integration - Task Master, Database Adapters, Atlassian, Figma, Playwright

🚀 First Time Here? Start Here!

New to this AI Agent toolkit? Follow this simple path:

1️⃣ Understand What You Have

You now have 9 specialist AI agents organized by department:

• 🛠️ Dev Team: Backend Developer, Frontend Developer, Quality Engineer
• 💼 Business: Business Analyst
• 🏗️ Infrastructure: Architecture Expert, Platform Engineer
• 📊 Data: Data Engineer
• ⚙️ Common: Initialize Agent, Prompt Engineer

2️⃣ First Action: Run /activate

After installing dependencies, simply type:

/activate

This initializes your project and guides you through setup much faster than manual invocation.

3️⃣ Then Choose Your Path

Option A: Use Task Master AI (Recommended for structured work)

npx task-master-ai
# AI breaks down your work into manageable tasks

Option B: Use Agent Shortcuts (Fastest way to get specialized help)

/as-dev-backend         # Backend Developer with auto prompt enhancement
/as-dev-frontend        # Frontend Developer with auto prompt enhancement
/as-dev-quality         # Quality Engineer with auto prompt enhancement
/as-business-analyst    # Business Analyst with auto prompt enhancement
/as-infra-architect     # Architecture Expert with auto prompt enhancement
/as-infra-platform      # Platform Engineer with auto prompt enhancement
/as-data-engineer       # Data Engineer with auto prompt enhancement

Option C: Talk to Agents Directly (Traditional way)

Act as agent-backend-developer from .claude/agents/dev/ and help me build a REST API

Option D: Use Workflow Commands (For Git operations)

/push              # Auto-review, commit, and push
/discard           # Discard uncommitted changes
/undo              # Interactive undo
/revert-last       # Revert last commit

4️⃣ Common First-Time Questions

❓ "Which agent do I use?" → See "Meet Your AI Agent Team" section below - agents are organized by department

❓ "Where are the agents located?" → .claude/agents/[department]/[agent-name].md

❓ "Do I need all 9 agents?" → No! Use only what you need. They're there when you need them.

❓ "What if I don't have API keys yet?" → The framework works without API keys, but Task Master AI requires them. See Configuration section.

Quick Start (< 5 Minutes)

The easiest way to get started:

1. Clone this repository to your project location

   git clone [framework-repo] my-project
   cd my-project

2. Install dependencies

   npm install

3. Configure your API keys (one-time setup)

   cp env.example .env
   # Edit .env and add your AI service API keys (Anthropic, OpenAI, etc.)

4. Open in Claude Code and run the activation command:

   /activate

   This will initialize your project and guide you through the setup process with faster response than manual agent invocation.

That's it! The framework provides:

• ✅ 9 specialist AI agents organized by department
• ✅ 7 agent shortcuts with auto prompt enhancement
• ✅ Task Master AI MCP for integrated task management
• ✅ Claude Skills for specialized capabilities
• ✅ 12 slash commands (setup, Git workflows, agent shortcuts)
• ✅ MCP servers for external integrations
• ✅ Compliance validation built into all agents

Why Use This Framework?

The Problem

• Teams use AI inconsistently → varied code quality
• Compliance is manual afterthought → audit failures
• Work breakdown takes hours → delayed starts
• Project context gets lost → slow ramp-up

The Solution

This Framework provides:

• ✅ Standardized AI assistance following company standards
• ✅ Automatic compliance validation (no manual checking)
• ✅ AI-powered work breakdown (minutes, not hours)
• ✅ Context preservation (never forget where you left off)

The Approach

Framework, Not Platform - Clone to your project, not deploy infrastructure

• Setup: < 30 minutes (vs hours for platforms)
• Maintenance: Pull updates (vs managing infrastructure)
• Autonomy: Your project, your control
• No CI/CD: Direct development workflow without pipeline complexity

Meet Your AI Agent Team

Agents are organized by department for better team alignment and navigation.

🛠️ Development Team (.claude/agents/dev/)

Backend Developer Agent

Use for: API design, security implementation, microservices, AI integration Specialization: ISO 27001 security controls, authentication, backend architecture Location: .claude/agents/dev/agent-backend-developer.md

Frontend Developer Agent

Use for: UI/UX, React components, accessibility, performance optimization Specialization: WCAG 2.2 AA compliance, ISO 27701 privacy controls, modern frontend Location: .claude/agents/dev/agent-frontend-developer.md

Quality Engineer Agent

Use for: Test automation, quality assurance, compliance validation, security testing Specialization: ISO/CMMI compliance, test frameworks, audit preparation Location: .claude/agents/dev/agent-quality-engineer.md

💼 Business & Analysis (.claude/agents/business/)

Business Analyst Agent

Use for: Requirements engineering, stakeholder analysis, compliance documentation Specialization: ISO/CMMI compliance, process modeling, document parsing Location: .claude/agents/business/agent-business-analyst.md

🏗️ Infrastructure & Operations (.claude/agents/infrastructure/)

Architecture Expert Agent

Use for: System design, technical leadership, distributed systems Specialization: Architecture Decision Records (ADRs), scalability, cloud patterns Location: .claude/agents/infrastructure/agent-architecture-expert.md

Platform Engineer Agent

Use for: Infrastructure as code, CI/CD pipelines, Kubernetes, cloud platforms Specialization: AWS/Azure/GCP, Oracle Database expertise, DevOps Location: .claude/agents/infrastructure/agent-platform-engineer.md

📊 Data Engineering (.claude/agents/data/)

Data Engineer Agent

Use for: Data pipelines, ETL processes, data warehouse, analytics Specialization: Data modeling, big data processing, streaming, data quality Location: .claude/agents/data/agent-data-engineer.md

⚙️ Common Utilities (.claude/agents/common/)

Initialize Agent

Use for: Project setup, environment configuration, framework initialization Specialization: Bootstrapping, scaffolding, dependency management Location: .claude/agents/common/agent-initalize.md

Prompt Engineer Agent

Use for: Prompt optimization, LLM configuration, intelligent agent routing Specialization: ISO/IEC 22365-3 v7 Framework, multi-provider AI management Location: .claude/agents/common/agent-prompt-engineer.md

💬 Slash Commands Reference

12 slash commands available - organized by category:

🚀 Project Setup

• /activate - Initialize project and setup environment

  /activate
  # Runs agent-initialize with automatic prompt enhancement
  # Faster than manual "Act as agent-initalize..." invocation

🔧 Git Workflow Commands

• /push - Auto-review, commit, and push changes

  /push
  # Reviews code, creates commit message, pushes to remote

• /undo - Interactive undo with smart detection

  /undo
  # Intelligently detects what to undo (uncommitted changes, last commit, etc.)

• /discard - Discard uncommitted changes

  /discard
  # Discards all uncommitted changes (git reset --hard)

• /revert-last - Revert the last commit

  /revert-last
  # Reverts the most recent commit

🤖 Agent Shortcuts (Auto Prompt Enhancement)

These shortcuts automatically enhance your prompt and invoke the specialist agent:

• /as-dev-backend - Backend Developer Agent

  /as-dev-backend Build a REST API for user authentication with JWT
  # Automatically enhances prompt with security requirements, ISO 27001 controls, etc.

• /as-dev-frontend - Frontend Developer Agent

  /as-dev-frontend Create a responsive dashboard with dark mode
  # Auto-adds WCAG 2.2 AA accessibility, ISO 27701 privacy requirements, etc.

• /as-dev-quality - Quality Engineer Agent

  /as-dev-quality Write tests for the login API endpoint
  # Auto-adds security testing, coverage requirements, compliance validation, etc.

• /as-business-analyst - Business Analyst Agent

  /as-business-analyst Decompose this feature into user stories
  # Auto-adds acceptance criteria format, ISO/CMMI requirements, etc.

• /as-infra-architect - Architecture Expert Agent

  /as-infra-architect Design a microservices architecture for our SaaS platform
  # Auto-adds scalability, ADRs, compliance requirements, etc.

• /as-infra-platform - Platform Engineer Agent

  /as-infra-platform Set up a Kubernetes cluster on AWS EKS
  # Auto-adds HA/DR, security hardening, cost optimization, etc.

• /as-data-engineer - Data Engineer Agent

  /as-data-engineer Build a real-time data pipeline for user analytics
  # Auto-adds data quality, governance, performance requirements, etc.

💡 Pro Tip: Agent shortcuts provide automatic prompt enhancement based on ISO/IEC 22365-3 v7 Framework, ensuring comprehensive requirements and deliverables.

Daily Workflow Example

# Morning: Check what to work on
npx task-master-ai
# Select: "What should I work on next?"
# Shows: Task 5.2 - Implement OAuth2 with ISO 27001 A.9 requirements

# Get AI help with implementation using agent shortcut (in Claude Code)
/as-dev-backend Implement OAuth2 authentication with refresh tokens
# Agent automatically enhances prompt with:
# - ISO 27001 A.9 security controls
# - OWASP authentication best practices
# - Token storage and rotation requirements
# - Provides production-ready implementation

# Code the implementation...

# Use Task Master AI MCP to update progress
# (If MCP is connected, Task Master tools are available directly in Claude Code)

# Get code review using agent shortcut
/as-dev-quality Review this OAuth2 implementation for security and compliance
# Agent automatically validates:
# - Security vulnerabilities (SQL injection, XSS, CSRF)
# - ISO 27001 compliance
# - Test coverage requirements
# - Performance benchmarks

# Use slash command for Git operations
/push
# Automatically reviews code, creates commit message, pushes to remote

# Compliance validation is built into all agent responses

What Teams Get

Core Framework

your-project/
├── .claude/                    # Claude Code agents and configuration
│   ├── agents/                 # 9 AI agents organized by department
│   ├── commands/               # Slash commands (init, git/push, etc.)
│   ├── skills/                 # Reusable capabilities
│   └── .mcp.json              # MCP server configuration
├── .taskmaster/                # Task Master AI integration
│   ├── config.json            # AI provider settings
│   ├── state.json             # Current task state
│   ├── tasks/                 # Task definitions
│   └── templates/             # Task templates
├── memory-bank/                # Context preservation system
└── package.json               # Dependencies (task-master-ai, parsers)

MCP Servers & Integrations

your-project/
├── .mcp/
│   ├── database-adapters/      # PostgreSQL, MySQL, SQLite MCP servers
│   ├── atlassian/              # Jira/Confluence integration
│   ├── figma/                  # Figma design integration
│   └── playwright/             # Browser automation

Documentation

your-project/
├── docs/
│   ├── compliances/            # ISO/CMMI/PCI-DSS standards
│   ├── epics/                  # Epic-level requirements
│   ├── user-stories/           # User stories
│   └── tasks/                  # Technical documentation

Key Features

🤖 AI Agents with Standards Built-In

• 9 specialist agents organized by department (Dev, Business, Infrastructure, Data, Common)
• Each agent has ISO/IEC 27001/27701/27002 compliance embedded
• CMMI Level 3 practices built into responses
• Role-specific expertise and best practices
• Designed for Claude Code, Claude Desktop, and Claude.ai

📋 Task Master AI via MCP

• AI-powered task breakdown and expansion
• Complexity analysis and subtask generation
• Progress tracking with AI-assisted updates
• MCP integration for seamless Claude Code experience
• Multi-provider support (Anthropic, OpenAI, Google, etc.)

✅ Compliance Automation

• ISO 27001: Information security management
• ISO 27701: Privacy information management
• ISO 27002: Security controls implementation
• CMMI Level 3: Defined process maturity
• Compliance validation built into all agent responses

🎯 Claude Skills & Commands

• Skills: Reusable capabilities (Jira, Confluence, Database, Figma, Document Parser, Context7)
• Workflow Commands: Git operations (/push, /undo, /discard, /revert-last) and project setup (/activate)
• Agent Shortcuts: Quick access to specialists (/as-dev-backend, /as-dev-frontend, /as-dev-quality, /as-business-analyst, /as-infra-architect, /as-infra-platform, /as-data-engineer)

🔧 MCP Server Integrations

• Task Master AI: Task management directly in Claude Code
• Database Adapters: PostgreSQL, MySQL, SQLite introspection
• Atlassian: Jira and Confluence integration
• Figma: Design file access and automation
• Playwright: Browser automation and testing

Setup Requirements

Prerequisites

• Node.js 18+ (for Task Master AI and dependencies)
• Claude Code (for AI agents and MCP integration)
• Git (for version control)
• AI Service API Keys (Anthropic recommended, or OpenAI, Google, etc.)

Optional (for MCP servers)

• Docker (for Atlassian MCP server)
• Database drivers (if using database adapter MCP servers)
• Python 3.8+ (for document parsing scripts)

Configuration

Environment Variables

# .env file (not committed to Git)

# Required: AI Service API Keys
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=...

# Optional: Task Master Settings
TASKMASTER_LOG_LEVEL=info

# Optional: Compliance Validation
COMPLIANCE_THRESHOLD=70
COMPLIANCE_AUTO_RUN=true

Task Master AI

# Interactive CLI
npx task-master-ai

# Or configure providers in .taskmaster/config.json

Configuration file (.taskmaster/config.json) includes:

• Models: Main, Research, and Fallback configurations
• Providers: Anthropic (recommended), OpenAI, Azure, Ollama, etc.
• Settings: Log level, default tasks/subtasks, project name

MCP Server Configuration

Edit .claude/.mcp.json to enable/disable MCP servers:

{
  "mcpServers": {
    "task-master-ai": {
      "command": "npx",
      "args": ["-y", "task-master-ai"],
      "env": {
        "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}"
      }
    }
  }
}

Available MCP Servers:

• task-master-ai - Task management (enabled by default)
• database-adapter - Database introspection (disabled by default)
• atlassian - Jira/Confluence (disabled by default, requires Docker)
• human-mcp - Human interaction (disabled by default)
• serena - Code understanding (disabled by default)

Using AI Agents

In Claude Code (Recommended - Agent Shortcuts):

/as-dev-backend Implement authentication with JWT and refresh tokens

These shortcuts automatically enhance your prompt with best practices and compliance requirements.

In Claude Code (Traditional - Direct Invocation):

Act as agent-backend-developer from .claude/agents/dev/ and help me implement authentication

Use this when you need more control over the initial prompt.

In Claude Desktop: Reference agents from your project context - they're automatically available

In Claude.ai: Copy-paste agent content from .claude/agents/[department]/[agent-name].md

📚 See .claude/README.md for detailed agent documentation 📚 See "Slash Commands Reference" section above for all 12 available commands

Documentation

• docs/compliances/ - ISO/CMMI/NF compliance standards (ISO-27001, ISO-27701, ISO-27002, ISO-9001, CMMI, PCI-DSS, 1NF/2NF/3NF)
• docs/epics/ - Epic-level requirements and specifications
• docs/user-stories/ - User stories and acceptance criteria
• docs/tasks/ - Technical task documentation and implementation details

Project Structure

ai-integration-workflow-standardize/
│
├── README.md                       # You are here
├── CLAUDE.md                       # Project instructions for Claude Code
├── package.json                    # Dependencies (task-master-ai v0.27.3, document parsers)
├── env.example                     # API key templates
│
├── .claude/                        # Claude Code Configuration
│   ├── README.md                   # Claude-specific documentation
│   ├── .mcp.json                   # MCP server configuration
│   ├── agents/                     # 9 AI Agents organized by department
│   │   ├── dev/                    # Development (3 agents)
│   │   │   ├── agent-backend-developer.md
│   │   │   ├── agent-frontend-developer.md
│   │   │   └── agent-quality-engineer.md
│   │   ├── business/               # Business Analysis (1 agent)
│   │   │   └── agent-business-analyst.md
│   │   ├── infrastructure/         # Platform & Ops (2 agents)
│   │   │   ├── agent-architecture-expert.md
│   │   │   └── agent-platform-engineer.md
│   │   ├── data/                   # Data Engineering (1 agent)
│   │   │   └── agent-data-engineer.md
│   │   └── common/                 # Common Utilities (2 agents)
│   │       ├── agent-initalize.md
│   │       └── agent-prompt-engineer.md
│   ├── commands/                   # Slash commands (12 total)
│   │   ├── activate.md              # Project initialization
│   │   ├── push.md                 # Git push workflow
│   │   ├── discard.md              # Discard changes
│   │   ├── undo.md                 # Interactive undo
│   │   ├── revert-last.md          # Revert last commit
│   │   ├── as-dev-backend.md       # Backend Developer agent shortcut
│   │   ├── as-dev-frontend.md      # Frontend Developer agent shortcut
│   │   ├── as-dev-quality.md       # Quality Engineer agent shortcut
│   │   ├── as-business-analyst.md  # Business Analyst agent shortcut
│   │   ├── as-infra-architect.md   # Architecture Expert agent shortcut
│   │   ├── as-infra-platform.md    # Platform Engineer agent shortcut
│   │   └── as-data-engineer.md     # Data Engineer agent shortcut
│   ├── skills/                     # Reusable capabilities
│   │   ├── jira-integration.skill
│   │   ├── confluence-integration.skill
│   │   ├── database-pool.skill
│   │   ├── figma-integration.skill
│   │   ├── document-parser.skill
│   │   └── context7-integration.skill
│   └── scripts/                    # Validation scripts
│       └── validate-bash.sh
│
├── .taskmaster/                    # Task Management
│   ├── config.json                 # Task Master AI settings
│   ├── state.json                  # Current state
│   ├── tasks/                      # Task definitions
│   └── templates/                  # Task templates
│
├── memory-bank/                    # Context Preservation
│   └── .gitkeep                    # (currently empty, ready for use)
│
├── .mcp/                           # MCP Servers
│   ├── database-adapters/          # PostgreSQL/MySQL/SQLite MCP servers
│   ├── atlassian/                  # Jira/Confluence integration
│   ├── figma/                      # Figma design integration
│   └── playwright/                 # Browser automation
│
├── scripts/                        # Automation Scripts
│   ├── document-parser.py          # Multi-format document parsing
│   ├── parse-single-pdf.py         # PDF-specific parsing
│   └── requirements-document-parser.txt
│
├── docs/                           # Documentation
│   ├── compliances/                # ISO/CMMI/PCI-DSS standards
│   ├── epics/                      # Epic-level requirements
│   ├── tasks/                      # Technical task documentation
│   └── user-stories/               # User stories and acceptance criteria
└──

Development Approach

This framework follows iterative improvement cycles:

• Core Framework: Production-ready AI agents and Task Master integration
• MCP Integrations: Expanding connectivity (Database, Atlassian, Figma, Playwright)
• Claude Skills: Building reusable capabilities for common workflows
• Documentation: Continuous updates based on team feedback

Contributing

This is an internal company framework. Improvements are welcome:

1. Fork the framework repository
2. Make your changes (agent enhancements, documentation, examples)
3. Test with your team
4. Submit for review with business case

Support

• Documentation: Check docs/ directory first
• Issues: Create issue in framework repository
• Questions: [Internal support channel]
• Compliance Standards: See docs/compliances/ for ISO/CMMI documentation

Success Metrics

Framework Adoption

• ✅ Setup Time: < 30 minutes (Target achieved)
• 🎯 Team Adoption: 80% within 3 months (In progress)
• 🎯 Daily Usage: 90% developers (Pilot testing)

Quality Impact

• ✅ Compliance: 100% validation pass rate
• 🎯 Code Quality: 40% fewer standards violations
• 🎯 Bugs: 30% reduction in security/compliance bugs

Productivity Gains

• 🎯 Planning: 50% faster task breakdown
• 🎯 Context Switching: 60% faster ramp-up
• 🎯 Documentation: 40% less documentation time

Frequently Asked Questions

Q: Is this a platform I need to deploy? A: No! It's a framework you clone to your project. No infrastructure or CI/CD needed.

Q: Which AI platform do I need? A: Claude Code (recommended), Claude Desktop, or Claude.ai. The framework is optimized for Claude's context window and capabilities.

Q: Do I need all 9 AI agents? A: All agents are available. Use the ones relevant to your work. They're organized by department (Dev, Business, Infrastructure, Data, Common) for easy navigation.

Q: What if I don't have Anthropic API keys? A: Task Master supports multiple providers (Anthropic, OpenAI, Azure, Google, etc.). Configure the providers you have access to in .taskmaster/config.json.

Q: How do I enable MCP servers? A: Edit .claude/.mcp.json to enable/disable servers. Task Master is enabled by default. Others (Database, Atlassian, Figma) are disabled by default.

Q: What's the difference between agent shortcuts and direct agent invocation? A: Agent shortcuts (e.g., /as-dev-backend) provide automatic prompt enhancement and faster response. Direct invocation (Act as agent-backend-developer from .claude/agents/dev/) gives you more control over the initial prompt. Shortcuts are recommended for most use cases.

Q: What are Claude Skills? A: Reusable capabilities in .claude/skills/ that agents can use. Includes Jira, Confluence, Database, Figma, Document Parser, and Context7 integrations.

Q: Can I use this on existing projects? A: Yes! Clone the framework into your project root, configure API keys, and start using agents immediately.

Q: How do updates work? A: Pull updates from the framework repo and merge into your project (like any dependency).

Version History

• v3.0.0 (November 2025) - Migrated to Claude Code, department-based agent organization, MCP integration
• v2.0.0 (October 2025) - Framework approach with Cursor IDE support
• v1.x (Pre-October 2025) - Enterprise platform (deprecated)

License & Compliance

This framework maintains enterprise-grade compliance standards:

• ✅ ISO/IEC 27001:2022 (Information Security Management)
• ✅ ISO/IEC 27701:2019 (Privacy Information Management)
• ✅ ISO/IEC 27002:2022 (Security Controls)
• ✅ CMMI Level 3 (Defined Process Maturity)

Compliance validation runs automatically. See docs/compliances/ for standards documentation.

Quick Links

• 🤖 AI Agents - 9 specialist agents organized by department
• 📚 Claude Documentation - Detailed agent guide and usage
• 🔧 MCP Configuration - MCP server settings
• 💬 Slash Commands - 12 commands total:
  • Setup: /activate
  • Git Workflows: /push, /undo, /discard, /revert-last
  • Agent Shortcuts: /as-dev-backend, /as-dev-frontend, /as-dev-quality, /as-business-analyst, /as-infra-architect, /as-infra-platform, /as-data-engineer
• 🎯 Claude Skills - Reusable capabilities
• 📋 Task Management - Task Master AI configuration
• ✅ Compliance Standards - ISO/IEC 27001/27701/27002, CMMI, PCI-DSS
• 📖 Project Instructions - Instructions for Claude Code

Project Status: ✅ Production Ready | 🚀 Claude Code Optimized Last Updated: November 11, 2025 Platform: Claude Code, Claude Desktop, Claude.ai Framework Version: 3.0.0