Claude Code: Enterprise AI Development Fleet

Build autonomous AI development teams with specialized agents, skills, and workflows.

Inspired by:

• AgentField - "Kubernetes for AI Agents" architecture
• Shotgun - Spec-driven development (5-phase pipeline)
• Claude Code Skill Factory - Factory patterns for generating skills/agents

Quick Start

# Basic installation (agents only)
npx @kartick.ns/claude-agents init

# Full installation (agents + skills + hooks + rules + memory)
npx @kartick.ns/claude-agents init --full

# Enterprise installation (everything + enterprise settings)
npx @kartick.ns/claude-agents init --enterprise

What's Included

Agents (12 Specialized Subagents)

| Agent | Model | Purpose | |-------|-------|---------| | code-reviewer | sonnet | Proactive code quality review | | security-auditor | opus | Vulnerability detection (proactive) | | debugger | opus | Bug investigation and fixes | | test-engineer | sonnet | Test creation and coverage | | architect | opus | System design and planning | | feature-builder | sonnet | Feature implementation | | api-developer | sonnet | REST/GraphQL API development | | refactorer | sonnet | Code improvement | | performance-analyst | sonnet | Performance optimization | | codebase-explorer | haiku | Code understanding | | doc-writer | sonnet | Technical documentation | | git-operator | haiku | Version control operations |

Skills (16 Workflow Commands)

| Skill | Purpose | |-------|---------| | /specify | Create codebase-aware specs (5-phase: Research→Specify→Plan→Tasks→Export) | | /plan | Design implementation plans | | /implement | Execute plans task by task | | /validate | Verify against specifications | | /review | Comprehensive code review | | /debug | Systematic debugging | | /orchestrate | Multi-agent coordination | | /ralph-loop | Autonomous iteration pattern | | /tdd | Test-driven development | | /explore | Deep codebase exploration | | /manual-qa | Manual QA testing (Playwright MCP for web, Mobile MCP for mobile) | | /automation-qa | E2E automation testing (Playwright/Detox/Maestro) | | /skill-factory | Generate new skills through guided Q&A | | /agent-factory | Generate new agents through guided Q&A | | /hook-factory | Generate new hooks through guided Q&A | | /export | Export specs to AGENTS.md, Cursor rules, or other formats |

Hooks (9 Automation Scripts)

| Hook | Event | Purpose | |------|-------|---------| | detect-secrets | PreToolUse | Block credential exposure | | validate-command | PreToolUse | Prevent dangerous commands | | validate-file-path | PreToolUse | Block path traversal | | auto-format | PostToolUse | Auto-format after edits | | run-lint | PostToolUse | Run linter on changes | | update-progress | Stop | Track progress | | log-agent-completion | SubagentStop | Log agent work | | load-env | SessionStart | Load environment | | save-session-summary | SessionEnd | Create session summary |

Rules (5 Path-Specific Standards)

| Rule | Applies To | |------|------------| | api | src/api/**/*, src/routes/**/* | | tests | **/*.test.ts, test/**/* | | components | src/components/**/* | | database | src/db/**/*, prisma/**/* | | security | src/auth/**/*, **/*password* |

Memory Bank

Persistent context across sessions:

• progress.md - Task tracking
• decisions.md - Architectural decisions
• patterns.md - Code patterns
• issues.md - Known issues
• glossary.md - Project terminology

Installation Options

# Install everything
npx @kartick.ns/claude-agents init --enterprise

# Install specific components
npx @kartick.ns/claude-agents init --with-skills --with-hooks

# Install specific team
npx @kartick.ns/claude-agents team full-stack

# Add specific agents
npx @kartick.ns/claude-agents add code-reviewer security-auditor

CLI Commands

| Command | Description | |---------|-------------| | init | Initialize agents (+ optional components) | | add <agents...> | Add specific agents | | team <name> | Install pre-configured team | | list | List available agents/teams | | update | Update to latest version | | validate | Validate installed files |

Init Options

| Option | Description | |--------|-------------| | --full | Install agents, skills, hooks, rules, memory | | --enterprise | Full + enterprise settings and MCP | | --with-skills | Install workflow skills | | --with-hooks | Install hook scripts | | --with-rules | Install path-specific rules | | --with-memory | Install memory bank templates | | --with-mcp | Install MCP configuration | | --with-settings | Install settings.json |

Pre-Built Teams

Essential (3 agents)

For solo developers and quick prototypes.

npx @kartick.ns/claude-agents team essential

Full-Stack (7 agents)

For web application development.

npx @kartick.ns/claude-agents team full-stack

Security (6 agents)

For security-sensitive applications.

npx @kartick.ns/claude-agents team security

Enterprise (12 agents)

Complete coverage for enterprise development.

npx @kartick.ns/claude-agents team enterprise

Workflow: Spec-Driven Development

/specify → /plan → /implement → /validate

1. Specify: Create requirements specification
2. Plan: Design implementation approach
3. Implement: Execute the plan
4. Validate: Verify against acceptance criteria

Workflow: Ralph Loop (Autonomous Iteration)

For complex tasks requiring multiple iterations:

# Start the loop
/ralph-loop "Migrate all API endpoints to Hono"

Progress tracked in git history and .claude/memory/ralph-progress.md.

Project Structure After Installation

your-project/
├── .claude/
│   ├── agents/           # Agent definitions
│   │   ├── code-reviewer.md
│   │   ├── debugger.md
│   │   └── ...
│   ├── skills/           # Workflow skills
│   │   ├── specify/SKILL.md
│   │   ├── plan/SKILL.md
│   │   └── ...
│   ├── rules/            # Path-specific rules
│   │   ├── api.md
│   │   ├── tests.md
│   │   └── ...
│   ├── memory/           # Persistent context
│   │   ├── progress.md
│   │   ├── decisions.md
│   │   └── ...
│   └── settings.json     # Hooks and permissions
├── hooks/                # Hook scripts
│   ├── auto-format.sh
│   ├── detect-secrets.sh
│   └── ...
└── .mcp.json             # MCP server configuration

Configuration

Settings (.claude/settings.json)

{
  "permissions": {
    "allow": ["Bash(npm:*)", "Read(*)", "Glob(*)"],
    "deny": ["Bash(rm -rf:*)"]
  },
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Write|Edit",
        "hooks": [
          {"type": "command", "command": "$CLAUDE_PROJECT_DIR/hooks/detect-secrets.sh"}
        ]
      }
    ]
  }
}

MCP Servers (.mcp.json)

{
  "mcpServers": {
    "github": {
      "type": "http",
      "url": "https://api.githubcopilot.com/mcp/",
      "headers": {"Authorization": "Bearer ${GITHUB_TOKEN}"}
    },
    "postgres": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres"],
      "env": {"POSTGRES_URL": "${DATABASE_URL}"}
    }
  }
}

Agent File Format

---
name: agent-name
description: |
  When to use this agent. Claude reads this for delegation.
tools: Read, Grep, Glob, Bash
model: sonnet
---

# System prompt content

## When Invoked
1. Step 1
2. Step 2

## Output Format
...

Skill File Format

---
name: skill-name
description: What this skill does
allowed-tools: Read, Grep, Glob
context: fork
model: sonnet
---

# Instructions for the skill

Resources

• Claude Code Documentation
• Multi-Agent Orchestration Guide

License

MIT