ai-dev-pipeline-marketplace

AI Development Pipeline

A comprehensive Claude Code plugin providing a 4-tier AI agent architecture for software development workflows including code review, implementation, and git safety.

Features

• Orchestrators - Coordinate complex workflows with human checkpoints
• Specialists - Deep expertise in code review, debugging, architecture analysis
• Workers - Execute well-defined tasks with safety gates
• Git Safety - Centralized git operations preventing accidental damage
• Multi-Model Consensus - AI decisions validated across multiple models via PAL

Quick Start

Installation via Plugin Marketplace

# In Claude Code, run:
/plugin marketplace add p4ndroid/ai-dev-pipeline-architecture

Or install manually:

# Clone and install
git clone https://github.com/p4ndroid/ai-dev-pipeline-architecture.git ~/.claude-plugins/ai-dev-pipeline

# Copy example settings to your project
cp ~/.claude-plugins/ai-dev-pipeline/examples/settings.json .claude/settings.json

See INSTALLATION.md for detailed setup instructions, PAL configuration, and troubleshooting.

Prerequisites

Requirement	Purpose	Installation
PAL MCP Server	Multi-model consensus, code review	Required - see PAL documentation
gh CLI	GitHub PR operations	brew install gh
git	Version control	Pre-installed

Verify Setup

# Check GitHub CLI
gh auth status

# Check git
git --version

Agent Architecture

┌─────────────────────────────────────────────────────────────────┐
│                     TIER 1: ORCHESTRATORS                        │
│  Coordinate workflows, spawn sub-agents, human checkpoints      │
├─────────────────────────────────────────────────────────────────┤
│  pr-review-manager     │ Full PR review lifecycle               │
│  architecture-lead     │ Architecture analysis & design         │
│  implementation-manager│ Task execution coordination            │
└─────────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────┐
│                     TIER 2: SPECIALISTS                          │
│  Deep expertise in one domain, spawned by orchestrators         │
├─────────────────────────────────────────────────────────────────┤
│  code-reviewer         │ PAL-based multi-model code review      │
│  debug-analyst         │ Systematic debugging & root cause      │
│  architecture-analyst  │ Deep codebase exploration              │
│  task-decomposer       │ Break down complex tasks               │
│  implementation-planner│ Feature implementation planning        │
│  test-runner           │ Test execution and reporting           │
└─────────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────┐
│                     TIER 3: WORKERS                              │
│  Execute well-defined tasks, mandatory safety gates             │
├─────────────────────────────────────────────────────────────────┤
│  git-operator          │ ALL git operations (centralized)       │
│  doc-writer            │ ALL documentation creation             │
│  task-implementer      │ Code implementation with tests         │
└─────────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────────┐
│                     TIER 4: UTILITIES                            │
│  Simple commands, run in main session                           │
├─────────────────────────────────────────────────────────────────┤
│  /git-status           │ Quick git state view                   │
│  /list-tasks           │ List implementation tasks              │
│  /workflow-status      │ Show active workflows                  │
└─────────────────────────────────────────────────────────────────┘

Usage Examples

Code Review Workflow

User: Review PR #42

Claude: [Spawns pr-review-manager]
        → Spawns code-reviewer for multi-model analysis
        → Creates review summary document
        → Presents checkpoint for approval
        → Implements fixes if requested
        → Merges when approved

Architecture Analysis

User: Analyze the authentication system architecture

Claude: [Spawns architecture-lead]
        → Spawns multiple architecture-analysts in parallel
        → Runs AI consensus on findings
        → Presents design options
        → Creates architecture report

Implementation

User: Implement the feature from task 10