AI Development Workflow System

Mission-Driven AI workflows for Claude Code with specialized agents, token efficiency, and GitHub integration.

Stop "vibe coding." Start building with purpose. This system provides a team of 10 specialized AI agents that transform GitHub issues into tested, aligned, and production-ready code.

---

Why Use This?

Problem: AI coding sessions are chaotic. You waste tokens on unnecessary context, make unplanned changes, and lose sight of the project's goals.

Solution: A structured, mission-driven workflow where specialized agents handle each step of the development process, from planning to final alignment.

Result:

• ✅ Mission-Driven Development - Ensure every commit aligns with your project's core goals.
• 🎯 50-80% Fewer Tokens - Smart, role-based context loading for each agent.
• 🧠 Superior Code Quality - A mandatory Plan -> Test -> Align -> Commit sequence prevents mistakes.
• 🔗 Full GitHub Integration - Turn your GitHub issues into completed tasks automatically.
• ♿ ADHD/Autistic-Friendly - Structured, predictable, and organized workflows reduce cognitive load.

---

Quick Start

1. Installation

# Clone the workflow into your project's .ai-workflow directory
git clone https://github.com/aidenestelle/AI-Development-Workflow-System.git .ai-workflow

2. Define Your Mission

Before you start, you must define your project's goals. This is a one-time setup.

# In Claude Code:
/create-project-alignment

Claude will ask you a few questions to create your alignment-guide.md. This guide ensures the AI never loses sight of your vision.

3. Your First Task

First, create an issue on your project's GitHub repository. Then:

# In Claude Code:
/create

Claude will read your open GitHub issues, create tasks in its internal task manager (Taskmaster), and then you can start work with:

/start

---

The 10 Specialized Agents

Each agent has a defined role, loading only the context it needs to excel at its job.

Agent	Budget	What It Does
🎯 Orchestrator	20-30k	The project manager; directs the other agents.
🧠 Planning	40-60k	Analyzes problems and creates robust implementation plans.
🎨 Frontend	80-120k	Implements UI/UX with React, Tailwind, etc.
⚙️ Backend	80-120k	Builds APIs and server-side logic.
🔗 Full-Stack	120-150k	Handles features that span the entire stack.
🗄️ Database	50-80k	Manages schema, migrations, and database operations.
✅ Testing	80-120k	Performs technical validation: builds, lints, and tests.
mision-driven Alignment	40-60k	The gatekeeper; ensures work aligns with project goals before commit.
🔧 Refactoring	120-150k	Cleans up code and splits large files.
📝 Documentation	30-50k	Manages the in-progress CONTEXT.md file.

---

Available Commands

/create-project-alignment  # One-time setup to define your project's mission.
/create                    # Create tasks from open GitHub issues.
/start                     # Begin the next available task.
/continue                  # Resume an in-progress task.
/plan                      # Enter a deep planning session for a complex task.
/finish                    # Kick off the final Test and Align validation sequence.
/do-all                    # Automate the entire workflow for all open tasks.

---

How It Works: The Full Lifecycle

Traditional Approach 😫

You: "Fix the navigation bug"
AI: *loads entire codebase* (800k tokens)
AI: *makes a quick patch without analysis*
Result: The bug returns, new bugs are created, and the "fix" doesn't match the desired user experience.

With This System ✅

1. You: Create a "Fix navigation bug" issue on GitHub.
2. You: /create Orchestrator: "I've found the issue and created TASK-007."
3. You: /start Orchestrator: "This is a bug. Spawning Planning Agent..." Planning Agent: "The root cause is a stale closure. The plan is to refactor the hook. Do you approve?"
4. You: "Approve" Orchestrator: "Spawning Frontend Agent to implement the approved plan..."
5. You: /finish Orchestrator: "Implementation is done. Spawning Testing Agent..." Testing Agent: "Build passed, lint passed, all acceptance criteria met. The code is technically sound." Orchestrator: "Tests passed. Spawning Alignment Agent..." Alignment Agent: "I've reviewed the changes against the alignment-guide.md. The fix aligns with our principle of 'Clarity over cleverness'. This is approved."
6. Orchestrator: "Alignment passed. Committing code and closing GitHub issue."

Result: A robust, well-planned fix that is both technically correct and aligned with your project's vision.

---

What You Get

.ai-workflow/
├── .claude/
│   ├── commands/    # Workflow command definitions
│   └── agents/      # 10 specialized agents
└── docs/
    ├── AI_RULESET.md              # The AI's master constitution
    ├── AGENTS.md                  # Defines each agent's role and context
    ├── AGENT_SELECTION_GUIDE.md   # A flowchart for agent selection
    ├── alignment-guide.md         # Your Project's Mission and Goals!
    ├── PLANNING_MODE.md           # The protocol for preventing bad code
    ├── TESTING_GUIDELINES.md      # The standards for technical validation
    └── CONTEXT.md                 # A temporary file for in-progress work

---

Contributing

We welcome contributions! See guide/CONTRIBUTING.md for guidelines.

License

This project is licensed under the MIT License - see the guide/LICENSE file for details.