Multiple ways to use AI Agent Workflow in your projects.
| Method | Best For | Command |
|---|---|---|
| One-liner install | Personal use, multiple projects | curl ... | bash |
| Git submodule | Team projects, version control | git submodule add ... |
| Direct clone | Contributing, customizing | git clone ... |
| Copy agents only | Minimal footprint | cp -r agents/ ... |
Install globally and use across all your projects:
curl -fsSL https://raw.githubusercontent.com/yourusername/ai-agent-workflow/main/scripts/install.sh | bashThis:
- Clones to
~/.ai-agents/ - Adds
agent-initcommand to your PATH - Sets up
AI_AGENTS_HOMEenvironment variable
After installation:
# Restart terminal, then:
agent-init my-new-project
cd my-new-project && claudeAdd to existing project:
cd my-existing-project
add-to-project.shBest for team projects where you want version-controlled agent prompts:
cd your-project
# Add as submodule
git submodule add https://github.com/yourusername/ai-agent-workflow.git .ai-agents
# Create symlink to agents
ln -s .ai-agents/agents ./agents
# Copy and customize CLAUDE.md
cp .ai-agents/templates/CLAUDE.md.template ./CLAUDE.md
# Create artifacts directory
mkdir -p artifacts
# Commit
git add .gitmodules .ai-agents agents CLAUDE.md artifacts
git commit -m "Add AI Agent Workflow"For teammates cloning the repo:
git clone --recursive your-repo-url
# Or if already cloned:
git submodule update --initUpdate to latest version:
cd .ai-agents && git pull origin main && cd ..
git add .ai-agents && git commit -m "Update AI Agent Workflow"Best for contributing or heavy customization:
# Clone anywhere you like
git clone https://github.com/yourusername/ai-agent-workflow.git ~/ai-agent-workflow
# Use the init script
~/ai-agent-workflow/scripts/init-project.sh my-project
# Or add to existing project
~/ai-agent-workflow/scripts/add-to-project.sh /path/to/projectMinimal approach - just copy what you need:
cd your-project
# Download just the agents folder
curl -L https://github.com/yourusername/ai-agent-workflow/archive/main.tar.gz | \
tar -xz --strip-components=1 ai-agent-workflow-main/agents
# Create minimal CLAUDE.md
cat > CLAUDE.md << 'EOF'
# Project: My Project
## Orchestrator-Driven Mode
Agent 0 drives development. See agents/agent-0-orchestrator.md for details.
## Project Context
- **Name**: My Project
- **One-liner**: [Description]
## Current State
Phase: Discovery
EOF
mkdir -p artifactsAdd to your .devcontainer/devcontainer.json:
{
"postCreateCommand": "curl -fsSL https://raw.githubusercontent.com/yourusername/ai-agent-workflow/main/scripts/install.sh | bash"
}However you install, your project should look like:
your-project/
├── CLAUDE.md # Project config (required)
├── agents/ # Symlink or copy (required)
├── artifacts/ # Generated docs (created as needed)
│ ├── problem-brief-v0.1.md
│ ├── prd-v0.1.md
│ └── ...
└── src/ # Your code
| Variable | Purpose | Default |
|---|---|---|
AI_AGENTS_HOME |
Global installation path | ~/.ai-agents |
Add to your ~/.zshrc or ~/.bashrc:
# Quick project initialization
alias agent-init='$AI_AGENTS_HOME/scripts/init-project.sh'
alias agent-add='$AI_AGENTS_HOME/scripts/add-to-project.sh'
# Quick access to agent docs
alias agents='ls $AI_AGENTS_HOME/agents/*.md'
alias agent='cat $AI_AGENTS_HOME/agents/agent-$1.md'Global install:
cd $AI_AGENTS_HOME && git pullSubmodule:
git submodule update --remote .ai-agentsDirect copy: Re-download the agents folder
Global install:
rm -rf ~/.ai-agents
# Remove lines from ~/.zshrc or ~/.bashrcSubmodule:
git submodule deinit .ai-agents
git rm .ai-agents
rm -rf .git/modules/.ai-agentsThe target moved. Recreate it:
rm agents
ln -s /path/to/ai-agent-workflow/agents ./agentsMake sure you're running claude from the directory containing CLAUDE.md.
chmod +x $AI_AGENTS_HOME/scripts/*.shAfter installation:
- Edit
CLAUDE.mdwith your project details - Run
claudein your project directory - Tell the orchestrator what you want to build!
See docs/CLAUDE_CODE_GUIDE.md for the full usage guide.