Workflow-Driven Agentic Software Factory

The Dark Factory Engine converts human-readable workflow definitions (Markdown) into executable OpenClaw skills that can be deployed as Coolify apps.

Core Concept: workflow.md → skill/ → Coolify deployment → Automated execution

┌─────────────────┐
│ Workflow.md     │ (Human-readable process definition)
│ (Templates)     │
└────────┬────────┘
         │ Compile
         ↓
┌─────────────────┐
│ OpenClaw Skill  │ (Executable code + metadata)
│ (Generated)     │
└────────┬────────┘
         │ Deploy
         ↓
┌─────────────────┐
│ Coolify App     │ (Running automation)
│ (Auto-scaled)   │
└─────────────────┘

# 1. Compile a workflow template
./bin/compile workflow-templates/deploy-service.md

# 2. Dry-run — see what would happen
./bin/deploy output/deploy-service-skill/ --dry-run

# 3. Deploy to Coolify (git push + app create + deploy)
./bin/deploy output/deploy-service-skill/

# 4. Monitor deployed skills
./bin/monitor                          # list all df-* apps
./bin/monitor deploy-service           # specific skill status
./bin/monitor deploy-service --logs    # tail logs
./bin/monitor --registry               # local deployment registry

Workflows are written in Markdown with structured sections:

# Workflow: Deploy Service

## Description
Deploys a service to Coolify with health checks and rollback capability.

## Inputs
- `service_name`: Name of the service to deploy
- `git_repo`: Repository URL
- `branch`: Git branch (default: main)

## Steps
1. **Clone Repository**
   - Checkout specified branch
   - Validate dockerfile exists

2. **Create Coolify App**
   - Generate app configuration
   - Set environment variables
   - Configure health checks

3. **Deploy & Verify**
   - Trigger deployment
   - Wait for healthy status
   - Run smoke tests

## Outputs
- `deployment_url`: Live service URL
- `health_status`: Service health check result
- `rollback_command`: Command to rollback if needed

## Guardrails
- Max deployment time: 10 minutes
- Auto-rollback on failure
- Requires manual approval for production

dark-factory-engine/
├── README.md                    # This file
├── bin/                        # Executable scripts
│   ├── compile                 # Workflow → Skill compiler
│   ├── deploy                  # Skill → Coolify deployer  
│   └── monitor                 # Execution monitor
├── workflow-templates/         # Sample workflows
│   ├── deploy-service.md
│   ├── run-tests.md
│   ├── backup-data.md
│   └── security-scan.md
├── compiler/                   # Compiler implementation
│   ├── parser.py              # Markdown parser
│   ├── generator.py           # Skill generator
│   └── templates/             # Code templates
├── examples/                   # Example generated skills
└── docs/                      # Documentation

• Workflow template parsing
• Basic skill generation
• Coolify integration (bin/deploy — git push + app create + deploy)
• Deployment monitor (bin/monitor — live status + logs)
• Execution engine (skill triggering via HTTP or cron)

• Meta-agent monitoring
• Multi-workflow orchestration
• Visual workflow builder
• Compliance evidence generation

1. Install Dependencies

   pip install -r requirements.txt

2. Test Compilation

   python compiler/compile.py workflow-templates/deploy-service.md

3. Deploy First Workflow

   # TBD - Coolify integration

This engine integrates with our existing infrastructure:

• OpenClaw: Generated skills run as OpenClaw agents
• Coolify: Auto-deployment and scaling
• Memory System: Execution logs → Qdrant vector memory
• Agent Spawning: Multi-workflow parallel execution

Part of the MachineMachine Dark Software Factory initiative