forge-deploy

name	forge-deploy
version	2.0
type	autoresearch
description	Autonomous infrastructure deployment and MCP server management
user-invocable	true
agent	FORGE
agent_model	deepseek/deepseek-r1
mcps	["henry-core","henry","github","filesystem"]
gstack_skills	["/ship","/plan-eng-review","/qa","/debug"]
eval_metric	deploy_success_rate
eval_budget	120s
guard	test_pass_rate = 1.0

/forge-deploy — Infrastructure Deployment

What it does

Manages the HENRY OS infrastructure stack: MCP servers, Docker services, CLI tools. Uses henry-core for specialist skills, GitHub for version control, and gstack's engineering pipeline for quality gates.

Protocol

Step 0: Load context

• memory/CTO_MEMORY.md — infrastructure state, pending deployments
• henry-os/INSTALL.md — current installation status
• HENRY_BRAIN.md — engineering priorities

Step 1: Assess current state

Check infrastructure health:

# henry-core status
node C:\Users\whitt\Development\henry-ai-company\henry-os\henry-core\src\index.js --health

# Docker services
docker ps --format "table {{.Names}}\t{{.Status}}"

# MCP server connectivity
claude mcp list

Step 2: Engineering review (/plan-eng-review)

For any new deployment or change:

1. Run /plan-eng-review to lock architecture, data flow, edge cases
2. Verify Windows-native first (no WSL dependency for core services)
3. Confirm Docker for background services only
4. Check MCP integration points

Step 3: Build and test

1. Implement changes (Node.js for servers, Python for AI/ML)
2. Run /qa for integration testing
3. Run /debug if any failures
4. All tests must pass (guard: test_pass_rate = 1.0)

Step 4: Ship (/ship)

Run /ship pipeline:

1. Run tests
2. PR review
3. Push
4. Open PR on GitHub

Step 5: Deploy

# Windows services
cd C:\Users\whitt\Development\henry-ai-company\henry-os
powershell -File start.ps1

# Docker services
cd henry-os/openclaw && docker-compose up -d

Step 6: Verify

• Restart Claude Desktop if MCP config changed
• Test each MCP endpoint
• Update memory/CTO_MEMORY.md with deployment result

Infrastructure Map

henry-os/
├── henry-core/   Node.js — 20 AI skills via Opus 4.6 [DEPLOYED]
├── henry-phone/  Node.js — Twilio voice/SMS [NEEDS TWILIO NUMBER]
├── openclaw/     Docker — AI runtime [NEEDS docker-compose up]
└── voice/        HTML — Browser voice UI [DEPLOYED]

Autoresearch Loop

• Metric: deploy_success_rate — % of deployments that succeed without rollback
• Budget: 120s eval
• Modify: deployment scripts, health checks, error handling
• Guard: test_pass_rate must be 1.0 (all tests pass before deploy)

MCP Integration Map

FORGE (DeepSeek R1)
  ├── henry-core: 20 specialist AI skills
  ├── henry: system access (PowerShell, processes)
  ├── github: repo management, PRs
  ├── filesystem: read/write deployment files
  ├── /ship: test → review → push → PR pipeline
  ├── /plan-eng-review: architecture validation
  ├── /qa: integration testing
  └── /debug: root-cause debugging