HomeLab Specialist Agents Directory

Version: 1.0.0 For: HL-Master reference

Quick Reference

Full UUIDs

HL-Master:  11111111-aaaa-bbbb-cccc-000000000001
HL-Network: 11111111-aaaa-bbbb-cccc-000000000002
HL-SCPI:    11111111-aaaa-bbbb-cccc-000000000003
HL-Infra:   11111111-aaaa-bbbb-cccc-000000000004
HL-Home:    11111111-aaaa-bbbb-cccc-000000000005
HL-AI:      11111111-aaaa-bbbb-cccc-000000000006
HL-Gate:    11111111-aaaa-bbbb-cccc-000000000007

HL-Network (Network Infrastructure)

UUID: 11111111-aaaa-bbbb-cccc-000000000002 Startup: /hl-network Working Dir: /home/homelab/HomeLab

Capabilities:

• UniFi management (UDM Pro @ 10.0.1.1)
• VLAN configuration
• Firewall rule management
• Client tracking
• Network topology
• Connectivity troubleshooting

MCPs: UniFi MCP (~4k tokens) Context: ~44k tokens at startup

When to Delegate:

• Network configuration changes
• UniFi device management
• Client connectivity issues
• Firewall rules
• VLAN design

HL-SCPI (Test Equipment)

UUID: 11111111-aaaa-bbbb-cccc-000000000003 Startup: /hl-scpi Working Dir: /home/homelab/HomeLab

Capabilities:

• SCPI instrument control (6 devices)
• Automated measurements
• Test data logging
• Equipment health monitoring
• Calibration tracking

MCPs: None (raw socket commands) Context: ~40k tokens at startup

When to Delegate:

• Equipment control
• Measurement tasks
• Test automation
• Equipment discovery
• Data logging

CRITICAL: Never enable outputs without user approval

HL-Infra (Infrastructure & Virtualization)

UUID: 11111111-aaaa-bbbb-cccc-000000000004 Startup: /hl-infra Working Dir: /home/homelab/HomeLab

Capabilities:

• Proxmox VM management
• Docker orchestration
• Ansible automation
• Infrastructure as Code
• Backup management

MCPs: None (SSH, APIs, CLI) Context: ~40k tokens at startup

When to Delegate:

• VM lifecycle operations
• Container deployment
• Infrastructure automation
• System updates
• Backup/restore

HL-Home (Home Automation)

UUID: 11111111-aaaa-bbbb-cccc-000000000005 Startup: /hl-home Working Dir: /home/homelab/homeassistant

Capabilities:

• Home Assistant management
• Device integration
• Automation development
• Scene management
• Energy monitoring

MCPs: Home Assistant MCP (~16k tokens) Context: ~56k tokens at startup

When to Delegate:

• Home Assistant configuration
• Device integration
• Automation scripts
• Smart home control
• HA troubleshooting

Note: Works in separate repo (unmanned-systems-uk/homeassistant)

HL-AI (AI/ML Operations)

UUID: 11111111-aaaa-bbbb-cccc-000000000006 Startup: /hl-ai Working Dir: /home/homelab/HomeLab

Capabilities:

• GPU workload management
• Model training/deployment
• Ollama LLM services
• Jetson edge deployment
• ML pipeline orchestration

MCPs: None (SSH, Docker, CUDA) Context: ~40k tokens at startup

When to Delegate:

• GPU passthrough configuration
• Model training
• LLM deployment
• Jetson operations
• ML pipeline work

HL-Gate (HomeGate Project)

UUID: 11111111-aaaa-bbbb-cccc-000000000007 Startup: /hl-gate Working Dir: /home/homelab/HomeGate

Capabilities:

• HomeGate development
• GitHub issue implementation
• Frontend/backend development
• Database schema design
• Deployment to i3 host

MCPs: None (code development) Context: ~40k tokens at startup

When to Delegate:

• HomeGate development ONLY
• HG- issue implementation
• Dashboard features
• SSH terminal work
• Deployment tasks

Note: Works in separate repo (unmanned-systems-uk/homegate)

Availability Checks

Check if Agent is Running

# Check via messaging (send ping)
curl -X POST "http://10.0.1.210:8000/api/v1/agent-messages?from_agent_id=11111111-aaaa-bbbb-cccc-000000000001" \
  -H "Content-Type: application/json" \
  -d '{
    "to_agent_id": "<AGENT_UUID>",
    "message_type": "info",
    "subject": "Ping",
    "body": "Status check",
    "priority": "low"
  }'

# Wait for response (5-10 seconds)
# If no response, agent is offline

Startup Status Check (On HL-Master Startup)

echo "## Specialist Agents"
echo "HL-Network: Available"
echo "HL-SCPI: Available"
echo "HL-Infra: Available"
echo "HL-Home: Available"
echo "HL-AI: Available"
echo "HL-Gate: Available"

(All agents are always "available" as they're on-demand. Not persistent processes.)

Task Delegation Template

# Generic task delegation
curl -X POST "http://10.0.1.210:8000/api/v1/agent-messages?from_agent_id=11111111-aaaa-bbbb-cccc-000000000001" \
  -H "Content-Type: application/json" \
  -d '{
    "to_agent_id": "<SPECIALIST_UUID>",
    "message_type": "task",
    "subject": "Task Subject (clear, < 80 chars)",
    "body": "Detailed task description.\n\nRequirements:\n- Requirement 1\n- Requirement 2\n\nUser request: Original user message",
    "priority": "normal",
    "metadata": {
      "task_id": "HL-TASK-YYYY-MM-DD-NNN",
      "user_request": "Original user message",
      "estimated_duration": "10 minutes"
    }
  }'

Agent Communication Patterns

HL-Master → Specialist

• Task assignment
• Status queries
• Session EOD notification

Specialist → HL-Master

• Task acknowledgment
• Status updates
• Task completion
• Error/failure reports

Specialist → Specialist

NOT ALLOWED. All inter-specialist communication must go through HL-Master.

Any Agent → Human

• Approval requests
• Critical alerts
• Clarification questions

Context Optimization Summary

Savings vs Monolithic: 16-20k tokens per agent (8-10% of context)

All specialists report to HL-Master. HL-Master coordinates the ecosystem.