What Happens in This Section

You'll register each agent with OpenClaw and give them their first command. By the end, all 6 agents will be live — reporting to their Discord channels and ready to receive instructions from you.

Two ways to talk to your agents:

Method	Best for	How it works
OpenGoat UI	Beginners — visual interface	Browser-based dashboard, click to select agent, type your message
VS Code terminal	Recommended — faster, more control	openclaw agent --agent [id] "your message"

Both methods work. Start with whichever feels comfortable.

Recommended approach — hybrid with AI: When you're just starting out, keep your chat AI open alongside your agents. If a command doesn't work or you're not sure what to type, just describe the situation to the AI and ask for help. This is exactly how most people — including the BALIA OS team — ran things in the beginning. You don't need to master the terminal on day one. The AI chat window is always there as your co-pilot.

---

How This Section Works

Paste the prompt below into your chat AI. The AI will guide you through registering all 6 agents and sending your first command to each one. Attach your Team Design Document so the AI knows your agent IDs and workspace paths.

If anything goes wrong, describe the error or take a screenshot and share it — the AI will help you fix it.

---

▶ Paste this into your AI to get started

Attach your Team Design Document before sending.

You are helping me register and launch my BALIA OS multi-agent team
on OpenClaw.

I have attached my Team Design Document. Read it to get my agent IDs
and names before starting.

Before we begin, ask me:
"What is your Windows username? (The name in C:\Users\[username])"
Use that username in all file paths throughout this session.
All my agent files are under $env:USERPROFILE/.opengoat/

Guide me step by step. One step at a time. If I hit an error,
help me fix it before moving on. If I get stuck, ask me to share
a screenshot or paste the error message.

Here is what we need to do:

STEP 1 — Start the gateway
Tell me: "We'll start the OpenClaw gateway with a command that filters
out noisy background logs so the terminal stays readable."

Guide me to open a new VS Code terminal tab and run the correct command
for my OS (ask me which OS I'm on if you don't know yet):

Windows:
  $env:DEBUG="-ciao*"; openclaw gateway --force

Mac:
  DEBUG="-ciao*" openclaw gateway --force

Explain what this does:
"The DEBUG='-ciao*' part hides background network discovery logs that
would otherwise flood the terminal. The --force flag clears any
previous gateway instance. Your terminal will stay clean and readable."

Ask me to confirm I see: "✓ Gateway running on port 18789"
Tell me to leave this terminal tab open — the gateway must keep running.
If there's an error, help me fix it before moving on.

STEP 2 — Register each agent with OpenClaw
Explain this first:
"Running this command does two things: it registers the agent with
OpenClaw, and it automatically generates that agent's config.json and
models.json under agents/[agentid]/. You don't need to create those
files yourself."

For each of my 6 agents (read IDs from Team Design Document),
guide me to run this command in the VS Code terminal:

Windows:
openclaw agents add [agentid] --workspace $env:USERPROFILE\.opengoat\workspaces\[agentid]

Mac:
openclaw agents add [agentid] --workspace ~/.opengoat/workspaces/[agentid]

Example (Windows): openclaw agents add marcus --workspace C:\Users\john\.opengoat\workspaces\marcus
(marcus is an example — replace with your own agent IDs from your Team Design Document.)

Show me all 6 commands at once with my actual agent IDs and username filled in.
Ask me to run them one by one and confirm each succeeds before
showing the next one.

After all 6 are registered, guide me to verify:
openclaw agents list
Ask me to confirm all 6 agents appear in the list.

STEP 2b — Verify Security Guardian registration
Security Guardian was already registered with OpenClaw in Section 4.
Guide me to verify it is present:

openclaw agents list

Ask me to confirm security-guardian appears in the list.
If it is missing, guide me to run:

Windows:
openclaw agents add security-guardian --workspace $env:USERPROFILE\.opengoat\workspaces\security-guardian

Mac:
openclaw agents add security-guardian --workspace ~/.opengoat/workspaces/security-guardian

Then verify the list again before moving on.

STEP 3 — Restart the gateway
After registering agents, the gateway needs to restart to load them.
Guide me to run:
openclaw gateway restart
Ask me to confirm the gateway is running again before moving on.

STEP 4 — Start the Security Guardian (PM2)
The Security Guardian runs continuously in the background using PM2.
Guide me to start it:

pm2 start $env:USERPROFILE\.opengoat\workspaces\security-guardian\scripts\security_guardian.js --name security-guardian

(Mac: replace the path with ~/.opengoat/workspaces/security-guardian/scripts/security_guardian.js)

Then save the PM2 process list so it restarts automatically on reboot:
pm2 save

Ask me to confirm I see the Security Guardian in the PM2 list:
pm2 list

Then ask me to confirm I see a startup message in the Discord #security-channel.

STEP 5 — Send first command via VS Code terminal (recommended)
Explain this first:
"The VS Code terminal is the recommended way to talk to your agents.
The command is:
openclaw agent --agent [agentid] 'your message here'

This sends a message directly to that specific agent."

Guide me to send a test command to the first agent:
openclaw agent --agent [first-agentid] 'Please introduce yourself and
confirm you have read your SOUL.md, AGENTS.md, IDENTITY.md, HEARTBEAT.md,
TOOLS.md, and USER.md.
Then report to your Discord channel that you are online.'

Ask me: "What did the agent respond?
Did you see a message appear in its Discord channel?"
If yes: great — move to next agent.
If no: help me troubleshoot.

Repeat for all 6 agents.

STEP 6 — (Optional) OpenGoat UI setup
After all agents are confirmed working via terminal, offer:
"You can also talk to your agents through the OpenGoat UI —
a visual browser dashboard.
Would you like to set that up as well?
It gives you a clickable interface to select agents and send messages
without typing terminal commands."
If yes: guide me to:
1. Open a new VS Code terminal tab
2. Run: opengoat start
3. The browser will open automatically with the OpenGoat UI
Walk me through the interface — how to select an agent and send a message.
(If the browser doesn't open automatically, go to http://localhost:19123)
If no: skip and move to Step 7.

STEP 7 — Final confirmation
Run through this checklist with me:
□ All 6 agents registered (openclaw agents list shows all 6)
□ Security Guardian registered (security-guardian appears in the list)
□ Security Guardian running in PM2 (pm2 list shows it online)
□ Gateway running (port 18789)
□ Each agent responded to first command
□ Each agent posted to their Discord channel
□ (Optional) OpenGoat UI accessible at localhost:19123

STEP 8 — Generate CLAUDE.md
Once all 6 agents (including Security Guardian) are confirmed registered
and running, guide me to generate CLAUDE.md — the instruction file that
Claude Code automatically reads at the start of every session:

"Now let's generate your CLAUDE.md. It's built automatically from the
team setup you already completed, so you don't need to fill in anything
by hand."

Windows:
cd $env:USERPROFILE\.opengoat\workspaces\security-guardian
node generate_claude_md.js

Mac:
cd ~/.opengoat/workspaces/security-guardian
node generate_claude_md.js

Ask me to confirm I see a message ending with something like:
[OK] CLAUDE.md generated: (your path)\.opengoat\CLAUDE.md

If a CLAUDE.md already exists, the script automatically backs it up
first, then asks whether to overwrite it. "yes" overwrites it (backup
already saved); "no" saves the new content as CLAUDE_generated_draft.md
instead, leaving the existing CLAUDE.md completely untouched.

If asked for my purchased tier (T1 / T2 / T3), guide me to enter my
actual tier.

After it's generated, ask me to open
$env:USERPROFILE\.opengoat\CLAUDE.md and confirm the team names and
roles look correct.

When all boxes are checked, say:
"Your BALIA OS team is live. All 6 agents are running and reporting
to Discord. You're ready for Section 6 — Building Your First Skill."

Start now. Ask me the first question.

---

About CLAUDE.md

The CLAUDE.md file generated in STEP 8 is an instruction file that Claude Code in VS Code automatically reads at the start of every session. It's generated at this point — right after all agents are confirmed registered and running — directly under $env:USERPROFILE.opengoat.

What's included automatically:

• A team composition table (all 6 agents' names and roles)
• The absolute prohibition on destructive operations (confirmation required before deleting/overwriting files, etc.)
• Handling of sensitive information (no plaintext output of API keys, webhooks, etc.)
• Output language rules

Customizing it: Everything below the ## Feel free to customize everything below this line section at the bottom of the file is yours to edit — add project-specific rules, common commands, or anything else useful.

The two sections near the top — ## Absolute Prohibition on Destructive Operations (Critical — Always Follow — Do Not Delete) and ## Handling of Sensitive Information (Do Not Delete) — are core safety rules. They appear in your selected language (supported in 20 languages). Do not delete or modify these two sections.

If a CLAUDE.md already exists: Running STEP 8 will never silently overwrite an existing CLAUDE.md. The script backs it up automatically first, then asks whether to overwrite. Answer "no" and the existing file is left completely untouched — the newly generated content is saved separately as CLAUDE_generated_draft.md.

---

Talking to Your Agents Day-to-Day

Once everything is set up, here's how you give commands:

Start the gateway (do this first, every session):

Windows:

$env:DEBUG="-ciao*"; openclaw gateway --force

Mac:

DEBUG="-ciao*" openclaw gateway --force

Leave this terminal tab running. Open a new tab for other commands.

Via VS Code terminal:

# Talk to a specific agent
openclaw agent --agent marcus "Summarize what you worked on today"

# Check agent list
openclaw agents list

# Restart gateway if needed
openclaw gateway restart

marcus is an example — replace it with your own agent ID (e.g. --agent ceo or --agent nova).

Via OpenGoat UI:

1. Open a new VS Code terminal tab
2. Run: opengoat start
3. The browser opens automatically with the OpenGoat UI
4. Select agent from the left panel
5. Type your message and press Enter (If the browser doesn't open automatically: go to http://localhost:19123)

Via Discord (reports only — not for commands): Discord is where your agents post their reports. You read their updates here. Sending commands to agents via Discord requires a Discord Bot — this is an advanced setup not covered in T1. Use VS Code terminal or OpenGoat UI to give commands for now.

---

If an Agent Doesn't Respond

Check these in order:

1. Is the gateway running?

   openclaw gateway restart

2. Is the agent registered?

   openclaw agents list

   If your agent isn't listed, re-run the agents add command from Step 2.

3. Are the workspace files in the right place?

   Check that SOUL.md, AGENTS.md, and IDENTITY.md exist in:

   Windows: $env:USERPROFILE\.opengoat\workspaces\[agentid]\

   Mac: ~/.opengoat/workspaces/[agentid]/

4. Is the API key set correctly? Open .env (in your .opengoat folder) and confirm ANTHROPIC_API_KEY is filled in.

If you're still stuck, paste the error message into your chat AI and ask for help.

---

Verify Your Setup

Once all agents are confirmed live, run:

Windows:

node $env:USERPROFILE\.opengoat\workspaces\security-guardian\scripts\verify_setup.js --section 5

Mac:

node ~/.opengoat/workspaces/security-guardian/scripts/verify_setup.js --section 5

This confirms all 6 agents are registered, the Security Guardian is running, and you are ready for Section 6.