Accelerated Intelligence

Stage 3: Rhetoric

The Art of Application and Expression

"Now use it, build with it, and teach it."

Stage 3: Rhetoric

"Now use it, build with it, and teach it."

Purpose of This Stage

Rhetoric is about output and expression. You know the parts. You understand how they connect. Now you use that knowledge to install, configure, customize, troubleshoot, and ultimately teach others.

Everything in Grammar and Logic was preparation for this stage. Installation commands that would have been mysterious incantations now make sense. Configuration choices that would have been guesswork are now informed decisions. Errors that would have been panic-inducing are now solvable puzzles.

In classical education, Rhetoric was reserved for the most mature students (roughly ages 14-18) because it requires both Grammar and Logic as prerequisites. Rhetoric is not just about speaking well. It is about using knowledge effectively in the real world: creating things, solving novel problems, persuading others, teaching, and making judgments under uncertainty.

Module 3.1: Installation and Setup

Learning Objectives:

  • Successfully install OpenClaw on a Mac
  • Complete the onboarding wizard with understanding of each step
  • Connect at least one messaging channel
  • Verify the setup is working and secure

The Installation Sequence

Because you now understand what each component is (Grammar) and how they relate (Logic), the installation process will make sense instead of feeling like magic:

Step 1: Install Node.js (version 22+)

This is the engine OpenClaw runs on. The recommended method on Mac is using Homebrew:

bash
brew install node

You can verify the installation by typing:

bash
node --version

It should show v22 or higher.

Step 2: Install OpenClaw globally

bash
npm install -g openclaw@latest

This downloads the OpenClaw package from the npm registry and makes it available as a command you can run from anywhere in Terminal. The -g flag means "global" - install it system-wide, not just in one folder.

Step 3: Run the onboarding wizard

bash
openclaw onboard --install-daemon

The wizard walks you through:

  • Choosing your AI model provider
  • Entering your API key
  • Connecting your first channel (WhatsApp is the most popular)
  • Configuring security settings
  • Installing recommended skills

The --install-daemon flag tells it to set up launchd so the Gateway auto-starts with your Mac.

Step 4: Run the doctor

bash
openclaw doctor

This checks your setup for security issues, permission problems, missing dependencies, and misconfigurations. Fix anything it flags before using the system.

Step 5: Test

Send a message to your assistant through your chosen channel. Verify it responds. Check the Gateway logs for any errors. Run:

bash
openclaw status --all

for a complete diagnostic report.

Why Installation Is in Rhetoric, Not Grammar

Installation is an act of application, not definition. If you had jumped straight to "npm install" without Grammar and Logic, you would not understand what the wizard is asking you, what your configuration choices mean, or why the doctor flagged something. Now you do. Every step above connects to concepts you already understand.

Best Practice: Never Install on Your Main Machine First

Multiple security researchers and experienced users recommend running OpenClaw on a separate, dedicated machine - a Mac Mini, a Raspberry Pi, or a cloud server. This isolates the agent from your primary computer with all its sensitive files, passwords, and accounts. Think of it as giving your new employee their own desk in their own office, not a chair at your personal workstation.

Module 3.2: Customization

Learning Objectives:

  • Customize the agent's personality and behavior
  • Configure channel routing and security
  • Set up scheduled tasks (cron jobs)
  • Install and evaluate community skills

Key Customization Points

SOUL.md - Personality

Give your agent a name, a personality, and a communication style. Be specific. "You are a professional assistant who communicates concisely and avoids jargon" produces very different behavior from "You are a casual, witty helper who uses humor." The more detail you provide, the more consistent the personality becomes.

AGENTS.md - Behavior Rules

Define what the agent should and should not do, how it should handle different situations, what tools it should prefer, and any hard boundaries. This is the master instruction set. Examples:

  • "Never send emails without my explicit approval"
  • "Always save important information to MEMORY.md before ending a session"
  • "When unsure, ask me rather than guessing"

Channel Routing

Configure which channels the agent monitors, which groups it responds in (all vs. only when @mentioned), and how it handles DMs vs. group chats. You can also set allowlists to restrict which users can interact with the agent.

Cron Jobs - Scheduled Tasks

Set up scheduled tasks. Examples:

  • Daily email summaries
  • Weekly project status checks
  • Morning briefings
  • Automated social media monitoring

Cron jobs run in isolated sessions so they do not pollute your main conversation.

Skills

Browse available skills, evaluate them by reading their SKILL.md files, and install the ones that match your workflow. Start with bundled skills before exploring community skills.

Module 3.3: Troubleshooting

Learning Objectives:

  • Use the three-layer diagnostic framework
  • Run openclaw doctor to identify and fix issues
  • Interpret error messages and logs
  • Know when to restart vs. reconfigure vs. reinstall

The Three-Layer Diagnostic Framework

When something goes wrong, your Logic-stage knowledge of the three-layer architecture becomes your diagnostic framework:

1. Is the Gateway running?

Check: Type openclaw status in Terminal. Is the process alive? Can you connect to the WebSocket?

If not, the entire system is down. Restart with:

bash
openclaw gateway start

If that fails, run:

bash
openclaw doctor --fix

to attempt auto-repair. Check the logs for error messages.

2. Is the Channel connected?

Check: Does the specific platform show as connected in the Control UI (the web dashboard)?

If WhatsApp is down but Discord works, the problem is the WhatsApp channel adapter, not the system. Check credentials, re-scan the QR code, or restart the specific channel.

3. Is the AI model responding?

Check: Do you have a valid API key? Is the model provider having an outage? Have you exceeded rate limits or run out of credits?

If the Gateway is up and the channel works but you get no response, the LLM layer is the likely culprit. Check your API key balance and the provider's status page.

Cross-cutting: Is there a skill conflict?

Did you recently install a new skill? Skills can introduce conflicting instructions or even prompt injection. Try disabling the new skill and testing again.

Cross-cutting: Is it a memory/session issue?

Did the session recently reset? Was compaction triggered? The agent may have lost context. Check if critical information was saved to workspace memory files.

Why This Framework Works

This diagnostic approach only works because you learned the architecture in Logic. Without that mental model, troubleshooting is random guessing - restart everything and hope. With it, you systematically narrow down where the problem lives in 60 seconds, then apply the right fix to the right layer.

Module 3.4: Teaching Others

Learning Objectives:

  • Explain the Trivium method to someone new
  • Walk a non-technical person through Grammar concepts
  • Demonstrate the value of understanding before doing
  • Recognize when someone is ready to advance stages

The Ultimate Proof of Mastery

The ultimate proof of mastery is the ability to teach. This is also where your own gaps reveal themselves - the concepts you thought you understood but cannot explain simply to someone else.

The Teaching Framework

Follow the Trivium itself:

Teach the Grammar first: Give them vocabulary. Define terms. Use analogies. Never assume they know what localhost, daemon, or API key means. If they cannot name the seven core components and explain each one, they are not ready for Logic.

Teach the Logic second: Walk them through the message flow. Explain cause and effect. Show them the three-layer architecture. Make sure they can predict what happens when settings change. If they cannot trace a message from phone to AI and back, they are not ready for Rhetoric.

Guide the Rhetoric third: Walk them through installation with commentary, not just commands. Every step should come with context: "We are installing Node.js because..." "The onboarding wizard is going to ask about channels because..." They should understand why before they do what.

Congratulations!

You have completed the three stages of the Trivium. You now have:

  • Grammar: A rich vocabulary of OpenClaw components and concepts
  • Logic: A mental model of how the system works and why
  • Rhetoric: The ability to install, configure, troubleshoot, and teach

You are no longer following instructions blindly. You are making informed decisions. When something breaks, you can diagnose it. When you want to customize, you understand the implications. When someone asks for help, you can explain clearly.

This is mastery.

Ready for the Final Quiz?

Take the Rhetoric Stage quiz to prove you can apply your knowledge. You need 70% or higher to complete the learning system.

The quiz will cover:

  • Installation and setup procedures
  • Customization best practices
  • Troubleshooting methodology
  • Teaching and knowledge transfer

Ready for the Final Challenge?

Take the Rhetoric Stage quiz to prove you can apply your knowledge in real-world scenarios. You need 70% or higher to complete the Trivium Learning System.