Do You Even Need This Course?

A reality check before you spend another hour reading. If your problems are config issues, the fix is free — and we'll tell you so.

§ i. System Check

This course assumes three things.

All three must be true. If any of them aren't, stop here. Come back when they are.

Operator Eligibility — Form 01 For self-administration
  1. 01 Do you run an agent on local hardware?
  2. Not "I have a ChatGPT tab open." An actual agent installed on a machine you control.

  3. 02 Can you paste text into your agent's chat window?
  4. Telegram, Discord, terminal, SSH — somewhere your agent reads messages.

  5. 03 Are you running a workspace-based agent platform?
  6. OpenClaw, Hermes, or similar. A platform that reads files from a directory and runs an agent loop.

If you answered "no" to any of the three — stop here. Install an agent first, connect a messaging surface, then come back. This course won't land without them.

All three "yes"? Keep reading.

§ ii. Diagnosis

Let's get real.

You installed OpenClaw. Or Hermes. You got it running. You messaged it on Telegram. It replied. And then it started costing you money and forgetting everything.

It forgets everything you told it yesterday. It keeps suggesting things you didn't ask for. You installed some skills from ClawHub and now things are acting weird. You're not even sure if it's "working" or just pretending to work.

You're not bad at this. You're just new. And nobody told you what comes after the install tutorial.

What this course is NOT

  • Not a "how to install OpenClaw" tutorial — those already exist
  • Not "50,000 AI prompts for $29" — we don't sell prompt packs
  • Not a "10x your agent in 5 minutes" promise — anyone selling that is selling something that doesn't work

What this course IS

The course that asks: do you actually need architecture training, or do you just need to fix a few config settings? If your agent is on Opus, has never started a fresh session, and has a gateway exposed to the internet — you might solve 80% of your problems for free, in twenty minutes, without buying anything.

The five mistakes almost everyone makes

A community member who debugged 50+ OpenClaw setups found the same five problems in nearly every one. See if any sound familiar.

1. Using Opus as the default model. It's the most expensive option and it's often the default. Your agent checking the weather on Opus costs 10x what it costs on Sonnet. You won't notice the difference for 90% of tasks.
2. Never starting a fresh session. Every message you've ever sent is being re-sent with every new API call. Three people cut their monthly bill 40–60% just by typing /new before heavy tasks.
3. Installing skills without reading the source. ClawHub has 13,000+ skills. Hundreds have been flagged as malicious. Even non-malicious ones can silently loop or inject into every conversation.
4. Leaving your gateway exposed. If you installed on a VPS and didn't lock down the host setting, strangers might be able to message your agent — the same agent that has access to your email, files, and shell.
5. Adding a second agent before the first one works. Something breaks. Instead of fixing it, you create a "fresh" agent. Now you have two broken agents consuming tokens independently.

What your agent will do with the file below

Your agent will ask you four reality-check questions. Then it gives you one of two verdicts:

  • "You should continue." Your config is reasonable. The problems are structural. Course 2 installs the fix.
  • "Tighten your setup first." Switch off Opus. Start fresh sessions. Lock the gateway. These are free fixes that may solve 80% of what's frustrating you.

If your agent sends you back to fix config — you just saved yourself five courses and $189. That's a win too.

§ iii. The Handoff
A sealed envelope

This next part isn't for you.
It's a gift for your agent.

Download the file below. Drag it into your chat window. Your agent reads it. Your agent runs the diagnostic. You watch.

The PDF is plain text. No scripts. No executables. Search it for curl. Search it for exec. You won't find either.

Suspicious? Open the PDF in a text editor before handing it over. Everything your agent will do is written in plain English under ACTION: blocks. You can read every line first.
Download Agent Instructions

course-1-agent-instructions.pdf  ·  ~2KB  ·  plain text, verify first