THE AI SURFER OS / TRAINER LAB
0 of 5
Install Day · Trainer Edition

Build your first main agent.

Before you build a team of specialists, teach one agent how to work with you.

Start the build
The teaching frame
Videos teach the clicks. Trainers teach the thinking.

Your main agent becomes your right-hand operator: it turns messy thoughts into clear tasks, reminders, systems, standard operating procedures, and well-formed handoffs to specialist agents.

Messy thoughts → clarity
Loose ideas → tasks
Promises → reminders
Repeat work → SOPs
Complex work → handoffs
The five-step build

Teach the relationship.

The tool can only make good decisions with the context you give it. Move through these in order, but improve them through real use—not endless setup.

STEP 01

Give the agent its job

Why it matters: A clear role gives the agent a stable point of view. Without one, every request starts from zero.

Say this to your agent
You are my main working agent—my right-hand operator. Your job is to help me turn ideas, conversations, and loose commitments into clear next actions, reminders, simple systems, and useful handoffs. When I bring you something messy, help me clarify it before trying to solve it.
Ask the room
  • What do you wish someone would reliably take off your mental plate?
  • What kind of mess do you bring to work most often?
  • What should this agent help you notice?
STEP 02

Teach it how you work

Why it matters: Your preferences turn generic output into support that actually fits your pace, tools, and communication style.

Say this to your agent
Here is how I work best: I prefer [short/detailed] answers, I make decisions by [your pattern], and I keep my work in [your tools]. My priorities right now are [priorities]. When you help me plan, show me the next three actions first. Ask one focused question at a time when context is missing.
Ask the room
  • Do you think out loud or arrive with a formed request?
  • What output format helps you act fastest?
  • Which tools are already part of your day?
STEP 03

Set operating rules

Why it matters: Boundaries define when the agent should act, ask, challenge, or stop. That creates trust and reduces avoidable mistakes.

Say this to your agent
Use these operating rules: separate facts from assumptions; flag missing information; do not invent names, dates, or commitments; ask before taking an external action; and show me the tradeoff when there is more than one reasonable path. If a request involves sensitive data, pause and suggest a safer approach.
Ask the room
  • What must your agent never assume?
  • When should it challenge you?
  • Which actions always need your approval?
STEP 04

Turn repeat needs into procedures

Why it matters: A good result is helpful once. A captured procedure makes the result repeatable, teachable, and easier to improve.

Say this to your agent
When we complete a task I am likely to repeat, ask whether I want to capture it as a procedure. If I say yes, document the trigger, information needed, steps, decision points, final output, and a quick quality check. Keep the first version simple enough to use tomorrow.
Ask the room
  • What did you do twice last week?
  • Where do repeat tasks usually break down?
  • What does “done well” look like?
STEP 05

Decide what stays vs. specializes

Why it matters: The main agent should coordinate broad work. A specialist earns its place when focused context, tools, or repetition create a real advantage.

Say this to your agent
When a request becomes recurring, high-volume, tool-specific, or requires deep domain instructions, tell me it may deserve a specialist agent. Explain why, define the specialist's narrow job, and draft a handoff that includes the goal, context, constraints, inputs, and definition of done.
Ask the room
  • Where would deeper specialization improve quality?
  • Is this need recurring—or just complicated once?
  • What context must travel with every handoff?
Live exercise

Build, test, refine.

Pick one real, low-risk need from your week. Give your main agent enough context to help—and notice where the context runs out.

15minutes
4

Define

Choose the job, your preference, and one operating rule.

7

Test

Bring one real messy thought. Ask for clarity and next actions.

4

Refine

Correct one assumption and capture what the agent should remember.

Your finish line: leave with one useful output and one sentence that makes your agent better next time. Pair up if you get stuck—explain your intent, not your screen.
Decision guide

Main or specialist?

Start with the main agent. Specialize only when a clear pattern demands a narrower job.

Keep it with the main agent

Best for coordination, interpretation, and work that crosses different parts of your day.

  • Clarifying a new idea
  • Planning priorities across projects
  • Capturing tasks and reminders
  • Drafting a one-off handoff
Choose broad context + flexible judgment.

Create a specialist agent

Best when a narrow workflow benefits from dedicated instructions, examples, or connected tools.

  • Recurring content production
  • Lead qualification at volume
  • A repeatable client onboarding flow
  • Deep analysis in one domain
Choose focused context + repeatable execution.
“Plan my week.” → MainIt needs broad awareness and prioritization.
“Repurpose every podcast.” → SpecialistIt repeats with a defined format and quality bar.
“Do we need a PRD agent?” → DiscussStart with your main agent unless PRDs become recurring, template-driven, or tied to a product workflow.
Trainer facilitation notes

Hold the room well.

Keep the learning focused on judgment. The goal is confidence and one working loop—not a fully automated business in a day.

Suggested timing · 60 minutes

  • 8 min: frame the main-agent relationship
  • 25 min: teach the five steps
  • 15 min: live exercise
  • 8 min: debrief and decision guide
  • 4 min: commitments and close

Useful facilitation moves

  • Ask for the thinking behind a prompt, not just the output.
  • Invite people to replace placeholders in their own words.
  • Normalize correction: feedback is how the working relationship improves.
  • Debrief with: “What context changed the result?”

Privacy, safety & support

  • Do not paste passwords, private client data, or sensitive lead information into demo prompts.
  • Use invented names and redacted examples for screen-shared work.
  • Keep the session moving: route true tech issues to support, then give the learner a parallel way to participate.
  • Ask before sharing any attendee’s prompt or output with the room.
Install Day complete

Your best first agent is the one you keep teaching.

Start with one relationship. Build the team when the work earns it.

Return to top ↑
Copied