Free 🧰 Practical LLM Lessons

LLM Lab 10: Mini Agent Loop

Combine planning, action, and self-checks to build a simple agent-like workflow in chat.

Published January 8, 2026

Previous

LLM Lab 9: Debugging Prompts

Next

Practical LLM: Electrician Safety-First Checklist

From single prompt to loop

Agents iterate: plan, act, observe, and adjust. You can mimic that loop manually inside a chat.

Can You Guess?

Which sequence best mirrors an agent loop?

Manual agent pattern

Follow the steps and watch each checkpoint light up as you progress.

1

Step 1: Plan

Ask the model to list the steps to solve the task.

2

Step 2: Act

Have it execute only the first step (draft, query, or calculation).

3

Step 3: Observe

Check the output against constraints or real data.

4

Step 4: Adjust

Tell it what passed or failed and request the next step.

5

Step 5: Loop

Repeat until the task is complete or blocked.

Try this now

Run a mini agent by hand

  • Goal: create a 3-step onboarding email sequence for a new app.
  • Step 1: ask the model for a plan only.
  • Step 2: pick one step to execute, review it, and give feedback.
  • Step 3: loop until you are happy, then ask for a final packaged answer.

    • Key Takeaways

    • ✓ Agents iterate instead of answering once.
    • ✓ Feedback between steps improves quality.
    • ✓ You can simulate tool use by checking and feeding results back.

    Quick poll

    What brought you to this free lesson?

    Lesson: LLM Lab 10: Mini Agent Loop

    ← Previous Lesson LLM Lab 9: Debugging Prompts Next Lesson → Practical LLM: Electrician Safety-First Checklist
    Back to Library Unlock More Lessons