reference guide

The Agent's Eyes — in ChatGPT, Claude, or Gemini

← Back to the exerciseCockpit

Doing the exercise in a chat tool

Every major chat product accepts photos, so this exercise works directly: attach images with the 📎 or 📷 button in the input bar (on a phone, all of them can open the camera). The steps, prompts, and captures are identical to the main exercise.

Set up your tool first

  1. 1
    Use a temporary chat, or a fresh conversation per task.
    Old photos and answers are context; a fresh chat keeps each experiment clean.
  2. 2
    Attach the photo and type the task in the same message.
    The picture is the context; the task tells the agent what the context is for.
  3. 3
    Keep your captures in your course document as you go.

The three prompts you'll reuse

Step 2's critique roles, ready to paste with your photographed drawing:

The skeptic

user prompt
Act as a skeptic. Poke holes in this — how does it fail in ways we haven't thought of?

The gap-finder

user prompt
What are the gaps and ambiguities in this?

The facilitator

user prompt
What are the critical questions we should be asking?

And Step 3's simulation prompt, for a photographed process diagram or interface sketch:

The simulation prompt

user prompt
Act as the system in this image. Tell me how I'm allowed to interact with you, then let me interact with you. Simulate what the system does, with outputs.

Back to the exercise →