Write the First Prompt and Output Format

Unit ID: PAW-M12-U02 Estimated active time: 25-30 minutes

Translate the canvas, do not summarise it

The prompt is the canvas rearranged into instructions. Nothing in the canvas should be dropped in translation — if it did not earn a place in the prompt, remove it from the canvas too.

A working order

Context first, then source, then task, then constraints, then output specification, then how to handle gaps.

Putting the task before the context makes the tool start answering before it has read the material, and putting constraints last means they are read after the shape is fixed.

The prompt skeleton

PROMPT — (task name), v1

CONTEXT
(audience line; situation; known facts, marked do-not-change)

SOURCE
--- BEGIN --- (material) --- END ---

TASK
(one sentence, matching the canvas)

CONSTRAINTS
Include:
Avoid:
Preserve exactly:

OUTPUT
Format:
Sections:
Closing sections:   assumptions | what I do not know | not covered

IF SOMETHING IS MISSING
Ask, or list it under "what I do not know". Do not fill it in.

Expect v1 to be wrong

It usually is, and that is the design. Module 7's repair loop is the next unit, and the first prompt exists to be improved rather than to be right.

Do not spend an hour polishing v1. Spend ten minutes and run it.

Version it from the start

Label it v1, date it, and keep it when you write v2.

Comparing versions is how you learn what actually helped, and Module 6's maintenance header is what makes the library usable in six months.

Mini practice

Write v1 from your canvas.

Check every canvas row appears somewhere in the prompt.

Run it once and keep the output without editing anything.

Mark complete