CCAO-F · Study Guide

← Domain 1: Prompting and Task Execution

1.3 · Lesson 3 of 4

Iterative Prompt Improvement

What you need to know

On the CCAO-F exam, the first claude.ai response is a draft to refine, not a verdict. Improvement comes from specific feedback that says what to keep, what to change, and why. Regenerating and hoping gives you a new sample of the same prompt instead of progress.

Write success criteria before you iterate

Decide what “done” looks like up front: length, reading level, required points, tone, audience. Score each draft against that list. When every item passes, stop. Without criteria, “better” has no finish line.

Feedback that moves the draft

Useful feedback is concrete:

  • Names the gap (“paragraph two assumes readers know our pricing tiers”).
  • States the fix (“add a one-line definition of each tier”).
  • Says what to preserve (“keep the pricing table and the closing CTA”).

Weak feedback sounds like “too long,” “try again,” or “make it more professional” with no target.

Same thread vs new chat

Keep iterating in the same conversation while the context is useful. Start a new chat when the thread has drifted, accumulated contradictions, or grown so long that earlier constraints are getting lost. If you restart, carry a short, edited summary of decisions and open items — not the entire messy history.

Worked example

A project brief draft buries the key ask in paragraph three. Instead of rewriting everything, send: “Move the request into the first sentence. Keep the rest of the structure and the timeline table as they are.”

That signal is exam-shaped: one named problem, one named fix, explicit preservation. The draft improves in one turn instead of five blind regenerations.

Exam traps

  • Regenerate until it looks good

    Without new information you get another random draw from the same prompt. That is not iteration.

  • "Make it better" is enough feedback

    Neither you nor Claude can tell when you are done unless success criteria exist.

  • Rewrite the whole prompt from scratch after every miss

    Wholesale rewrites hide which change helped. Change one element at a time against a stable checklist.

  • Fix style and ignore what should stay

    If you do not say what to preserve, a useful apology or table can disappear in the next draft.

Practice scenario

Claude’s first draft of a customer update is too formal and about two paragraphs too long. The sender clicks regenerate five times. What is the better move?

Choose one answer

Build exercise

Iterate a Prompt With Targeted Feedback

1.3 · Intermediate · ~30 min · claude.ai

What you will practise

  • Why specific feedback beats regenerating and hoping.
  • How to write success criteria before the first draft.
  • How to name exactly what to keep and what to change.
  • How to stop when the criteria are met.

Full steps, hints, and the skill check live on the domain exercises page. Keep this lesson open while you work so you can apply what you just studied.

Open Domain 1 exercises

Path: /learn/claude-associate/exercises/1-prompting-task-execution

Sources

View progress