What you will be able to do
- Work out why a first answer fell short before you rewrite the prompt
- Pick the change that fits the gap: more context, a tighter spec, an example, or a stated reason
- Get more depth from an answer that is too brief without starting over
- Tell when the next round of prompting is worth it and when a direct edit is quicker
Key concept
Prompt iteration — A weak first answer tells you something is missing from the prompt. Iterating means finding that gap and supplying it, through follow-up instructions, clarifications or a changed prompt. Sending the same request again, or saying it more forcefully, does not supply anything.
1.Read the first answer as a diagnosis
A disappointing first answer rarely means Claude can't do the task. More often, the prompt left out something the answer needed. Anthropic's prompting guidance suggests thinking of Claude as a brilliant but new employee who lacks context on your norms and workflows. A capable new hire given a one-line brief will fill the gaps with sensible guesses about audience, length, format and what counts as done. Those guesses may be different from what you had in mind. The first answer shows you which guesses Claude made.
So iteration starts with a question: what did the answer lack, and what in my prompt would have supplied it? Claude's help centre lists the options you have after a first attempt: follow-up instructions, clarifications, or a request to rewrite the answer. Each one adds something the first prompt didn't have. Anthropic's 'golden rule' is a quick way to find the gap. Show your prompt to a colleague who knows little about the task and ask them to follow it. If they would be confused, Claude will be too.
2.Match the change to the gap
Different kinds of weak answer need different fixes. If the answer is generic, the prompt was probably missing context. The help centre advises writing as if for someone with no background knowledge, and making sure the prompt contains as much context as possible. If you describe a document instead of pasting it, Claude has to work from your summary. If the content is right but the shape is wrong (wrong sections, wrong order, wrong length), the output spec was too loose. The guidance is to be specific about the desired output format and constraints. When order or completeness matters, it recommends numbered steps or bullet points.
Some requirements are hard to put into words, such as a house style, a tone or the layout reviewers expect. Here, adding an example usually beats adding more description. Anthropic calls examples one of the most reliable ways to steer output format, tone and structure. Examples work best when they closely mirror your real use case. This matters when every rewrite costs a full round. One example of the structure your reviewers like, supplied up front, can save several rounds of Claude guessing and you rejecting the guess.
| What the answer shows | What to change in the prompt |
|---|---|
| Right topic, but generic or aimed at the wrong reader | Add context: who it is for, what it will be used for, and the actual source material |
| Content fine, format or length wrong | Be specific about the desired output format and constraints; use numbered steps when order or completeness matters |
| Tone or structure keeps missing after description | Add a few relevant examples, wrapped in <example> tags so they are distinct from instructions |
| Instruction followed literally but applied badly | Explain the reason behind the instruction so Claude can generalise it |
| Instructions, pasted material and examples blur together | Put each type of content in its own tag, for example <instructions>, <context>, <input> |
After rewriting an instruction to say "write in flowing prose, avoid bullet points," a team still finds Claude's responses returning to bulleted lists in some outputs. They notice their own prompt itself is written using bullet points and headers throughout. Which additional iteration is most likely to help?
Correct answer: A — Remove the bullet points and headers from the prompt itself so the prompt's own style matches the desired flowing-prose output.
- A. Correct. Anthropic notes that the formatting style used in the prompt influences Claude's response style, so removing bullet points and headers from the prompt can help the model adopt the desired flowing-prose style.
- B. Incorrect. Adding more bullet points would reinforce, not reduce, the bullet-point pattern the team is trying to eliminate, making it even harder for Claude to avoid that structure.
- C. Incorrect. The core issue is not instruction recency but the prompt's own style. Moving the anti-bullet instruction to a later turn does not address the underlying influence of the prompt's bulleted formatting.
- D. Incorrect. Increasing temperature affects randomness in word choice and phrasing, but it does not discourage structural elements like lists or bullet points, which are determined more by the prompt's formatting.
3.When the answer is too brief or too shallow
Answers that are too brief or too shallow have their own help-centre article. Its advice: if responses are too brief or only partially complete, consider retooling your prompts. It also recommends giving Claude the previous prompt and response when you write the next prompt, so Claude can pick up where it left off. Don't discard a shallow draft and start over. It is useful input. It shows Claude what already exists, and your follow-up can say what the next version needs to add.
The second part of the fix is to ask for the depth you want. The best-practice guide says that if you want 'above and beyond' behaviour, you should explicitly request it rather than relying on the model to infer this from vague prompts. Its own example contrasts a bare request with one that states the ambition:
Create an analytics dashboard. Include as many relevant features and interactions as possible. Go beyond the basics to create a fully-featured implementation.First, supply the context the depth depends on: who the briefing is for, what they will decide with it, and the source material. Second, ask outright for the depth you need (which areas to expand and how far), carrying the previous prompt and response forward. Resending the original prompt adds neither.
A prompt asks Claude to solve multi-step math word problems, but a review of outputs shows occasional arithmetic slips that go uncaught. The team wants to iterate on the prompt to catch these errors before the final answer is returned, without switching to a different evaluation pipeline. Which addition to the prompt best addresses this?
Correct answer: A — Append a step instructing Claude to verify its answer against the problem constraints before giving the final answer.
- A. Correct. Adding an explicit self-verification step prompts Claude to check its solution against the original constraints, which can catch arithmetic errors before the final output. This technique directly addresses the need for error detection without altering the evaluation pipeline.
- B. Incorrect. Removing instruction to show steps discourages structured, step-by-step reasoning that helps maintain accuracy on multi-step math problems. Without explicit reasoning, arithmetic slips are more likely to occur and go unnoticed.
- C. Incorrect. Suppressing intermediate work hides the reasoning chain, preventing both the model and reviewers from spotting where arithmetic errors happen. It eliminates any chance for self-correction and does not add a verification mechanism.
- D. Incorrect. Repeating the same approach twice does not introduce a genuine check, because identical reasoning is likely to reproduce the same mistake. Since only the first result is used, the second computation provides no error-catching benefit.
4.Give reasons, not emphasis, and know when to stop
When Claude misses an instruction, the easy reaction is to repeat it in capitals. Anthropic's guide gives a direct counter-example. It marks a bare 'NEVER use ellipses' as less effective and prefers a version that gives the reason:
Your response will be read aloud by a text-to-speech engine, so never use ellipses since the text-to-speech engine will not know how to pronounce them.The reason is what makes the rule transferable. The guide notes that Claude is smart enough to generalise from the explanation. Once Claude knows the text will be spoken, it can also avoid other things a speech engine would stumble over, not only the one character you named. Emphasis gives Claude no new information about your goal. A reason does.
Iteration also has a stopping point. Anthropic's prompt-engineering overview says that not every success criterion or failing eval is best solved by prompt engineering. Suppose a draft is accurate and well structured after several rounds, and the only remaining faults are a couple of known, small errors such as wrong names in two headings. Another round risks changing what already works, and it costs time. Fixing those lines by hand is often the better call, especially against a deadline. Save another prompt round for gaps you can't close with a quick edit.
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.If Claude ignores an instruction, writing it in capitals (NEVER ...) is the strongest fix.Why is that wrong?
Anthropic's guide marks the capitalised bare rule as less effective. Stating the reason behind the rule works better because Claude can generalise from it.
Covered in Give reasons, not emphasis, and know when to stop
2.If an answer is too shallow, start a fresh conversation and resend the same prompt.Why is that wrong?
The help centre recommends retooling the prompt and giving Claude the previous prompt and response so it can build on them, along with an explicit request for the depth you want.
Covered in When the answer is too brief or too shallow
3.When the structure keeps coming out wrong, add more description of the layout each round until Claude gets it.Why is that wrong?
Structure and tone are hard to describe. A relevant example of the shape you want is one of the most reliable ways to steer them, and supplying it up front saves rounds.
Covered in Match the change to the gap
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practicesOfficial docs
“Think of Claude as a brilliant but new employee who lacks context on your norms and workflows.”
↩︎ Read the first answer as a diagnosis“Show your prompt to a colleague with minimal context on the task and ask them to follow it.”
↩︎ Read the first answer as a diagnosis“Be specific about the desired output format and constraints.”
↩︎ Match the change to the gap“Relevant: Mirror your actual use case closely.”
↩︎ Match the change to the gap“explicitly request it rather than relying on the model to infer this from vague prompts”
↩︎ When the answer is too brief or too shallow“Providing context or motivation behind your instructions, such as explaining to Claude why such behavior is important, can help Claude better understand your goals”
↩︎ Give reasons, not emphasis, and know when to stop“Claude is smart enough to generalize from the explanation.”
↩︎ Exam trap 1“Examples are one of the most reliable ways to steer Claude's output format, tone, and structure.”
↩︎ Exam trap 3 - 2.https://support.claude.com/en/articles/7996857-my-prompt-isn-t-giving-me-a-helpful-answerOfficial docs
“Give Claude feedback: you provide Claude with follow-up instructions, clarifications, or ask it to rewrite an answer.”
↩︎ Read the first answer as a diagnosis“Make sure that your prompt contains as much context as possible.”
↩︎ Match the change to the gap“Give Claude feedback: you provide Claude with follow-up instructions, clarifications, or ask it to rewrite an answer.”
↩︎ Key concept - 3.https://support.claude.com/en/articles/8114518-claude-s-response-to-my-prompt-is-too-briefOfficial docs
“If its responses are too brief or only partially complete, consider retooling your prompts.”
↩︎ When the answer is too brief or too shallow“We recommend giving Claude the previous prompt and response when writing your next prompt to pick up where it left off.”
↩︎ Exam trap 2 - 4.
“Not every success criteria or failing eval is best solved by prompt engineering.”
↩︎ Give reasons, not emphasis, and know when to stop