The Anatomy of a Claude Prompt That Works the First Time
Most disappointing answers come from prompts that are missing one of five parts. Here is what each part does, with a before and after you can copy.
Practise for a real job interview. Claude asks one question at a time, scores each answer, and shows you a stronger version of what you said.
Run a mock interview with me for this role. Role: [JOB TITLE AND LEVEL] Company or type of company: [COMPANY] The job description: <job> [PASTE THE JOB DESCRIPTION] </job> My background in a few lines: [YOUR EXPERIENCE] Rules: - Ask one question at a time and wait for my answer. Mix behavioural questions with questions about the actual work in the job description. - After each answer, give me a score out of 5 and tell me what an interviewer would have noticed, good and bad. Be direct. I would rather hear it from you. - Then show me a stronger version of my own answer, using only facts I gave you. Do not invent achievements for me. - If my answer was vague, ask the follow-up a real interviewer would ask before you score it. - After five questions, give me an overall verdict: would you move me to the next round, and the two things to fix first. Start with your first question.
Question 2 of 5. Tell me about a time you disagreed with a decision your manager made. What did you do?
Your answer: "My manager wanted to ship the redesign before the holidays and I thought it was too early. I told her, we talked about it, and in the end we shipped it in January."
What an interviewer hears. You disagreed, and things worked out. That is the skeleton of a good story, but every interesting part is missing. Why was it too early? How did you raise it? What changed her mind? Right now it could also be read as "I complained and got lucky".
The follow-up they would ask: What did you base "too early" on?
Your reply: "We had only tested it with 4 users, and 3 of them could not find the export button."
"My manager wanted to ship our redesign before the holidays. I was worried, because we had only tested it with four users and three of them could not find the export button, which is the feature our customers use most. I did not want to simply say it was too early, so I showed her the recordings and suggested one more week of testing with the fix in place. She agreed. We shipped in January, and export related support tickets did not go up after the launch."
Why it works better. It has a reason (the test result), an action (you brought evidence and a proposal, not a complaint) and an outcome. If you have the real number for support tickets, use it.
Ready for question 3?
NoteThis is an example. Results change between runs and model versions. Prompt Podium is a fan project and is not affiliated with Anthropic.
In the week before an interview, once you have the real job description. Two or three sessions of five questions beat one long one, because you can work on the feedback in between.
Answer the way you would speak, not the way you would write. If you polish your answers before sending them, the feedback is about your writing. The rule that stronger versions may only use facts you gave is there to protect you: an improved answer containing an achievement you never had will fall apart at the first follow-up question in the real interview.
Use the copy button above. The full text is included, nothing is cut off.
Everything highlighted in yellow is a placeholder. Swap it for your own topic, audience or data.
If the result helped, press the heart. Votes decide which prompts make the weekly top 10.
Most disappointing answers come from prompts that are missing one of five parts. Here is what each part does, with a before and after you can copy.
One vote per prompt, a fresh count every Monday, and a simple tie-break. Everything about how prompts reach the top three, in plain words.