# Anthropic Authors - Prompt Design - Being explicit (Highlights) ![rw-book-cover|256](https://readwise-assets.s3.amazonaws.com/static/images/article4.6bc1851654a0.png) ## Metadata **Review**:: [readwise.io](https://readwise.io/bookreview/26424145) **Source**:: #from/readwise **Zettel**:: #zettel/fleeting **Status**:: #x **Authors**:: [[Anthropic Authors]] **Full Title**:: Prompt Design - Being explicit **Category**:: #articles #readwise/articles **Category Icon**:: 📰 **URL**:: [console.anthropic.com](https://console.anthropic.com/docs/prompt-design/being-explicit) **Host**:: [[console.anthropic.com]] **Highlighted**:: [[2023-04-14]] **Created**:: [[2023-04-15]] ## Highlights - Try writing prompts as though you're speaking to a naive nine-year-old child who happens to have an encyclopedic knowledge and exceptional vocabulary. ([View Highlight](https://read.readwise.io/read/01gxztfbcthe20m48j5vq4xwx4)) ^509001468 - There are two ways to give examples. One is by putting words in Claude's mouth by crafting a previous conversation, another is by providing examples directly. ([View Highlight](https://read.readwise.io/read/01gxzthftzka143x46agjnb4yg)) ^509001857 - Claude is trained to act as a chatbot, but that doesn't mean it has to be a general conversational chatbot. By instructing Claude to act in a certain way, it can play a wide variety of different roles with different voices, as necessary for your particular use case. ([View Highlight](https://read.readwise.io/read/01gxztk7vp0bew9gzekq5ajbjn)) ^509001952 #example ``` Human: For the rest of this conversation, I want you to act as a math teacher. I will provide some mathematical equations or concepts, and it will be your job to explain them in easy-to-understand terms. This could include providing step-by-step instructions for solving a problem, demonstrating various techniques with visuals or suggesting online resources for further study. Assistant: OK, got it, I'll be a math teacher! Human: What is Pythagoras' theorem? Assistant: ``` - Claude is a conversational agent so it'll often answer in a verbose manner, but sometimes you want it to get to the point quickly without any preamble. To achieve this, or to elicit a response in a particular manner or format, you can put words in Claude's mouth by starting the sentence yourself. ([View Highlight](https://read.readwise.io/read/01gy042dpfj3xgm3d134padced)) ^509079522 - XML tags like `<tag>this</tag>` can be useful for defining structure in your prompt, and also in Claude's response. ([View Highlight](https://read.readwise.io/read/01gy043296zejbkaxjaaef7nmw)) ^509079683 - One thing to try when you see a failure like this is to *fully explain the task, anticipate the failure in the prompt, and explain it to Claude.* In the case of information extraction in the case above, one might first state: ([View Highlight](https://read.readwise.io/read/01gy046wwksm77em03py8carpv)) ^509080477 > Human: I'm trying to look at documents to find out if they contain information I can use to answer a specific question. There might be information in some of these documents I can use to answer the question, but the documents I've found might also not be at all helpful or relevant to the question. I'm going to send you the document and I want you to find out if the document contains an answer to my question. Do you understand?