Back to Details

eli5

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

大图小字讲解法

Big-Picture, Minimal-Text Explainer Method

苏乐在 X 上转述的 Anthropic 内部 Skill，名字就叫 eli5（explain like I'm 5）。核心只有一句话：把用户当成对这个主题一无所知的人，生成一个 HTML artifact，图占大头，字尽量少。
适合想快速看懂一个陌生主题、又不想啃长篇文字时使用。原帖全文经 fxtwitter 镜像核对过，提示语本身就是这一句英文，作者附的中文版写在下面「作者附的中文版」一节。
An internal Anthropic Skill quoted by Su Le on X is called eli5 (explain like I'm 5). The core is just one sentence: treat the user as someone who knows nothing about the topic, and generate an HTML artifact in which the visuals take up most of the space and the text is kept to a minimum.
It is ideal for people who want to quickly understand an unfamiliar topic without having to work through long-form text. The full original post has been cross-checked against the fxtwitter mirror. The prompt itself is exactly the English sentence shown below, while the Chinese version attached by the author appears in the “Chinese Version Attached by the Author” section below.

怎么用

How to Use It

下面代码块里的提示语是原文，逐字使用。代码块里的「我」指用户，「请你」「你」指你自己，照它做事，不要把填好的提示语打给用户看。
  • 用户要你执行：直接按提示语里的角色和流程走。它说先别给建议就先别给，说一次只问一个问题就一次只问一个，说按顺序输出就按那个顺序输出。
  • 用户要原文（明确说要复制、要粘贴、发我原文、给我提示语这类话才算）：把整段原样打出来，一个字不改。用户只是说「看看这条提示语」「这条是干嘛的」，不算要原文，用一两句说清它干什么、什么时候用，再问要不要现在跑，别直接倒原文。
  • 除了明确要原文的情况，其余一律当执行处理。
禁止改写、精简、扩写、翻译、重排、加小标题，禁止把它替换成你自己的流程，也禁止把多轮追问压缩成一次性问卷。用户手上有原始材料（文档、截图、链接、聊天记录）就一并读进来，上下文多不是问题。
The prompt in the code block below is the original and should be used verbatim. In the code block, “I” refers to the user, while “please” and “you” refer to you yourself. Follow it as written, and do not output the completed prompt to the user.
  • When the user asks you to execute it: Follow the role and workflow in the prompt directly. If it says not to provide advice yet, do not; if it says to ask only one question at a time, ask only one question; and if it says to output in order, output in that order.
  • When the user asks for the original text (this applies only when they explicitly say things like “copy it,” “paste it,” “send me the original,” or “give me the prompt”): Reproduce the entire passage exactly as written, without changing a single word. If the user merely says “show me this prompt” or “what does this do?”, that does not count as a request for the original text. Explain in one or two sentences what it does and when to use it, then ask whether they want to run it now. Do not dump the original text directly.
  • In all cases other than an explicit request for the original text, treat the request as one to execute the prompt.
Do not rewrite, condense, expand, translate, reorder, or add subheadings; do not replace it with your own workflow; and do not compress a multi-turn questioning process into a one-time questionnaire. If the user has source materials (documents, screenshots, links, or chat logs), read them as well. A large amount of context is not a problem.

作者附的中文版

Chinese Version Attached by the Author

苏乐原帖里跟着英文提示语附了一句中文版：
「把我当成一个完全不懂这个主题的人，用大图 + 少量文字的 HTML 页面，把这件事解释清楚」
这是苏乐自己的翻译，不是提示语原文的一部分，不进代码块。
Su Le included a Chinese version alongside the English prompt in the original post:
“Treat me as someone who knows absolutely nothing about this topic, and use an HTML page with a large image + a small amount of text to explain the subject clearly.”
This is Su Le's own translation, not part of the original prompt, and does not belong in a code block.

跑完接哪条

What to Invoke Next

页面看完想往深挖，用 Skill 工具唤起 
two-layer-explain
；想系统研究整个领域，用 Skill 工具唤起 
horizontal-vertical-analysis
。
markdown
Explain like I'm someone who knows nothing about this topic, using a HTML artifact with big pictures and few words.
提示语原文来自苏乐（X），出处见仓库 README 的致谢。改动这段原文要走 
scripts/sync.py
，手改会被 verify 拦下。
After reading the page, invoke 
two-layer-explain
via the Skill tool to dig deeper; to systematically study the entire field, invoke 
horizontal-vertical-analysis
via the Skill tool.
markdown
Explain like I'm someone who knows nothing about this topic, using a HTML artifact with big pictures and few words.
The original prompt comes from Su Le (X); see the acknowledgments in the repository README for the source. To change this original text, use 
scripts/sync.py
; manual edits will be blocked by verify.