A cookbook for ChatGPT prompts, served in Markdown

What it does The repository hosts a Chinese-language guide that treats prompt engineering as a set of repeatable recipes. Each chapter introduces a specific technique—such as instruction prompting, role prompting, or few-shot prompting—and breaks it down into a rigid template of task, instruction, and role, followed by worked examples like customer-service replies or legal documents. It is essentially a markdown textbook hosted on GitHub rather than a software project.

The interesting bit The guide attempts to systematize what is often treated as intuition, offering literal formulas such as 作为[角色]生成[任务] or 按照以下指示生成[任务]：[指令]. This paint-by-numbers approach to LLM interaction is either refreshingly concrete or slightly mechanical, depending on your taste.

Key highlights

• Discrete chapters covering instruction, role, standard, zero/one/few-shot, “let’s think about this,” and self-consistency prompting
• Explicit three-part prompt formulas: task, instruction, and role
• Worked examples across domains like customer service, legal writing, and product reviews
• 1,004 stars suggest a sizable audience finds the cookbook format useful
• Pure documentation: no code, dependencies, or runtime; just markdown chapters

Caveats

• The repository is essentially a markdown book, not a tool or library you can run
• The README source shows truncation artifacts (e.g., ``), so the later chapters may be incomplete in the hosted version
• The prose is functional but reads like instructional textbook material rather than original research

Verdict Grab it if you are new to LLMs and want a structured, formulaic starting point for writing prompts in Chinese. Skip it if you are looking for code, automation, or research-level prompt engineering.