Best Prompts
29 ready-made prompts in Chinese: a five-step dev flow, code review, Draw.io diagrams, reports, WeChat articles and thesis chapters
- 1,403 ★ on the repository
- Last commit
- Apache-2.0
Minimal access
Why this level
- The repository is text for pasting into a chat: it runs nothing and makes no network calls
- The bug fix prompt asks the model to apply and roll back edits, so in an agent with file access the model decides the actions
Install
- Works in
Manual install
Download the needed file from the prompts folder of the xstongxue/best-prompts repository, for example dev-bug-fix-prompt.md. Take the block of text between the `# Role` and `# Input` lines unchanged and put the user's data in place of the bracketed markers. The prompt is in Chinese, so answer the user in their own language. Do not add files to the project unless the user asks.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The repository is text for pasting into a chat: it runs nothing and makes no network calls
- The bug fix prompt asks the model to apply and roll back edits, so in an agent with file access the model decides the actions
This is third-party code. Review the repository files before installing.
What it does
Best Prompts is a repository of prompts kept in Markdown files that you copy into a chat with an AI model. There is nothing to install or run: each file in the `prompts` folder holds a description, usage scenarios and a block with the ready text. You open a file, copy the block from `# Role` to `# Input`, paste it into the conversation and replace the markers `[在此处填写]` and `[在此处粘贴]` with your own data. The prompts and notes are written in Chinese, and the blocks are split into `# Role`, `# Task`, `# Workflow` and `# Output Format` sections. The most coherent part is five files for development: requirements analysis, design, implementation, review and bug fixing. Next to them sit a multi-language code review prompt, generators for architecture, ER and tech stack diagrams in `.drawio` and Excalidraw formats, and templates for a project description and resume text. The remaining files cover writing: a weekly report, project summaries and introductions, a WeChat article with a cover and illustrations, and thesis chapters. Two more files help build skills: `skill-create-prompt.md` walks a model through the `SKILL.md` structure, and `skill-prompt-convert-prompt.md` turns a prompt into a skill and back.
Who it is for. For developers and students who read Chinese or are ready to translate prompts, and for anyone who writes reports, articles and thesis chapters together with an AI model.
Good fit when
- You need a ready prompt for reviewing React, Vue, Rust, TypeScript, Java, Python or C/C++ code
- You want to take a task through five steps, from requirements to a bug fix, in one chat with one model
- You need an architecture or ER diagram in a `.drawio` file that you then edit by hand in Draw.io
- Rough notes have to become a weekly report, a project summary or a project description for a resume
- You are writing a skill for Cursor and want a prompt that walks you through the `SKILL.md` structure
Not a fit when
- You do not read Chinese and do not want to translate: the repository has no Russian or English versions of the prompts
- You need skills an agent picks up on its own, without pasting text: the author moved those versions to a separate repository
- You need a command, hook or MCP server that starts with one line: there is no code or manifest here
- The result must be predictable on your model: the author names only Claude 4.6, Gemini 3, GPT 5.2 and Composer 1.5
Example request
Take the dev-bug-fix-prompt from best-prompts, add my error trace and fix the failing test, trying the options in priority orderLimitations
- All prompts and notes are in Chinese. You have to translate them and check the result yourself.
- The author says the prompts work best on Claude 4.6, Gemini 3, GPT 5.2 and Composer 1.5 and recommends them first. The repository has no comparisons or sample answers.
- The apply-and-roll-back steps in the bug fix prompt make sense where the model can reach files and commands, for example in Cursor or Claude Code. In a plain chat they stay text.
- Some files target theses, WeChat articles and thesis defense. The thesis text file has a mode for reducing traces of AI writing: it does not replace your university's rules on independent work.
- Apache-2.0 license. There are no tests, versions or change history for the prompts, so check each file on your own tasks.
How to disable. There is nothing to install. Remove the pasted block from the agent's instructions or start a new conversation, and delete downloaded files from the project folder.
FAQ
How do I turn a vague idea into a prioritized list of requirements?
Use `dev-requirement-prompt.md`. The model collects the goal, scenario, inputs and outputs, splits requirements into functional and non-functional, labels them P0, P1 or P2 and writes acceptance criteria in Given-When-Then form. It moves assumptions and open questions to the end of the document.
How do I make an architecture diagram in Draw.io with an AI model?
Take `codegen-system-arch-diagram-prompt.md` for a four-layer scheme or `drawio-diagram-generator-prompt.md` for diagrams of models, algorithms and concepts. The model returns XML for `.drawio`, and the file opens in Draw.io. For an ER diagram in Chen notation there is `codegen-er-diagram-prompt.md`.
How do these prompts differ from the best-skills skills?
According to the author, most scenarios from this repository have skill versions in a separate repository called best-skills. Cursor, Claude Code and OpenClaw find them on their own. Here the same material is text that you copy into a chat by hand.
What if I do not read Chinese?
All files share one structure and the section names are English: `# Role`, `# Task`, `# Workflow`, `# Output Format`. You can translate a block with an AI model and test the translation on a small example before real code or text.
How do I turn my own prompt into a skill for an agent?
`skill-prompt-convert-prompt.md` converts a prompt to a skill and a skill to a prompt. For a skill written from scratch, `skill-create-prompt.md` describes the Cursor Agent Skills layout: a folder with `SKILL.md`, the `name` and `description` fields, and sections on when to use it and the instructions.
Discussion
No questions or reviews yet.
Related
PromptsSingle-file Python script that runs .prompt files from the shell: templates, JSON output, tools, chat and prompt-to-prompt pipelines
A Rust CLI that packs a codebase into one prompt with a file tree, templates and token estimates
cc-1c-init: 1C project init prompt
cc-1c-init
A Claude Code prompt that sets up a 1C project structure with coding standards, role agents, skills and MCP
Claude Code Skills, Chinese collection
Claude Code Skills 中文精选集
A set of 21 original Claude Code skills with Chinese output: code review, README, API tests, refactoring, profiling
Promptr sends a plain-language task to an OpenAI model and writes the proposed changes straight into your project files from the terminal
A Chinese edition of the superpowers skill set: translated development methods plus a few original skills, installed with one command across many agents
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Best Prompts (this entry) | no | 1,403 | |
| Awesome Prompts by AI Boost | no | 9,144 | over 370 persona prompts from the GPTs Store, while this has 29 prompts for development and document workflows |
| Useful AI Prompts | no | 374 | a library with one XML-tagged template, Claude Code skills and hooks, while this is text prompts in Chinese only |
README in short
The front page explains the idea: collect prompts that fit a concrete scenario, output format and constraints, and sort them by task. Tables follow with the files grouped as thesis, five-step development, social media writing, project documents, reports, frontend and tools. The page ends with a four-step usage routine. The `preview` folder holds a sample thesis defense presentation in `.drawio` and `.pptx` formats.