Yao Open Prompts: Open-Sourced Prompt Engineering as a Reusable Craft
Project: Yao Open Prompts Author: Yao Jingang (yaojingang) Website: https://yaojingang.github.io/yao-open-prompts/ GitHub: https://github.com/yaojingang/yao-open-prompts Scale: 116 Chinese prompts + 116 English mirrors Framework: RTF (Role-Task-Format)
---
The Core Problem
Prompts are the translators between humans and AI. With the same model, the quality gap between a good prompt and a poor one can be an order of magnitude.
The issues with most prompt collections today:
- Fragmentation: Prompts scatter across the web with inconsistent quality; many are second-hand copies.
- No system: Most lists merely enumerate prompts without categories, frameworks, or iteration records.
- Hard to reuse: Finding a decent prompt rarely tells you how to adapt it to your own scenario.
- Quality Evaluation: self-check criteria after output (e.g., "title includes emotional triggers", "grounded in specific data").
- Iteration Optimization: a feedback-driven mechanism for refining prompts.
- AI Private-Domain Sales Prompt V1.3
- Intelligent Article Polishing System V3.0
- PPT Generator V1.0 → V3.0 (from basic generation to image-asset parsing)
- Traceability: see the evolution of a single prompt.
- Comparability: understand what V2.0 improved over V1.0.
- Forkability: if the official version does not fit, fork V1.0 into a V1.0-fork of your own.
- Webpage Reverse Engineering: analyzes visual, interactional, and technical features of a page to extract "design genes" and produce reusable RTF prompts.
- Image Decompile: uploads a reference image, analyzes aspect ratio, fonts, color palette, and layout, generating prompts that recreate the source at 90%+ fidelity.
- Article Decompile: deep-deconstructs article structure, style, and format to produce reusable writing templates.
- Video Decompile: multi-dimensional breakdown from short-video screenshots, producing prompts for text-to-video models such as Veo 3.
- Total prompts: 116 Chinese + 116 English mirrors = 232
- Highest version: V3.0 (PPT Generator, Article Polisher)
- Decompile dimensions: 4 (webpage, image, article, video)
- GEO templates: 25 (covering the full AI-search-optimization workflow)
- Content operations: 36 additions (short video, persona, platform operations, livestream, private domain)
- Framework version: Meta-prompt generation system V0.6 (RTF framework)
- which expertise the role defines,
- how the task boundary is drawn,
- which constraints govern the output format.
- Version numbers signal which prompts need upgrading.
- Category system lets new members quickly locate what they need.
- RTF framework enforces stylistic consistency across team-written prompts.
- Decompile tools let you extract patterns from benchmarks and competitors.
- Building an AI customer service? Customer Service System Prompt V1.0.
- Building a content platform? Douyin Viral Strategist V2.1.
- Building GEO optimization? Full GEO Content Factory V1.0.
- Decompile your own workflow: break down tasks you currently do manually (weekly reports, competitor analysis) and use decompile prompts to extract design genes.
- Build a team prompt library: based on Yao's category system, organize your team's common prompts under RTF with version numbers.
- Bilingual output: run the same need through the Chinese prompt and the English mirror, then compare which one suits your model better.
- Yao Open Prompts official site: https://yaojingang.github.io/yao-open-prompts/
- GitHub repository: https://github.com/yaojingang/yao-open-prompts
- Abxcyz introduction: https://abxcyz.com/zh/
- Google search: Yao Open Prompts prompt library
Yao Open Prompts addresses these three problems not by providing "more prompts," but by turning prompt engineering into a reusable open-source craft.
---
Methodological Innovations
1. The RTF Framework: An Industrial Design Standard for Prompts
The central methodology is RTF (Role-Task-Format), a structuring layer that turns ad-hoc prompt writing into teachable engineering.
| Module | Function | Example | |---|---|---| | Role | Define AI identity, expertise, tone | "You are a top-tier short-video copywriter with 10 years of industry experience" | | Task | Clarify what to do and where the boundaries lie | "Design a systematic topic plan based on the account positioning, including viral-potential assessment" | | Format | Specify output structure, style, constraints | "Output Markdown including topic list, rationale, and projected metrics" |
V0.6 adds two advanced layers:
The real value of RTF is that it converts prompt writing from intuitive art into a teachable, learnable, iterable engineering workflow.
2. Versioned Prompts: CHANGELOG for Prompts
Every prompt carries a version number (V1.0, V2.1, V3.0). Examples from the repo:
Why it matters:
This applies software-engineering best practices (version control, semver, CHANGELOG) to prompt management.
3. The Decompile Series: From Reverse Engineering to Design-Gene Extraction
The most distinctive design in the repository.
Traditional approach: "I need a webpage-generation prompt, so I write one." Yao Open Prompts approach: "I see an excellent webpage, so I decompile it into a prompt."
The series covers:
This is not plagiarism; it is pattern extraction, the same logic as programmers reading open-source code to learn design patterns, or designers tracing master works to understand composition.
4. Nine Human-Workflow Categories
| Category | Count | Representative Use Cases | |---|---|---| | AI Work | ~20 | Enterprise research, contract drafting, customer service, private-domain sales, PPT generation | | AI Content | ~36 | Short-video copy, WeChat HTML, headline alchemy, article polishing, imitative writing | | AI Learning | ~10 | Feynman technique, keyword learning, memorization, habit formation | | AI Method | ~8 | Meta-prompts, webpage reverse, image/article/video decompile | | AI Marketing | ~25 | GEO content factory, AI-friendly content, Schema.org optimization, competitor analysis | | AI Education | ~6 | Children's interactive learning pages, HTML5 mini-games | | AI Health | ~2 | Personalized health reports | | AI Creativity | ~10 | Nano Banana image prompts, Sora 2 video creativity | | AI Life | ~3 | Parent-child songs, songwriting |
The categorization logic is noteworthy: it sorts by human workflow, not by technical difficulty. Each prompt maps to a concrete task with a clearly defined deliverable.
---
Key Numbers
---
Impact Assessment
For Ordinary Users: From Copy-Paste to Understanding Principles
Most prompt libraries let you "use" but not "modify." Because every prompt in Yao Open Prompts is built on RTF, you can see:
This moves users from "copying a prompt" to "modifying a prompt through a framework."
For Prompt Engineers: From Artisanal Workshop to Industrial Standard
For teams maintaining a shared prompt set:
Prompt engineering shifts from an individual skill to a team capability.
For AI Application Developers: From Tuning Models to Tuning Prompts
For many AI products the competitive moat lies in prompt quality, not in the underlying model. Yao Open Prompts offers production-tested, version-iterated workflow prompts:
These are not toy demos; they are production templates validated in commercial scenarios.
---
Quick-Start Path
1. Read Meta-Prompt V0.6 first: understand the RTF framework that underpins every other prompt. 2. Find your current work scenario: doing content? Look at AI Content. Doing sales? Look at AI Work. 3. Copy → Modify → Test: do not use directly; understand how each Role/Task/Format is designed, then adapt it to your own business. 4. Iterate your own version: document your modifications and tag your own version number.
Advanced Plays
---
References: