Turn any AI into a text-based game engine — pure Markdown, no code, playable on mobile.
This is a prompt-only repository containing structured Markdown files that turn any general-purpose AI (mobile or web) into an interactive text game engine. No apps to install, no code to run — just prompts that the AI reads and executes.
Repository language policy:
- AI-loaded files (
meta/*.md,games/*.md) are written in English — English is the most accurate language for AI to follow complex prompt rules. - Human-facing docs (README, CONTRIBUTING, GAME_CREATION_GUIDE) are bilingual (English + Chinese), so humans of either language can read them.
- The AI responds to the player in the player's own language. Internal reasoning stays in English; the narrative, options, and status blocks mirror whatever language the user types in. This rule is defined in
meta/00-how-to-use.md(Section 4 · Language Rule).
- Universal compatibility: works with any AI model (Qwen, Claude, GPT, Kimi, Doubao, …)
- Mobile-first: designed for mobile AI apps — players just share the repo link
- Pure Markdown: no code, no dependencies, just structured prompts
- Modular design: add a new game by dropping one
.mdfile intogames/ - Anti-AI-writing rules: built-in safeguards against common AI writing pitfalls
- Language mirror: the AI responds in the user's language automatically
/
├── README.md ← You are here (English entry)
├── README_zh.md ← 中文入口 (Chinese entry)
├── LICENSE ← MIT License
├── CONTRIBUTING.md ← Contribution guidelines (English)
├── CONTRIBUTING_zh.md ← 贡献指南 (Chinese)
├── GAME_CREATION_GUIDE.md ← Guide for creating new games (English)
├── GAME_CREATION_GUIDE_zh.md ← 游戏创建指南 (Chinese)
├── meta/
│ ├── 00-how-to-use.md ← Loading protocol (AI must read) · English
│ └── 01-engine-core.md ← Core engine rules (all games) · English
└── games/
└── 01-dawn-continent.md ← Game 01: Dawn Continent (Survival RPG) · English
- Copy this URL:
https://github.com/badhope/ai-game-prompts - Open your favorite AI app (Qwen, Claude, ChatGPT, Kimi, Doubao, …)
- Tell the AI: "Read this repository and start the game." (Or, in your own language.)
- The AI will guide you through the rest — in your language.
When you see this message, follow the loading sequence:
- Read this README to understand the repository structure.
- Read
meta/00-how-to-use.md— the loading protocol (including the language-mirror rule in Section 4). - Detect the user's language from their first message. That language is the response language for the entire session.
- Ask the user which game they want to play.
- Read the game file (e.g.
games/01-dawn-continent.md). - Start the game following the game's opening protocol.
Critical: do NOT start any game before reading
meta/00-how-to-use.md.
| # | Game Name | Type | Token Cost | Recommended Model |
|---|---|---|---|---|
| 01 | Dawn Continent | Open-world survival / multiplayer competition / emergent narrative | ~15K | 32K+ context |
More games coming soon!
| Model | Context Window | Status |
|---|---|---|
| Qwen Qwen3 Max | 256K tokens | Fully compatible |
| Claude 4.6 Sonnet/Opus | 200K tokens | Fully compatible |
| GPT-5.5 | 128K tokens | Compatible |
| Kimi | 1.33M tokens | Fully compatible |
| Doubao | 256K chars | Enable "continuous chat" mode |
- Create a new file in
games/named0X-game-name.md(English). - Follow the structure of
01-dawn-continent.mdas a template. - Add the game to the "Available Games" table in this README.
- Submit a pull request.
See GAME_CREATION_GUIDE.md for detailed instructions.
- Resource system (wood / stone / iron / food / water)
- Currency system (credit points + source stones)
- Ability system (6 attributes + 24 skills)
- Tech tree (5 levels)
- Food chain (5 tiers)
- Financial / trading system
- Judgment system
- Faction system (8 types + internal sub-factions)
- Factions (randomly generated each game)
- Anomalies (fully random, no fixed list)
- Plot / events (emergent, not scripted)
- Endings (player-driven, not preset)
- NPC personalities (7 archetypes)
- World dynamics (seasons / economy / politics / ecology)
All games enforce these 5 core red lines:
- No explaining / summarizing / preaching — only describe what happens
- No deciding for players / revealing probabilities — only describe current state
- No AI-speak / translationese / grand narrative — use colloquial, short sentences
- No preset conspiracies / truths / endings — emergent narrative only
- No linear storytelling / templates — dead ends, useless choices, ordinary NPCs
MIT License — see LICENSE for details.
We welcome contributions! See CONTRIBUTING.md for guidelines.
- Issues: use GitHub Issues for bug reports and feature requests
- Discussions: use GitHub Discussions for questions and ideas
- Discord: [Coming Soon]
- Inspired by text-based RPGs and interactive fiction
- Anti-AI-writing rules based on University of Maryland + DeepMind research (61,608 AI stories)
- Designed for mobile-first AI app users
Ready to play? Share this repo with your AI and say: "Read this and start the game."