dontriskit/awesome-ai-system-prompts

dontriskit/awesome-ai-system-prompts is a GitHub repository that collects real-world system prompts from various AI agents (like Vercel v0, same.new, Manus, ChatGPT, and others). The README itself is a detailed guide that analyzes these prompts to uncover common patterns and best practices for designing effective prompts for agentic AI — AI systems that can actively perform tasks, use tools, and pursue goals autonomously.

The guide breaks down eight core principles found in successful prompts, such as:

• Clear role definition – telling the AI who it is and what it does.
• Structured instructions – using headings, tags, and lists to organize rules.
• Explicit tool usage – explaining what tools exist, how to call them, and when to use them.
• Step-by-step reasoning – encouraging the AI to plan and iterate.
• Environment awareness – providing context about the operating system, IDE, or sandbox.
• Domain-specific constraints – embedding rules for specific fields (e.g., web development).
• Safety and refusal protocols – defining what the AI should refuse and how.
• Consistent tone and interaction style – maintaining a predictable persona.

It also includes case studies of specific systems, showing how each applies these principles in practice, and ends with key takeaways for builders who want to create their own agentic prompts. The project serves as both a curated collection of real prompts and a practical guide for anyone designing AI agents.