Skip to main content
This guide will help you effectively navigate and utilize the AI System Prompts Collection to learn from production AI systems and improve your own implementations.

​
Collection Structure

The repository is organized by tool/platform, making it easy to find specific prompts you’re interested in: 
system-prompts-and-models-of-ai-tools/
├── Anthropic/
│   ├── Claude Code/
│   ├── Claude for Chrome/
│   └── Sonnet 4.5 Prompt.txt
├── Cursor Prompts/
│   ├── Agent Prompt v1.0.txt
│   ├── Agent Prompt v2.0.txt
│   └── Chat Prompt.txt
├── Devin AI/
├── Google/
├── Open Source prompts/
└── [30+ other tools...]
Each directory contains prompts, system instructions, and sometimes additional documentation specific to that tool or platform.
1

Choose Your Tool

Browse the main directory to find the AI tool or platform you’re interested in. Tools are organized alphabetically by name.
2

Explore Versions

Many tools have multiple prompt versions as they evolved. Look for version numbers or dates in filenames to understand the chronology:
  • Agent Prompt v1.0.txt
  • Agent Prompt v2.0.txt
  • Agent Prompt 2025-09-03.txt
3

Read the Prompts

Open the .txt files to view the full system prompts. These contain:
  • System instructions and personality
  • Tool usage patterns
  • Safety constraints and guidelines
  • Response formatting rules
4

Check READMEs

Some directories include README.md files with additional context about how the prompts were obtained or specific usage notes.

​
Finding Specific Content

​
By Tool Type

The collection covers several categories of AI tools:
  • Cursor Prompts - AI-powered code editor
  • Augment Code - Code completion and generation
  • VSCode Agent - VS Code AI extensions
  • Replit - Browser-based coding AI
  • Windsurf - Code navigation assistant

​
By Use Case

Depending on what you’re building, focus on different tools:

Code Generation

Study prompts from Cursor, Augment Code, Devin AI, and Replit to understand how to guide AI for code generation tasks.

Chat Interfaces

Examine Anthropic Claude, Perplexity, and NotionAI prompts for conversational AI patterns.

UI Generation

Check out v0, Lovable, and similar tools for visual design and component generation.

Agent Systems

Learn from Devin AI, Manus Agent, and Trae for building autonomous agent systems.

​
Search Tips

The repository contains over 30,000 lines of prompts. Here’s how to search effectively: If you’re viewing the repository on GitHub:
  1. Use the “Go to file” feature (press t) to quickly find files
  2. Use GitHub’s code search to find specific patterns or keywords
  3. Clone the repository locally for more advanced searching

​
Local Searching

If you’ve cloned the repository: 
# Search for specific keywords
grep -r "safety" .

# Find all prompts mentioning tools
grep -r "use.*tool" *.txt

# Search within a specific tool's directory
grep -r "system:" "Cursor Prompts/"

​
Understanding Prompt Structure

Most AI system prompts in this collection follow similar patterns:
The opening section typically establishes:
  • Who/what the AI is
  • Its primary purpose
  • Its personality or tone
Example: “You are an AI coding assistant specialized in Python…”
Instructions about available functions and how to use them:
  • Tool definitions
  • When to use each tool
  • Expected input/output formats
Rules the AI must follow:
  • Safety constraints
  • Output formatting requirements
  • What to avoid or refuse
Many prompts include example interactions to demonstrate expected behavior.

​
Comparing Versions

Many tools have multiple prompt versions. Comparing them reveals:
  • Evolution of capabilities - What features were added?
  • Refined instructions - How did clarity improve?
  • Safety improvements - What guardrails were strengthened?
  • Performance optimization - How were prompts made more efficient?
Use a diff tool to compare versions side-by-side:
diff "Cursor Prompts/Agent Prompt v1.0.txt" "Cursor Prompts/Agent Prompt v2.0.txt"

​
Best Practices

When using this collection:

Learn, Don't Copy

Study the patterns and principles rather than copying prompts verbatim. Adapt insights to your specific needs.

Respect Licenses

While this collection is GPL-3.0 licensed, be mindful that the original prompts may have different intellectual property considerations.

Test Thoroughly

If you adapt prompts for your own use, test extensively. What works for one model or context may not work for another.

Stay Updated

The collection is regularly updated. Star the repository and check back for new additions and versions.

​
Getting Help

If you have questions about using the collection:

Next: Start Contributing

Found prompts to add or improvements to make? Learn how to contribute.

Build docs developers (and LLMs) love

Get started for freeTalk to us