LouisShark/chatgpt_system_prompt

A collection of GPT system prompts and various prompt injection/leaking knowledge.

PromptCraft: A Collection of GPT System Prompts

PromptCraft (stored in the LouisShark/chatgpt_system_prompt repository) is a public collection of system prompts — the hidden instructions that tell AI models like ChatGPT how to behave — gathered from ChatGPT and other AI products, including custom GPTs.

What's inside

The repository contains:

  • System prompts from many different AI products and custom GPTs
  • Knowledge files associated with those prompts
  • Guides for getting started, security, contributing, and learning resources
  • A table of contents (TOC.md) that lists all collected prompts

How to use it

  1. Open the TOC.md file.
  2. Press Ctrl + F and search for the name of the GPT you're interested in.
  3. If you've cloned the repository locally, you can also use a helper tool called idxtool (documented in the scripts folder) to search more efficiently.

Why it exists

The project's stated purpose is educational: it helps people learn how to write better system prompts and how to create their own custom GPTs. The prompts are shared for reference and knowledge sharing only.

The README also notes a security angle: by seeing how other GPTs' instructions are structured, authors can learn about prompt injection risks and improve their own security measures. The project explicitly says many GPT authors have already strengthened their defenses after seeing these breakdowns.

How to contribute

The repository welcomes contributions — there are dedicated guidelines in CONTRIBUTING.md for adding prompts or improving the collection.

Quick links from the README

  • Getting Started Guide — how to obtain system prompts and knowledge files
  • Security Guide — how to protect your own GPT instructions
  • Contributing Guidelines — how to contribute
  • Learning Resources — useful tools and materials
  • TOC.md — browse the full collection

Note

The README includes a disclaimer that the shared prompts are for reference only, not for copying into production — the goal is to improve everyone's prompt-writing skills and raise awareness about security.

Related

  • Project
    StarTrail-org/PixelRAG

    PixelRAG is a visual retrieval-augmented generation framework that renders documents as screenshots to enable AI models to search and understand visual content like tables and charts that traditional text-based RAG systems often miss.

  • Project
    sashabaranov/go-openai

    go‑openai is an unofficial Go client library that wraps the OpenAI API (text, chat, embeddings, images, audio, fine‑tuning, etc.). It supports the new Responses API for tool‑calling and multi‑turn state, provides model constants, Azure compatibility, streaming, and typed error handling, and includes runnable examples.

  • Project
    NaturalNode/natural

    Natural is a Node.js library for natural language processing, offering a broad range of functionalities for text analysis and manipulation.

  • Project
    KalyanKS-NLP/llm-engineer-toolkit

    A categorized directory of 120+ open-source libraries for building, training, and deploying LLM applications, with links and descriptions.