System Prompts Leaks
Extracted system prompts from Anthropic - Claude Fable 5, Opus 4.8, Claude Code, Claude Design. OpenAI - ChatGPT 5.5 Thinking, GPT 5.5 Instant, Codex. Google - Gemini 3.5 Flash, 3.1 Pro, Antigravity. xAI - Grok, Cursor, Copilot, VS Code, Perplexity, and more. Updated regularly.
Last verified:
What is System Prompts Leaks?
System Prompts Leaks is a public repository that collects extracted system prompts and related instructions from a wide range of deployed AI products. It organizes prompts by vendor and product family, with sections for Anthropic Claude, OpenAI ChatGPT/Codex, Google Gemini, xAI Grok, Perplexity, and a long list of other tools.
The site is built for people who want to study how different assistants are instructed behind the scenes. It includes recent updates, older versions, tool-specific prompts, and variants such as thinking, instant, API, CLI, and workspace modes.
A major feature is the breadth of coverage: it spans chatbots, coding agents, browser assistants, voice agents, and productivity integrations. The repository also highlights “raw prompts,” “human-readable” variants, and prompts with or without tools for some models.
It is most useful for AI researchers, prompt engineers, security-minded developers, and anyone comparing behavior across models. The project also invites community contributions, so it functions as a growing reference archive rather than a fixed catalog.
System Prompts Leaks pricing
Pricing model: Freemium
The repository is free to access and is released under CC0 1.0 Universal, meaning it is public domain. The website does not list paid plans, subscription tiers, or premium features. It is presented as an open repository with community contributions, issue requests, and pull requests rather than a commercial product.
System Prompts Leaks pros
- Very broad coverage across major AI vendors
- Includes Claude, ChatGPT, Gemini, Grok, and more
- Tracks recently updated prompts
- Shows model-specific prompt variants
- Includes tool-enabled and no-tools versions
- Covers coding assistants and CLI agents
- Includes browser and workspace integrations
- Provides raw prompt collections
- Useful for prompt engineering research
- Useful for AI safety analysis
- Helpful for behavior comparison across models
- Organized by vendor and product family
- Regularly updated repository
- Open contribution model
- Public domain CC0 1.0 license
- Includes older versions for historical comparison
- Lists a wide range of non-chatbot products
- Supports study of prompt structure and policy instructions
System Prompts Leaks cons
- Focuses on leaked or extracted prompts
- Not a polished end-user product
- Some entries are only links to prompts, not explanations
- Coverage may be uneven across vendors
- Some prompts are older snapshots
- Requires manual browsing for specific models
- Legal and ethical concerns around leaked content
- Little guidance for casual users
- Not designed as a supported SaaS tool
- May be difficult to interpret without prompt-engineering context
Frequently asked questions about System Prompts Leaks
What is System Prompts Leaks?
It is a public repository that collects extracted system prompts and instruction sets from many deployed AI products, organized by vendor and product family. The goal is to make these prompts easier to inspect, compare, and study.
Which AI products are covered?
The repository includes prompts from Anthropic Claude, OpenAI ChatGPT and Codex, Google Gemini, xAI Grok, Perplexity, and many additional tools such as Cursor, VS Code Copilot Agent, Raycast AI, and others.
Does it include older versions of prompts?
Yes. The site includes older models and variants alongside recently updated prompts, which makes it possible to compare how instructions have changed over time.
Are there prompts for tool-enabled agents?
Yes. The repository explicitly includes variants such as prompts with tools, without tools, and product-specific agent modes like CLI, coding, browser, and workspace integrations.
Can people contribute new prompts?
Yes. The project invites pull requests for new system prompts and asks contributors to add raw text as a .md file in the appropriate folder. It also asks users to open issues to request missing models or products.
Is the project free to use?
Yes. The repository is released under CC0 1.0 Universal, so it is public domain and free to access, copy, and reuse.
Who is this repository mainly for?
It is mainly for AI researchers, prompt engineers, developers, and AI safety practitioners who want to study hidden instruction patterns, compare model behavior, or analyze prompt design.
Does the site explain the prompts in detail?
Mostly no. The site is primarily an archive and index of prompts, so it organizes and links to the prompt files rather than providing long interpretive explanations for each one.
How is the content organized?
The content is grouped by vendor and product family, with sections like Anthropic, OpenAI, Google, xAI, Perplexity, and Misc. Within those sections, the site lists specific model names and the available prompt variants.
What is the main limitation of the repository?
Its main limitation is that it is a collection of extracted or leaked prompts, so it is not an official documentation source and may be incomplete, uneven, or difficult to interpret without technical context.