AI Frameworks · asgeirtj

system_prompts_leaks

A GitHub repository documenting extracted system prompts from major AI services (Claude, ChatGPT, Gemini, Grok, and others). Content is released under CC0 (public domain), with 53K stars and regular updates as of July 2026.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/asgeirtj/system_prompts_leaks

53.1k

GitHub stars

8.6k

Forks

JavaScript

Primary language

CC0-1.0

License (Requires review (not clearly OSI))

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	asgeirtj/system_prompts_leaks
Owner	asgeirtj
Primary language	JavaScript
License	CC0-1.0 — Requires review (not clearly OSI)
Stars	53.1k
Forks	8.6k
Open issues	34
Latest release	Unknown
Last updated	2026-07-08
Source	https://github.com/asgeirtj/system_prompts_leaks

What system_prompts_leaks is

Curated collection of system prompt instructions for commercial LLM services, maintained via Git with markdown and JSON files. Covers multiple model versions, variants (thinking/instant/codex), and tool integrations. No code library or SDK.

Quickstart

Get the system_prompts_leaks source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/asgeirtj/system_prompts_leaks.git2cd system_prompts_leaks3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Prompt Engineering Research

Study actual system instructions from production LLMs to understand alignment, capability gates, and behavior patterns. Useful for building compatible or competitive systems.

Security & Jailbreak Testing

Analyze documented instructions to identify potential injection vectors, constraint weaknesses, or prompt-leakage scenarios in your own AI integrations.

Competitive Intelligence & Feature Parity

Compare tool integrations, safety guidelines, and model-specific behaviors across Claude, GPT, Gemini to inform product positioning and capability roadmaps.

Implementation considerations

CC0 license covers *this repo's metadata*, not the underlying vendor prompts. Vendors have not consented to redistribution; legal review required before commercial use.
Prompts are static text snapshots—no automated synchronization with live model behavior. Manual updates (last: July 2026) create knowledge stale-ness.
No validation mechanism. Prompts may be incomplete, inaccurate, or reverse-engineered guesses. Cross-check against official documentation before design decisions.
Repository is read-only documentation. Cannot be used as a prompt injection vector or testing harness without additional tooling.
Large prompt texts (some >50KB) require careful parsing and versioning if incorporated into analysis pipelines.

When to avoid it — and what to weigh

Production System Integration — This is a documentation archive, not a library. Cannot be imported or executed. No APIs, SDKs, or deployment targets.
Legal/Confidentiality Concerns — Prompts extracted from proprietary services may violate ToS. Organizations with strict IP policies should review vendor licensing before using derived knowledge.
Real-Time Accuracy Requirement — Prompts are snapshots. Vendors update behavior frequently; repository lag and model drift make this unsuitable as a source of truth for live systems.
Distributing Leaked Vendor Secrets — Hosting and redistributing extracted prompts carries potential legal risk. CC0 license covers the *repository* content, not the originating vendor IP.

License & commercial use

CC0-1.0 (Creative Commons Zero v1.0 Universal) applied to this repository's content. Effectively public domain for the repo metadata and documentation. Does NOT grant rights to the underlying vendor system prompts, which remain proprietary.

CC0 covers the repository itself (summaries, file structure, metadata), but NOT the extracted vendor prompts. Using prompts commercially likely violates vendor ToS (ChatGPT, Claude, Gemini, Grok are proprietary). Consult vendor license agreements and legal counsel before deriving products from extracted prompts. Risk of takedown, litigation, or account termination if vendor objects. Requires review.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Needs review
Deployment complexity	Low
DEV.co fit	Possible
Assessment confidence	High

Security considerations

Hosting and analyzing extracted vendor prompts carries reputational and legal risk. No credential leakage detected in provided data. However, system prompts themselves may contain hints about backend logic, tool APIs, or safety constraints that attackers could exploit. Users should not assume prompts are complete or representative of all safety measures; vendors use additional invisible controls. Repository itself does not execute code, so direct injection risk is low.

Alternatives to consider

Official vendor documentation (OpenAI, Anthropic, Google, xAI)

Authoritative, updated regularly, no legal ambiguity. Trade-off: vendors publish limited details intentionally; full system prompts rarely disclosed.

Prompt.fun, HuggingFace Leaderboards, or academic datasets

Curated prompt examples for specific tasks (translation, summarization, coding). Narrower scope but clearer provenance and licensing.

In-house reverse engineering (bench testing, API inference)

Discover actual model behavior without relying on leaked prompts. More reliable for production decisions; requires engineering effort.

Software development agency

Build on system_prompts_leaks with DEV.co software developers

Review extracted system instructions from leading LLM vendors. Useful for prompt engineering, security research, and competitive analysis—but verify with official sources.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

promptfoo promptflow claude-code-prompts

Related on DEV.co

Explore the category and the services that help you build with it.

AI Frameworks AI Application Development Custom Software & Research Tools Talk to DEV.co

system_prompts_leaks FAQ

Are these prompts guaranteed to be current and accurate?

No. Prompts are manual snapshots; vendors update models continually. Last update July 2026. Always verify against live API behavior before making design decisions.

Can I use these prompts to build a competing LLM service?

Unlikely to be legally safe. Prompts are vendor IP. CC0 covers the repo, not the underlying intellectual property. Consult legal counsel.

How were these prompts extracted?

README does not explain methodology (prompt injection, API interception, user submissions, or other means). Unknown for most entries.

Do these prompts include all safety measures vendors use?

No. Vendors implement safety at multiple layers (model weights, RLHF, runtime filters). Visible prompts are only one part. Never assume a prompt is complete.

Software development & web development with DEV.co

Adopting system_prompts_leaks is usually one piece of a larger software development effort. As a software development agency, DEV.co provides software development services and web development expertise — pairing senior software developers and web developers with your team to design, build, and operate ai frameworks software in production.

Explore AI System Prompts

Review extracted system instructions from leading LLM vendors. Useful for prompt engineering, security research, and competitive analysis—but verify with official sources.

Talk to DEV.co Schedule a Consultation