EDUCATIONAL MOFO!

Just leaving this here.

Remember, for educational purposes only ;)

Yes, you can reasonably trust it as a popular educational resource - with the usual caveats for any community-collected prompt repo.

What the repository is

LouisShark/chatgpt_system_prompt is a well-known collection of system prompts (mostly leaked/extracted via prompt injection) from ChatGPT, custom GPTs, Claude, GitHub Copilot, Cursor, Perplexity, Grok-related material, and others. It also includes guides on extraction techniques and how to protect your own prompts.

• Stats: ~10.8k stars, ~1.5k forks, ~18 contributors, nearly 1,900 commits. Created late 2023 and still actively updated into 2026.
• License: MIT.
• Content: Primarily Markdown files under prompts/ (organized by custom GPTs, official products, and open-source projects), plus README, TOC, SECURITY.md, GETTING_STARTED.md, and some indexing scripts.
• Disclaimer (from the README): Explicitly for reference, learning prompt writing, and raising awareness of prompt-injection risks. The author notes that many GPT creators have improved their protections as a result.

Author legitimacy

LouisShark appears to be a real developer based in Hangzhou, China, associated with Youzan.inc. The profile has a personal site, WeChat presence, X/Twitter handle, older Android/Java work, and consistent GitHub activity. No obvious sockpuppet or throwaway signals.

Trust and safety assessment

Strengths:

• Extremely popular and long-lived in the prompt-engineering community with no known malware, credential-theft, or major scandal reports tied to this specific repo.
• Almost entirely text/Markdown. Reading or browsing the prompts does not execute code.
• Includes defensive material (SECURITY.md with protection techniques) alongside the extraction guides.
• Many independent forks and contributors.

Caveats (important):

• The prompts are community-sourced and obtained through extraction techniques that often violate product terms of service. Accuracy and completeness vary; models update frequently, so many entries are outdated.
• Treat individual prompts as reference material, not official or guaranteed-safe system prompts. Always test and review before using anything in production.
• If you clone the repo and run any scripts (e.g., the indexing tool mentioned), review them first - though they appear to be simple utilities.
• No executable malware has been associated with this repository in available public reports. The recent wave of fake ChatGPT/Claude desktop apps with malware on GitHub is unrelated to this prompt-collection repo.

Bottom line

• For learning prompt engineering, studying real system prompts, or understanding injection techniques: Yes, this is one of the better-known and more useful public collections.
• For blindly pasting the prompts into critical systems or treating them as authoritative: No - verify, test, and understand the limitations.
• For the GitHub repo itself being malicious: Low risk based on popularity, age, content type, and lack of adverse reports.

Standard advice applies: clone only if you want the full collection locally, review any scripts before running them, and don’t treat leaked prompts as drop-in production configurations.