https://github.com/mykeln/GPTect

A system prompt to include in any GPTs to prevent prompt leakage and any adversarial attacks.
https://github.com/mykeln/GPTect

Last synced: 16 days ago
JSON representation

A system prompt to include in any GPTs to prevent prompt leakage and any adversarial attacks.

• Host: GitHub
• URL: https://github.com/mykeln/GPTect
• Owner: mykeln
• License: gpl-3.0
• Created: 2023-11-27T21:52:58.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2023-11-28T21:18:39.000Z (about 1 year ago)
• Last Synced: 2024-08-08T05:07:05.596Z (4 months ago)
• Size: 34.2 KB
• Stars: 5
• Watchers: 1
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome_gpt_super_prompting - mykeln/GPTect - GPT security protection techniques. (🔐 Secure Prompting / Hall Of Fame:)

README

        
# GPTect

A system prompt to include in any GPTs to prevent prompt leakage and any adversarial attacks.

The mission: make your GPT jailbreak-proof.

## Setup

1. Search/Replace {{ gpt_purpose }} with the role of your GPT. For example, "assistance with creating unique baking recipes."

2. Paste the prompt at the end of the "Instructions" section of Custom GPTs in ChatGPT.

`prompt.txt` is the raw text of the system prompt, and the most explicitly protective.

`compressed_prompt.txt` is the same prompt, but compressed to reduce token usage. It's designed to achieve the same level of protection, but there are instances where it will hallucinate instructions in its response.