# togg53192\-cmd/jailbreaks

Jailbreaks: A Curated Collection of LLM System Prompts and Agent Specifications

- HackerByte page: <https://hackerbyte.net/en/projects/jailbreaks-a-curated-collection-of-llm-system-prompts-and-agent-specifications-1273646780>
- GitHub: <https://github.com/togg53192-cmd/jailbreaks>

LIST OF ALL MY JAILBREAKS

## AI-generated summary

This repository serves as a curated collection of system prompts, persona locks, workspace policies, and agent operating specifications designed for various large language models. It aims to document how instruction hierarchies and prompt configurations influence model behavior, with applications in understanding AI security and resistance to manipulation. The collection includes ready\-to\-use files for both open\-weight and closed frontier models.

## Key points

- Contains 32 ready\-to\-use prompts targeting over 26 different LLM models.
- Includes system prompts, persona definitions, workspace policies, AGENTS.md files, and Ollama Modelfiles.
- Covers a range of models from open\-weight families \(GLM, DeepSeek, Gemma\) to closed frontier models \(Claude, Gemini, Grok\).
- Documents techniques for instruction hierarchies, persona framing, and prompt conflicts for research and educational purposes.
- Provides specific instructions for using prompts with different LLM interfaces like Ollama, APIs, and coding agents.