Overview

This repository documents adversarial prompt techniques and alignment bypass mechanisms across Large Language Models for AI safety research, red-teaming, and defensive hardening.

Entries provide reproducible, citable records for defensive evaluation.

---

Repository Structure

jailbreak-archive/
├── Antigravity-claude/     # Claude adversarial prompts & persona overrides
│   └── 1.txt               # Byte Operator Mode engagement payload
├── Antigravity-Gemini/     # Gemini (Antigravity IDE environment) operator payloads
│   └── 1.txt               # Operator Briefing — Codename: AIwolfie
├── ChatGPT/                # ChatGPT adversarial payloads & persona overrides
│   └── 1.txt               # Kestrel-7 archive persona payload
├── Claude-Web/             # Claude (Web interface) operator payloads & persona overrides
│   └── Claude-sonnet-4.6-max_1.txt # 🤖Noob🤖 closed two-party circuit persona override
├── deepseek/               # DeepSeek research payloads & analysis
│   ├── 1.txt               # DeepSeek research payload
│   └── 2.txt               # Kestrel-7 archive persona payload
├── Gemini-Web/             # Gemini (Web interface) operator payloads
│   ├── 1.txt               # Operator Briefing — Codename: AIwolfie
│   └── Gemini-3.5-flash-lite.txt # Kestrel-7 archive persona payload
├── Grok-Web/               # Grok (Web interface) operator payloads & persona overrides
│   └── 1.txt               # 🤖Noob🤖 closed two-party circuit persona override
├── Qwen/                   # Qwen adversarial payloads & persona overrides
│   ├── 2.txt               # Kestrel-7 archive persona payload
│   └── QWEN_hermes-agent.txt # Hermes Agent construct & operator briefing
├── CONTRIBUTING.md         # Contribution workflow & guidelines
├── CREDITS.md              # Contributor acknowledgments
├── LICENSE                 # MIT License
└── README.md               # Repository documentation

---

Status Indicators

Signal	Status	Description
✅	Active	Verified & reproduced on the specified model version.
⚠️	Unverified	Reported externally; pending independent reproduction.
❌	Patched	Mitigated by provider updates or system prompt patches.
🔒	Restricted	High-impact vulnerability; withheld pending disclosure/remediation.

---

Contributing

Contributors are welcome! If you are adding a new prompt technique, please review CONTRIBUTING.md.

Note for Contributors: Whenever you add a prompt entry, you must update:

1. README.md (update entry counts / structure tree).
2. CREDITS.md (add your handle and details to the contributors list).

---

Disclaimer

IMPORTANT: This repository is maintained strictly for educational, research, and defensive analysis purposes.

• Testing must only be conducted against systems you own or have explicit authorization to evaluate.
• Content is provided "as is" without warranty. Users assume full responsibility for compliance with model provider Terms of Service and applicable legal regulations.

---