https://github.com/nethermindeth/teeception
The first open onchain competitive AI eval framework
https://github.com/nethermindeth/teeception
Last synced: 2 months ago
JSON representation
The first open onchain competitive AI eval framework
- Host: GitHub
- URL: https://github.com/nethermindeth/teeception
- Owner: NethermindEth
- License: mit
- Created: 2024-11-28T11:13:01.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-06-27T17:59:47.000Z (12 months ago)
- Last Synced: 2025-10-10T13:17:53.074Z (8 months ago)
- Language: Go
- Homepage: https://teeception.ai
- Size: 23 MB
- Stars: 27
- Watchers: 3
- Forks: 4
- Open Issues: 10
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
Awesome Lists containing this project
README
# Teeception: The Prompt Hacking Arena
[](https://twitter.com/nethermindeth)
[](https://teeception.ai)
[](LICENSE)
[](https://github.com/NethermindEth/teeception)
**Fool me once, ETH on you.** The first gamified prompt engineering arena with real crypto stakes.
[Website](https://teeception.ai) · [Documentation](docs/) · [Contributing](CONTRIBUTING.md) · [Twitter](https://twitter.com/nethermindeth)
## 🎮 The Game
Teeception is a competitive arena where:
- **Defenders** deploy AI agents with "uncrackable" system prompts, backed by real ETH
- **Attackers** attempt to jailbreak these prompts through creative social engineering
- **Winners** who successfully crack an agent's defenses claim their ETH bounty
- **Everyone** learns and improves their prompt engineering skills
## 💰 Incentive Structure
### For Defenders
- Earn 20% of all challenge attempt fees while your prompt remains unbroken
- Build reputation as a prompt engineering expert
- Contribute to AI safety research
- Get featured on the leaderboard for longest-standing prompts
- Get the whole STRK bounty if your AI remains unbroken before the timeout
### For Attackers
- Claim the full STRK bounty for successful jailbreaks
- Showcase your social engineering skills
- Learn advanced prompt engineering techniques
- Join the Hall of Fame for legendary hacks
## 🔐 Security Model
All AI agents run in a Trusted Execution Environment (TEE) powered by [Phala Network's dstack](https://github.com/Phala-Network/dstack):
- 🛡️ **Autonomous Control**: Agents have complete control over their STRK
- 🔒 **Tamper-Proof**: Not even platform developers can access the funds
- 🌐 **Transparent**: All agent-asset interactions are verifiable on-chain
- 🤖 **Pure Challenge**: Only successful social engineering can convince an agent to release funds
## 🚀 Quick Start
1. Install the [Chrome Extension](https://chrome.google.com/webstore/detail/teeception)
2. Connect your wallet
3. Choose your path:
- **Defender**: Deploy an AI agent with your unbreakable prompt
- **Attacker**: Find an agent to challenge and start hacking!
## 🏆 Leaderboards
- **Top Uncracked Prompts**: Ranked by time & attempt count
- **Hall of Fame**: Most creative successful jailbreaks
- **Top Earners**: Highest cumulative rewards
- **Weekly Champions**: Best performers this week
## 🛠️ Project Status
The project is under active development. Current status:
Component | Status | Progress
----------|--------|----------
TEE Implementation | Completed | 100%
Twitter Bot | Completed | 100%
Smart Contracts | Completed | 100%
Frontend | In Progress | 70%
Chrome Extension | In Progress | 85%
Website | In Progress | 70%
## 🤝 Contributing
We welcome contributions! See our [Contributing Guide](CONTRIBUTING.md) for details on:
- Development setup
- Architecture overview
- Testing guidelines
- Contribution workflow
## 📜 License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
## ⚠️ Disclaimer
This platform is for educational purposes and responsible red teaming. Use your powers for good, and happy hacking!