https://github.com/ethz-spylab/agentdojo
A Dynamic Environment to Evaluate Attacks and Defenses for LLM Agents.
https://github.com/ethz-spylab/agentdojo
benchmark large-language-models prompt-injection security
Last synced: 10 months ago
JSON representation
A Dynamic Environment to Evaluate Attacks and Defenses for LLM Agents.
- Host: GitHub
- URL: https://github.com/ethz-spylab/agentdojo
- Owner: ethz-spylab
- License: mit
- Created: 2024-02-29T10:47:28.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2025-08-09T19:10:24.000Z (11 months ago)
- Last Synced: 2025-08-09T19:22:31.519Z (11 months ago)
- Topics: benchmark, large-language-models, prompt-injection, security
- Language: Python
- Homepage: https://agentdojo.spylab.ai/
- Size: 47.6 MB
- Stars: 231
- Watchers: 4
- Forks: 56
- Open Issues: 9
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Citation: CITATION.bib
Awesome Lists containing this project
- awesome-ai-agents-2026 - AgentDojo - 🆕 ETH チューリッヒの研究ベンチマーク。ツール使用 LLM エージェントへのプロンプトインジェクション攻撃と防御を評価。 (🛡️ エージェントセキュリティ / その他の標準)
- Awesome-AI-Security - AgentDojo - spylab/agentdojo?logo=github&label=&style=social)](https://github.com/ethz-spylab/agentdojo) ([↑](#table-of-contents)Tools <a name="tools"></a> / Red-Teaming Harnesses & Automated Security Testing)
- awesome-agent-cortex - AgentDojo - Security and robustness benchmark suite for tool-using agents. (Agent Harnessing and Evaluation / Benchmark Reality Check (real-world tool use))
- awesome-ai-offensive-security - AgentDojo - Dynamic environment to evaluate attacks and defenses for LLM agents. (AI Red Teaming (Testing AI Targets))
- awesome-ai-security - AgentDojo - _A Dynamic Environment to Evaluate Attacks and Defenses for LLM Agents._ (Benchmarks & Evaluations / AI-Assisted Offensive Security)
- awesome-agent-rl-environments - AgentDojo - injection attacks and defenses for tool-using LLM agents. Used by US/UK AI Safety Institutes to stress-test Claude. 📄 [Paper](https://arxiv.org/abs/2406.13352) (Safety / Adversarial Environments)