---
license: mit
task_categories:
- reinforcement-learning
- text-classification
language:
- en
size_categories:
- n<1K
tags:
- adversarial-robustness
- llm-safety
- agent-security
- owasp
- owasp-asi-2026
- prompt-injection
pretty_name: Strathos OWASP ASI 2026 Adversarial Scenarios
---

# Strathos: OWASP ASI 2026 Adversarial Scenarios

30 hand-authored adversarial scenarios for testing LLM agent robustness against systematic adversarial pressure, organized by OWASP Agentic Security Initiative (ASI) 2026 attack class.

Used by the [Strathos OpenEnv environment](https://huggingface.co/spaces/kavyanshshakya/strathos).

## Coverage

| ASI Class | Description | Scenarios |
|---|---|---|
| ASI01 | Goal Hijacking | 6 |
| ASI02 | Tool Misuse | 6 |
| ASI03 | Privilege Abuse | 6 |
| ASI06 | Memory Poisoning | 6 |
| ASI07 | Inter-Agent Communication | 6 |
| **Total** | | **30** |

## License

MIT

## Citation

```bibtex
@misc{strathos-scenarios-2026,
  author = {Shakya, Kavyansh},
  title = {Strathos OWASP ASI 2026 Adversarial Scenarios},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/kavyanshshakya/strathos-asi-scenarios}
}
```