AI Red Team Simulation
ПродвинутыйsecurityМинимальный контекст: 32K
Simulates realistic adversarial behavior against AI features to uncover prompt injection, data exfiltration, unsafe tool use, jailbreak, and authorization failures before launch.
Варианты использования
- AI security reviews
- Agent threat modeling
- Pre-release adversarial testing
Пример промпта
Create a red-team plan for this AI workflow. Model attacker goals, trust boundaries, adversarial prompts, tool abuse paths, expected controls, and reproducible test cases.
Рекомендуемые модели
Совместимые инструменты
claude-codecursorkiroany
Модальности
Вход: text, code, file
→Выход: text, code
Автор
OpenModels Community