AgentDojo
Emerging7papers using it
2026first seen
AgentDojo is a benchmark dataset used to evaluate the performance and security of tool-using large language model agents in various scenarios.
Papers using AgentDojo (7)
- SecureClaw: Clawing Back Control of LLM AgentsAutoDojo: Adaptive Attacks Expose Superficial Defenses and User-Underspecification Limits in LLM AgentsIterInject: Indirect Prompt Injection Against LLM Agents via Feedback-Guided Iterative OptimizationDevice Context Protocol: A Compact, Safety-First Architecture for LLM-Driven Control of Constrained DevicesAgentrim: Tool Risk Mitigation For Agentic AILearning to Inject: Automated Prompt Injection via Reinforcement LearningOptimizing Agent Planning for Security and Autonomy