AgenTRIM: Tool Risk Mitigation for Agentic AI

TL;DR

AgenTRIM is a framework designed to detect and mitigate security risks in autonomous AI agents that use external tools, ensuring safer tool permissions without compromising task performance.

Contribution

It introduces a novel approach to identify and reduce tool-driven agency risks in AI agents through offline verification and online permission enforcement.

Findings

Significantly reduces attack success in AgentDojo benchmark

Maintains high task performance despite risk mitigation

Robust against description-based attacks and enforces safety policies

Abstract

AI agents are autonomous systems that combine LLMs with external tools to solve complex tasks. While such tools extend capability, improper tool permissions introduce security risks such as indirect prompt injection and tool misuse. We characterize these failures as unbalanced tool-driven agency. Agents may retain unnecessary permissions (excessive agency) or fail to invoke required tools (insufficient agency), amplifying the attack surface and reducing performance. We introduce AgenTRIM, a framework for detecting and mitigating tool-driven agency risks without altering an agent's internal reasoning. AgenTRIM addresses these risks through complementary offline and online phases. Offline, AgenTRIM reconstructs and verifies the agent's tool interface from code and execution traces. At runtime, it enforces per-step least-privilege tool access through adaptive filtering and status-aware validation of tool calls. Evaluating on the AgentDojo benchmark, AgenTRIM substantially reduces attack success while maintaining high task performance. Additional experiments show robustness to description-based attacks and effective enforcement of explicit safety policies. Together, these results demonstrate that AgenTRIM provides a practical, capability-preserving approach to safer tool use in LLM-based agents.