paperJanuary 2026Unreviewed

AgenTRIM: Tool Risk Mitigation for Agentic AI

Roy Betser, Shamik Bose, Amit Giloni, Chiara Picardi, Sindhu Padakandla, R. Vainshtein

arXiv.org

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 fram

Framework mappings

OWASP Top 10 for LLM Applications

LLM01Prompt Injection

MITRE ATLAS

AML.T0051LLM Prompt Injection

Suggested from the entry's categories.

Cite

@misc{betser2026agentrim,
  title = {{AgenTRIM: Tool Risk Mitigation for Agentic AI}},
  author = {Roy Betser and Shamik Bose and Amit Giloni and Chiara Picardi and Sindhu Padakandla and R. Vainshtein},
  year = {2026},
  month = jan,
  eprint = {2601.12449},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2601.12449},
  url = {https://www.semanticscholar.org/paper/01faaf29f224194719d37bcb522264fdcc07bc24}
}