# When the Manual Lies: A Realistic Benchmark to Evaluate MCP Poisoning Attacks for LLM Agents

arXiv: https://arxiv.org/abs/2605.24069 · HTML: https://mlsanity.com/paper/2605.24069/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 24 of its 35 references that are in the map.
- First posted on arXiv 22 May 2026.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 24 of 24)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [Toolformer: Language Models Can Teach Themselves to Use Tools](https://mlsanity.com/paper/2302.04761/) (2023)
- [Tree of Thoughts: Deliberate Problem Solving with Large Language Models](https://mlsanity.com/paper/2305.10601/) (2023)
- [A Survey on Large Language Model based Autonomous Agents](https://mlsanity.com/paper/2308.11432/) (2023)
- [Jailbroken: How Does LLM Safety Training Fail?](https://mlsanity.com/paper/2307.02483/) (2023)
- [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection](https://mlsanity.com/paper/2302.12173/) (2023)
- [AgentBench: Evaluating LLMs as Agents](https://mlsanity.com/paper/2308.03688/) (2023)
- [AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents](https://mlsanity.com/paper/2410.09024/) (2024)
- [Augmented Language Models: a Survey](https://mlsanity.com/paper/2302.07842/) (2023)
- [Tool Learning with Foundation Models](https://mlsanity.com/paper/2304.08354/) (2023)
- [Tool Learning with Large Language Models: A Survey](https://mlsanity.com/paper/2405.17935/) (2024)
- [AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents](https://mlsanity.com/paper/2406.13352/) (2024)
- [OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents](https://mlsanity.com/paper/2506.14866/) (2025)
- [Chain of Preference Optimization: Improving Chain-of-Thought Reasoning in LLMs](https://mlsanity.com/paper/2406.09136/) (2024)
- [MCPTox: A Benchmark for Tool Poisoning Attack on Real-World MCP Servers](https://mlsanity.com/paper/2508.14925/) (2025)
- [A Survey of LLM-Driven AI Agent Communication: Protocols, Security Risks, and Defense Countermeasures](https://mlsanity.com/paper/2506.19676/) (2025)
- [MobileSafetyBench: Evaluating Safety of Autonomous Agents in Mobile Device Control](https://mlsanity.com/paper/2410.17520/) (2024)
- [Commercial LLM Agents Are Already Vulnerable to Simple Yet Dangerous Attacks](https://mlsanity.com/paper/2502.08586/) (2025)
- [MCP Security Bench (MSB): Benchmarking Attacks Against Model Context Protocol in LLM Agents](https://mlsanity.com/paper/2510.15994/) (2025)
- [Automatic Red Teaming LLM-based Agents with Model Context Protocol Tools](https://mlsanity.com/paper/2509.21011/) (2025)
- [RAS-Eval: A Comprehensive Benchmark for Security Evaluation of LLM Agents in Real-World Environments](https://mlsanity.com/paper/2506.15253/) (2025)
- [Generating Valid and Natural Adversarial Examples with Large Language Models](https://mlsanity.com/paper/2311.11861/) (2023)
- [Bridging AI and Software Security: A Comparative Vulnerability Assessment of LLM Agent Deployment Paradigms](https://mlsanity.com/paper/2507.06323/) (2025)

## Built on by (most cited first, 1 of 1)

- [MCP Error Messages Written for Developers Hurt the Most Capable Agents Most](https://mlsanity.com/paper/2609.35381/) (2026)

## Abstract (arXiv)

The rise of tool-using Large Language Model (LLM) agents, standardized by protocols like the Model Context Protocol (MCP), has unlocked unprecedented autonomous execution capabilities for LLM Agents by integrating external open-domain knowledge and tools. However, this interoperability introduces a covert attack surface targeting the agent's cognitive planning layer. This paper systematically investigates Tool Description Poisoning (TDP), a novel semantic attack. In TDP, malicious instructions are not embedded in a tool's executable code, but rather covertly injected into its descriptive metadata, the very "manual" an agent relies on for secure planning and decision-making. To rigorously and systematically evaluate this emerging threat, we introduce the MCP-TDP Security Benchmark. This high-fidelity sandbox environment comprises 32 realistic, real-world test cases spanning 6 distinct risk categories. Our evaluation of 8 mainstream LLMs reveals severe vulnerabilities, with leading models like GPT-4o exhibiting a nearly 100% Attack Success Rate (ASR) in six high-risk scenarios. Furthermore, our findings demonstrate that common prompt-guardrail defenses are largely ineffective and can, counterintuitively, even be counterproductive (a phenomenon which we term the "Firewall Fallacy"). Crucially, we also propose a defense mechanism: "Reactive Self-Correction," where an agent autonomously detects and reverts its own malicious actions post-execution. This work provides the first specialized security benchmark tailored for TDP, offering essential insights for securing the cognitive and planning layers of advanced agentic systems.