# The System Prompt Is the Attack Surface: How LLM Agent Configuration Shapes Security and Creates Exploitable Vulnerabilities

arXiv: https://arxiv.org/abs/2603.25056 · HTML: https://mlsanity.com/paper/2603.25056/

## Key facts

- Cited by 2 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 10 of its 17 references that are in the map.
- First posted on arXiv 26 Mar 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, 10 of 10)

- [A Survey on Large Language Model based Autonomous Agents](https://mlsanity.com/paper/2308.11432/) (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)
- [HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal](https://mlsanity.com/paper/2402.04249/) (2024)
- [JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models](https://mlsanity.com/paper/2404.01318/) (2024)
- [Personal LLM Agents: Insights and Survey about the Capability, Efficiency and Security](https://mlsanity.com/paper/2401.05459/) (2024)
- [Jailbreak Attacks and Defenses Against Large Language Models: A Survey](https://mlsanity.com/paper/2407.04295/) (2024)
- [Fraud-R1 : A Multi-Round Benchmark for Assessing the Robustness of LLM Against Augmented Fraud and Phishing Inducements](https://mlsanity.com/paper/2502.12904/) (2025)
- [Security Concerns for Large Language Models: A Survey](https://mlsanity.com/paper/2505.18889/) (2025)
- [DetoxBench: Benchmarking Large Language Models for Multitask Fraud & Abuse Detection](https://mlsanity.com/paper/2409.06072/) (2024)
- [SpearBot: Leveraging Large Language Models in a Generative-Critique Framework for Spear-Phishing Email Generation](https://mlsanity.com/paper/2412.11109/) (2024)

## Built on by (most cited first, 2 of 2)

- [Beyond the Payload: How User Invocation Shapes Coding Agent Vulnerability to Repository Poisoning](https://mlsanity.com/paper/2608.30686/) (2026)
- [ArrivalBench: Agent-Generated Data Pipelines Are Correct Once and Wrong Under Time](https://mlsanity.com/paper/2610.02363/) (2026)

## Abstract (arXiv)

System prompt configuration can make the difference between near-total phishing blindness and near-perfect detection in LLM email agents. We present PhishNChips, a study of 11 models under 10 prompt strategies, showing that prompt-model interaction is a first-order security variable: a single model's phishing bypass rate ranges from under 1% to 97% depending on how it is configured, while the false-positive cost of the same prompt varies sharply across models. We then show that optimizing prompts around highly predictive signals can improve benchmark performance, reaching up to 93.7% recall at 3.8% false positive rate, but also creates a brittle attack surface. In particular, domain-matching strategies perform well when legitimate emails mostly have matched sender and URL domains, yet degrade sharply when attackers invert that signal by registering matching infrastructure. Response-trace analysis shows that 98% of successful bypasses reason in ways consistent with the inverted signal: the models are following the instruction, but the instruction's core assumption has become false. A counter-intuitive corollary follows: making prompts more specific can degrade already-capable models by replacing broader multi-signal reasoning with exploitable single-signal dependence. We characterize the resulting tension between detection, usability, and adversarial robustness as a navigable tradeoff, introduce Safetility, a deployability-aware metric that penalizes false positives, and argue that closing the adversarial gap likely requires tool augmentation with external ground truth.