# AgentAuditor: Human-Level Safety and Security Evaluation for LLM Agents

arXiv: https://arxiv.org/abs/2506.00641 · HTML: https://mlsanity.com/paper/2506.00641/

## Key facts

- Cited by 90 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning.
- Builds on 1 of its 3 references that are in the map.
- First posted on arXiv 31 May 2025.

## Placed in

- LLM reasoning / [Reasoning for agents, planning and code](https://mlsanity.com/lr/reasoning-for-agents-planning-and-code/)
- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 1 of 1)

- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)

## Built on by (most cited first, 30 of 90)

- [MemGen: Weaving Generative Latent Memory for Self-Evolving Agents](https://mlsanity.com/paper/2509.24704/) (2025)
- [ToolSafe: Enhancing Tool Invocation Safety of LLM-based agents via Proactive Step-level Guardrail and Feedback](https://mlsanity.com/paper/2601.10156/) (2026)
- [A-MemGuard: A Proactive Defense Framework for LLM-Based Agent Memory](https://mlsanity.com/paper/2510.02373/) (2025)
- [LLM-based Agents Suffer from Hallucinations: A Survey of Taxonomy, Methods, and Directions](https://mlsanity.com/paper/2509.18970/) (2025)
- [ATBench: A Diverse and Realistic Agent Trajectory Benchmark for Safety Evaluation and Diagnosis](https://mlsanity.com/paper/2604.02022/) (2026)
- [A Survey on Autonomy-Induced Security Risks in Large Model-Based Agents](https://mlsanity.com/paper/2506.23844/) (2025)
- [Building a Foundational Guardrail for General Agentic Systems via Synthetic Data](https://mlsanity.com/paper/2510.09781/) (2025)
- [The Attack and Defense Landscape of Agentic AI: A Comprehensive Survey](https://mlsanity.com/paper/2603.11088/) (2026)
- [Who&When Pro: Can LLMs Really Attribute Failures in AI Agents?](https://mlsanity.com/paper/2607.09996/) (2026)
- [RAS-Eval: A Comprehensive Benchmark for Security Evaluation of LLM Agents in Real-World Environments](https://mlsanity.com/paper/2506.15253/) (2025)
- [TraceSafe: A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories](https://mlsanity.com/paper/2604.07223/) (2026)
- [SafetyRepro: Configuration-Conditional Rank Instability on Alignment Benchmarks](https://mlsanity.com/paper/2605.25492/) (2026)
- [Graph-Augmented Large Language Model Agents: Current Progress and Future Prospects](https://mlsanity.com/paper/2507.21407/) (2025)
- [A Vision for Access Control in LLM-based Agent Systems](https://mlsanity.com/paper/2510.11108/) (2025)
- [CentaurEval: Benchmarking Human-in-the-Loop Value in Agentic Coding](https://mlsanity.com/paper/2512.04111/) (2025)
- [Reflect-Guard: Enhancing LLM Safeguards against Adversarial Prompts via Logical Self-Reflection](https://mlsanity.com/paper/2605.24834/) (2026)
- [It's a TRAP! Task-Redirecting Agent Persuasion Benchmark for Web Agents](https://mlsanity.com/paper/2512.23128/) (2025)
- [Learning How to Remember: A Meta-Cognitive Management Method for Structured and Transferable Agent Memory](https://mlsanity.com/paper/2601.07470/) (2026)
- [TAME: A Trustworthy Test-Time Evolution of Agent Memory with Systematic Benchmarking](https://mlsanity.com/paper/2602.03224/) (2026)
- [HINTBench: Horizon-agent Intrinsic Non-attack Trajectory Benchmark](https://mlsanity.com/paper/2604.13954/) (2026)
- [Risky-Bench: Probing Agentic Safety Risks under Real-World Deployment](https://mlsanity.com/paper/2602.03100/) (2026)
- [When Actions Go Off-Task: Detecting and Correcting Misaligned Actions in Computer-Use Agents](https://mlsanity.com/paper/2602.08995/) (2026)
- [Learning When to Act or Refuse: Guarding Agentic Reasoning Models for Safe Multi-Step Tool Use](https://mlsanity.com/paper/2603.03205/) (2026)
- [Beyond Accuracy: Policy Invariance as a Reliability Test for LLM Safety Judges](https://mlsanity.com/paper/2605.06161/) (2026)
- [AtelierEval: Agentic Evaluation of Humans & LLMs as Text-to-Image Prompters](https://mlsanity.com/paper/2605.22645/) (2026)
- [TRACES: Proactive Safety Auditing for Multi-Turn LLM Agents via Trajectory-State Modeling](https://mlsanity.com/paper/2605.27690/) (2026)
- [Pre-Registering the Detectable Effect: A Paired-MDE Budget for 4-bit Quantization Benchmarks, with a Pilot Audit](https://mlsanity.com/paper/2605.28873/) (2026)
- [GRASP: Plan-Guided Graph Retrieval with Adaptive Fusion and Reranking on Semi-Structured Knowledge Bases](https://mlsanity.com/paper/2605.30237/) (2026)
- [Safety Testing LLM Agents at Scale: From Risk Discovery to Evidence-Grounded Verification](https://mlsanity.com/paper/2607.01793/) (2026)
- [To trust or not to trust: Attention-based Trust Management for LLM Multi-Agent Systems](https://mlsanity.com/paper/2506.02546/) (2025)

## Abstract (arXiv)

Despite the rapid advancement of LLM-based agents, the reliable evaluation of their safety and security remains a significant challenge. Existing rule-based or LLM-based evaluators often miss dangers in agents' step-by-step actions, overlook subtle meanings, fail to see how small issues compound, and get confused by unclear safety or security rules. To overcome this evaluation crisis, we introduce AgentAuditor, a universal, training-free, memory-augmented reasoning framework that empowers LLM evaluators to emulate human expert evaluators. AgentAuditor constructs an experiential memory by having an LLM adaptively extract structured semantic features (e.g., scenario, risk, behavior) and generate associated chain-of-thought reasoning traces for past interactions. A multi-stage, context-aware retrieval-augmented generation process then dynamically retrieves the most relevant reasoning experiences to guide the LLM evaluator's assessment of new cases. Moreover, we developed ASSEBench, the first benchmark designed to check how well LLM-based evaluators can spot both safety risks and security threats. ASSEBench comprises 2293 meticulously annotated interaction records, covering 15 risk types across 29 application scenarios. A key feature of ASSEBench is its nuanced approach to ambiguous risk situations, employing "Strict" and "Lenient" judgment standards. Experiments demonstrate that AgentAuditor not only consistently improves the evaluation performance of LLMs across all benchmarks but also sets a new state-of-the-art in LLM-as-a-judge for agent safety and security, achieving human-level accuracy. Our work is openly accessible at https://github.com/Astarojth/AgentAuditor.