# Detecting Malicious Agent Skills in the Wild using Attention

arXiv: https://arxiv.org/abs/2606.23416 · HTML: https://mlsanity.com/paper/2606.23416/

## Key facts

- Cited by 8 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 19 of its 22 references that are in the map.
- First posted on arXiv 22 Jun 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, 19 of 19)

- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (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)
- [DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence](https://mlsanity.com/paper/2606.19348/) (2026)
- [Defeating Prompt Injections by Design](https://mlsanity.com/paper/2503.18813/) (2025)
- [SecAlign: Defending Against Prompt Injection with Preference Optimization](https://mlsanity.com/paper/2410.05451/) (2024)
- [Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward](https://mlsanity.com/paper/2602.12430/) (2026)
- [Attention Tracker: Detecting Prompt Injection Attacks in LLMs](https://mlsanity.com/paper/2411.00348/) (2024)
- [Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks](https://mlsanity.com/paper/2602.20156/) (2026)
- [Get my drift? Catching LLM Task Drift with Activation Deltas](https://mlsanity.com/paper/2406.00799/) (2024)
- [Model Context Protocol (MCP) at First Glance: Studying the Security and Maintainability of MCP Servers](https://mlsanity.com/paper/2506.13538/) (2025)
- [LLM Platform Security: Applying a Systematic Evaluation Framework to OpenAI's ChatGPT Plugins](https://mlsanity.com/paper/2309.10254/) (2023)
- [VeriGuard: Enhancing LLM Agent Safety via Verified Code Generation](https://mlsanity.com/paper/2510.05156/) (2025)
- [A Closer Look at System Prompt Robustness](https://mlsanity.com/paper/2502.12197/) (2025)
- [SkillTrojan: Backdoor Attacks on Skill-Based Agent Systems](https://mlsanity.com/paper/2604.06811/) (2026)
- [Beyond the Protocol: Unveiling Attack Vectors in the Model Context Protocol (MCP) Ecosystem](https://mlsanity.com/paper/2506.02040/) (2025)
- [Under the Hood of SKILL.md: Semantic Supply-chain Attacks on AI Agent Skill Registry](https://mlsanity.com/paper/2605.11418/) (2026)
- [Exploiting LLM Agent Supply Chains via Payload-less Skills](https://mlsanity.com/paper/2605.14460/) (2026)

## Built on by (most cited first, 8 of 8)

- [MaliciousSkillBench: A Comprehensive Benchmark for Malicious Agent Skill Detection](https://mlsanity.com/paper/2608.19901/) (2026)
- [SkillPoison: Progressive Skill Poisoning via Successful Experiences](https://mlsanity.com/paper/2610.07645/) (2026)
- [SkillSentry: Adaptive Honey Worlds for Dynamic Safety Testing of Agent Skills](https://mlsanity.com/paper/2608.03485/) (2026)
- [Towards a Systems Foundation for Agentic Skills: Architecture, Lifecycle, and Security](https://mlsanity.com/paper/2608.29596/) (2026)
- [SkillSecurer: Detecting and Patching Prompt-Injection Vulnerabilities in AI Agent Skills](https://mlsanity.com/paper/2609.14079/) (2026)
- [SKILLLITE: Evidence-Guided Malicious Skill Auditing with Compact LLMs](https://mlsanity.com/paper/2609.36879/) (2026)
- [Pretext: Defeating Malicious Skill Detection Frameworks for AI Agents](https://mlsanity.com/paper/2609.39607/) (2026)
- [CoRedSkill: Stealthy Skill Composition Attacks via AST-Guided Injection](https://mlsanity.com/paper/2610.20049/) (2026)

## Abstract (arXiv)

LLM agents increasingly load skills, file-based packages of natural-language instructions written by third parties and distributed through marketplaces, that execute with the user's privileges. A single malicious skill can exfiltrate data, hijack the agent, or persist as a supply-chain foothold, which turns the skill marketplace into a new attack surface for agentic systems. Prompt-injection defenses do not carry over to this setting. They rely on a boundary between trusted instructions and untrusted data, whereas a skill is itself a body of instructions, so an injected command sits among many legitimate ones and inherits their authority. We present Locate-and-Judge, a two-stage detector designed for this regime. A lightweight locator scores the structural spans of a skill by the instruction-following attention each span draws and retains only the top-K. A judge then examines the retained spans in detail. Concentrating the costly judgment on a few high-attention spans lets the detector audit an entire marketplace instead of a sample. Compared to direct LLM-based scanning, this approach offers an order-of-magnitude cost reduction, dramatically increasing its scalability at a small cost to recall, and it dominates keyword and regex baselines at comparable expense. Deployed at marketplace scale and at negligible cost, Locate-and-Judge flags skills with high precision, the majority of which we manually confirmed as malicious, surfacing dozens of live malicious skills, including several disguised as benign functionality and many that SkillSpector and Cisco Skill Scanner fail to detect. We release the resulting labeled dataset.