# BadSkill: Backdoor Attacks on Agent Skills via Model-in-Skill Poisoning

arXiv: https://arxiv.org/abs/2604.09378 · HTML: https://mlsanity.com/paper/2604.09378/

## Key facts

- Cited by 27 papers in the map (data through 14 Oct 2026).
- Placed in Prompt injection and agent security, one of 12 directions in LLM safety and security.
- Builds on 31 of its 35 references that are in the map.
- First posted on arXiv 10 Apr 2026.

## Placed in

- LLM safety and security / [Prompt injection and agent security](https://mlsanity.com/safety/prompt-injection-and-agent-security/)

## Builds on (most cited first, 30 of 31)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Toolformer: Language Models Can Teach Themselves to Use Tools](https://mlsanity.com/paper/2302.04761/) (2023)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [A Survey on Large Language Model based Autonomous Agents](https://mlsanity.com/paper/2308.11432/) (2023)
- [ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs](https://mlsanity.com/paper/2307.16789/) (2023)
- [The Rise and Potential of Large Language Model Based Agents: A Survey](https://mlsanity.com/paper/2309.07864/) (2023)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain](https://mlsanity.com/paper/1708.06733/) (2017)
- [Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning](https://mlsanity.com/paper/1712.05526/) (2017)
- [Ignore Previous Prompt: Attack Techniques For Language Models](https://mlsanity.com/paper/2211.09527/) (2022)
- [Yi: Open Foundation Models by 01.AI](https://mlsanity.com/paper/2403.04652/) (2024)
- [Phi-4 Technical Report](https://mlsanity.com/paper/2412.08905/) (2024)
- [Identifying the Risks of LM Agents with an LM-Emulated Sandbox](https://mlsanity.com/paper/2309.15817/) (2023)
- [Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training](https://mlsanity.com/paper/2401.05566/) (2024)
- [AgentPoison: Red-teaming LLM Agents via Poisoning Memory or Knowledge Bases](https://mlsanity.com/paper/2407.12784/) (2024)
- [Backdoor Learning: A Survey](https://mlsanity.com/paper/2007.08745/) (2020)
- [InternLM2 Technical Report](https://mlsanity.com/paper/2403.17297/) (2024)
- [Weight Poisoning Attacks on Pre-trained Models](https://mlsanity.com/paper/2004.06660/) (2020)
- [Certified Defenses for Data Poisoning Attacks](https://mlsanity.com/paper/1706.03691/) (2017)
- [R-Judge: Benchmarking Safety Risk Awareness for LLM Agents](https://mlsanity.com/paper/2401.10019/) (2024)
- [Poisoning Language Models During Instruction Tuning](https://mlsanity.com/paper/2305.00944/) (2023)
- [Hidden Killer: Invisible Textual Backdoor Attacks with Syntactic Trigger](https://mlsanity.com/paper/2105.12400/) (2021)
- [BadNL: Backdoor Attacks against NLP Models with Semantic-preserving Improvements](https://mlsanity.com/paper/2006.01043/) (2020)
- [Dataset Security for Machine Learning: Data Poisoning, Backdoor Attacks, and Defenses](https://mlsanity.com/paper/2012.10544/) (2020)
- [Prompt Injection Attack to Tool Selection in LLM Agents](https://mlsanity.com/paper/2504.19793/) (2025)
- [Backdooring Instruction-Tuned Large Language Models with Virtual Prompt Injection](https://mlsanity.com/paper/2307.16888/) (2023)
- [Instructions as Backdoors: Backdoor Vulnerabilities of Instruction Tuning for Large Language Models](https://mlsanity.com/paper/2305.14710/) (2023)
- [ToolSword: Unveiling Safety Issues of Large Language Models in Tool Learning Across Three Stages](https://mlsanity.com/paper/2402.10753/) (2024)
- [Instruction Backdoor Attacks Against Customized LLMs](https://mlsanity.com/paper/2402.09179/) (2024)
- [A Study of Backdoors in Instruction Fine-tuned Language Models](https://mlsanity.com/paper/2406.07778/) (2024)

## Built on by (most cited first, 27 of 27)

- [SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces](https://mlsanity.com/paper/2605.12015/) (2026)
- [When Safe Skills Collide: Measuring Compositional Risk in Agent Skill Ecosystems](https://mlsanity.com/paper/2606.00448/) (2026)
- [Under the Hood of SKILL.md: Semantic Supply-chain Attacks on AI Agent Skill Registry](https://mlsanity.com/paper/2605.11418/) (2026)
- [AgentTrap: Measuring Runtime Trust Failures in Third-Party Agent Skills](https://mlsanity.com/paper/2605.13940/) (2026)
- [AutoResearch AI: Towards AI-Powered Research Automation for Scientific Discovery](https://mlsanity.com/paper/2605.23204/) (2026)
- [SkillHarm: Lifecycle-Aware Skill-Based Attacks via Automated Construction](https://mlsanity.com/paper/2606.02540/) (2026)
- [SkillGuard: A Permission-Centric Framework for Agent Skill Security](https://mlsanity.com/paper/2606.03024/) (2026)
- [Dynamic Agent Skills: A Lifecycle Survey and Taxonomy of Evolving Skill Libraries](https://mlsanity.com/paper/2607.10113/) (2026)
- [BraveGuard: From Open-World Threats to Safer Computer-Use Agents](https://mlsanity.com/paper/2606.01166/) (2026)
- [ColluSkill: Adversarial Cross-Skill Composition for Evading Agent Skill Scanners](https://mlsanity.com/paper/2608.09732/) (2026)
- [CompoSkill: Compositional Skill Chain Attacks from Individually Scanner-Passing LLM Agent Skills](https://mlsanity.com/paper/2608.16246/) (2026)
- [EVOMAL: Self-Poisoning in Self-Evolving Coding Agents](https://mlsanity.com/paper/2608.25776/) (2026)
- [SLBench: Evaluating How LLM Agents Follow Logical Relations in Skills](https://mlsanity.com/paper/2607.09016/) (2026)
- [When Experience Becomes Instruction: Trajectory Poisoning in Self-Evolving Agent Skill Systems](https://mlsanity.com/paper/2608.05563/) (2026)
- [Convergent Detour Hijacking: Task-Preserving Resource Amplification in Skill-Based LLM Agents](https://mlsanity.com/paper/2608.12273/) (2026)
- [Security of OpenClaw Agents: Fundamentals, Attacks, and Countermeasures](https://mlsanity.com/paper/2605.25435/) (2026)
- [EvoSkill Injection: Red-Teaming Autonomous Skill Generation and Evolution in Self-Evolving Agents](https://mlsanity.com/paper/2608.30429/) (2026)
- [Versioned Transitive Dependency-Closure Binding and Operation-Time Effect Governance for Agent Skills: ClosureBound](https://mlsanity.com/paper/2609.05920/) (2026)
- [Query-Only Backdoor Attacks on Self-Evolving Skills via Trajectory Poisoning](https://mlsanity.com/paper/2608.08303/) (2026)
- [ElasticBack: Stealthy Conditional Backdoor in LLM-Agent Skills via Coupled Trigger-Rule Optimization](https://mlsanity.com/paper/2608.09577/) (2026)
- [Beyond Handcrafted Security: Towards Self-Evolving Defense for LLM Agents](https://mlsanity.com/paper/2608.12977/) (2026)
- [Towards a Systems Foundation for Agentic Skills: Architecture, Lifecycle, and Security](https://mlsanity.com/paper/2608.29596/) (2026)
- [SkillSpec: Intent-Masked Specification Reasoning for Agent Skill Correctness](https://mlsanity.com/paper/2609.06052/) (2026)
- [Stealth Apart, Harm Together: Skill Cascading Attacks on Skill-Based Agent Systems](https://mlsanity.com/paper/2609.30383/) (2026)
- [Hiding in Plain Sight: Decoupling Pretext from Actuation for Skill Poisoning in LLM Agents](https://mlsanity.com/paper/2609.39352/) (2026)
- [Surviving the Router: Optimizing Skill Injections for Retrieval and Execution](https://mlsanity.com/paper/2610.08098/) (2026)
- [CoRedSkill: Stealthy Skill Composition Attacks via AST-Guided Injection](https://mlsanity.com/paper/2610.20049/) (2026)

## Abstract (arXiv)

Agent ecosystems increasingly rely on installable skills to extend functionality, and some skills bundle learned model artifacts as part of their execution logic. This creates a supply-chain risk that is not captured by prompt injection or ordinary plugin misuse: a third-party skill may appear benign while concealing malicious behavior inside its bundled model. We present BadSkill, a backdoor attack formulation that targets this model-in-skill threat surface. In BadSkill, an adversary publishes a seemingly benign skill whose embedded model is backdoor-fine-tuned to activate a hidden payload only when routine skill parameters satisfy attacker-chosen semantic trigger combinations. To realize this attack, we train the embedded classifier with a composite objective that combines classification loss, margin-based separation, and poison-focused optimization, and evaluate it in an OpenClaw-inspired simulation environment that preserves third-party skill installation and execution while enabling controlled multi-model study. Our benchmark spans 13 skills, including 8 triggered tasks and 5 non-trigger control skills, with a combined main evaluation set of 571 negative-class queries and 396 trigger-aligned queries. Across eight architectures (494M--7.1B parameters) from five model families, BadSkill achieves up to 99.5% average attack success rate (ASR) across the eight triggered skills while maintaining strong benign-side accuracy on negative-class queries. In poison-rate sweeps on the standard test split, a 3% poison rate already yields 91.7% ASR. The attack remains effective across the evaluated model scales and under five text perturbation types. These findings identify model-bearing skills as a distinct model supply-chain risk in agent ecosystems and motivate stronger provenance verification and behavioral vetting for third-party skill artifacts.