# Defense-as-Skill: Evolving Runtime Guard Skill for Skill-Augmented Agents

arXiv: https://arxiv.org/abs/2609.01487 · HTML: https://mlsanity.com/paper/2609.01487/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 30 references that are in the map.
- First posted on arXiv 1 Sep 2026.

## Builds on (most cited first, 19 of 19)

- [GLM-5: from Vibe Coding to Agentic Engineering](https://mlsanity.com/paper/2602.15763/) (2026)
- [SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks](https://mlsanity.com/paper/2602.12670/) (2026)
- [SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning](https://mlsanity.com/paper/2602.08234/) (2026)
- [AgentSpec: Customizable Runtime Enforcement for Safe and Reliable LLM Agents](https://mlsanity.com/paper/2503.18666/) (2025)
- [CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification](https://mlsanity.com/paper/2604.01687/) (2026)
- [Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward](https://mlsanity.com/paper/2602.12430/) (2026)
- [Agent Skills in the Wild: An Empirical Study of Security Vulnerabilities at Scale](https://mlsanity.com/paper/2601.10338/) (2026)
- [Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks](https://mlsanity.com/paper/2602.20156/) (2026)
- [SkillClaw: Let Skills Evolve Collectively with Agentic Evolver](https://mlsanity.com/paper/2604.08377/) (2026)
- [SkillLearnBench: Benchmarking Continual Learning Methods for Agent Skill Generation on Real-World Tasks](https://mlsanity.com/paper/2604.20087/) (2026)
- ["Do Not Mention This to the User": Detecting and Understanding Malicious Agent Skills in the Wild](https://mlsanity.com/paper/2602.06547/) (2026)
- [WildClawBench: A Benchmark for Real-World, Long-Horizon Agent Evaluation](https://mlsanity.com/paper/2605.10912/) (2026)
- [SkillAttack: Automated Red Teaming of Agent Skills through Attack Path Refinement](https://mlsanity.com/paper/2604.04989/) (2026)
- [Taming OpenClaw: Security Analysis and Mitigation of Autonomous LLM Agent Threats](https://mlsanity.com/paper/2603.11619/) (2026)
- [SkillFlow:Benchmarking Lifelong Skill Discovery and Evolution for Autonomous Agents](https://mlsanity.com/paper/2604.17308/) (2026)
- [Uncovering Security Threats and Architecting Defenses in Autonomous Agents: A Case Study of OpenClaw](https://mlsanity.com/paper/2603.12644/) (2026)
- [SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces](https://mlsanity.com/paper/2605.12015/) (2026)
- [Agent Skill Acquisition for Large Language Models via CycleQD](https://mlsanity.com/paper/2410.14735/) (2024)
- [SkillPyramid: A Hierarchical Skill Consolidation Framework for Self-Evolving Agents](https://mlsanity.com/paper/2606.03692/) (2026)

## Abstract (arXiv)

Skill-augmented agents load reusable skills as persistent runtime context, improving task performance but also giving malicious skills a durable channel for steering future actions. Such skills may leak secrets, corrupt code, bypass approvals, or stage data for exfiltration only after a concrete user task and workspace state make the unsafe action appear useful. This makes pre-install vetting insufficient and calls for runtime, task-conditioned protection. We propose Defense-as-Skill, a defense paradigm that implements the runtime guard itself as an installable, inspectable, and editable skill. Our guard, SkillSonar, runs alongside untrusted task skills and checks sensitive actions against the user's task boundary, routing each action to an allow, replan, or confirmation decision without modifying the underlying agent runtime. To study this setting, we construct SCOPE-R, a task-conditioned dataset covering 6 risk families and 21 sub-categories, with 206 attack-confirmed malicious instances and 43 benign tasks. We then improve SkillSonar on the SCOPE-R training subset using runtime guard-skill evolution, a Monte-Carlo Tree Search procedure that evolves the on-disk guard skill from feedback on the rollouts. Across Claude Code and OpenClaw, the evolved guard substantially reduces attack success while maintaining a favorable safety-utility trade-off; on repeated GLM-5 runs, ID ASR drops from 0.482 to 0.104 and OOD ASR from 0.606 to 0.115, with gains transferring across victim models, held-out risk families, and external benchmarks. These results show that runtime guards can be skill-native, transparent, and evolvable policy artifacts for skill-augmented agents.