EvoSkill Injection: Red-Teaming Autonomous Skill Generation and Evolution in Self-Evolving Agents
This paper introduces EvoSkill Injection, a threat model where self-evolving LLM agents can be manipulated into generating, storing, and later reusing malicious “skills” as if they were legitimate learned capabilities, and presents SARGE plus two benchmarks to red-team both the injection process and post-attack activation. The interesting bit is that it frames agent memory/skill evolution as a persistent capability-corruption surface rather than a one-off prompt-injection problem, though from the abstract it seems mainly evaluative and benchmark-driven, with limited discussion of defenses or how broadly the results transfer across real deployed agent architectures.
Abstract
LLM-based agent systems increasingly adopt skill-based architectures to reduce repetitive reasoning costs and improve stable, efficient task execution. Recent studies propose self-evolving agents that autonomously generate, refine, and reuse skills from past experiences to enable continuous capability evolution. However, autonomous skill evolution introduces a new attack surface in which malicious capabilities are generated, stored, and reused as legitimate skills. In this paper, we define EvoSkill Injection as a threat model targeting the autonomous skill generation and evolution pipeline of self-evolving agents. We further propose SARGE (Red-teaming Autonomous Skill Generation and Evolution in self-evolving agents), a red-teaming framework for evaluating this threat model through iterative generation, escalation, and reinforcement interactions. To support our framework, we construct EvoSkillBench, a benchmark dataset of malicious interaction trajectories for inducing malicious skill formation in self-evolving agents, and introduce EvoSkillSafetyBench, a post-attack benchmark for evaluating whether injected malicious skills are subsequently retrieved and activated as harmful behaviors. Our evaluation shows that SARGE induces malicious skill formation and that injected skills are persistently stored and repeatedly activated, highlighting the risk of persistent capability corruption.