Researchers have introduced AntiSkillBench, a comprehensive end-to-end benchmark designed to evaluate privacy leakage, attribute disclosure, and behavioral impersonation within the persona-skill pipeline. This pipeline involves distilling personal interaction histories into portable, executable artifacts known as persona skills to enable downstream agent personalization. To facilitate this evaluation, the authors constructed a dataset comprising 7,500 persona-grounded dialogue traces derived from 50 behaviorally diverse profiles across various task scenarios. The benchmark evaluates three distinct skill-distillation strategies and tests four defense configurations, spanning online and post-hoc interventions such as active risk suppression and passive provenance protection. Experimental results across three frontier agent backbones indicate that persona-skill risks are persistent and transcend specific distillation protocols. These vulnerabilities extend beyond the leakage of explicit attributes to include sensitive communication styles and personality traits. Furthermore, the study reveals that current defense mechanisms demonstrate limited effectiveness and fail to generalize across different risk types or distillation strategies. By exposing these systemic weaknesses, AntiSkillBench provides a rigorous framework for developing privacy-preserving and authenticity-aware persona skills in autonomous agents.