技能级联攻击：分散隐蔽、组合致害的智能体系统威胁
原标题：Stealth Apart, Harm Together: Skill Cascading Attacks on Skill-Based Agent Systems
AI 摘要
该论文提出「技能级联攻击」这一新威胁范式：攻击者将恶意目标分散到多个技能中，每个技能单独看都无害且能通过单技能扫描，但组合执行后产生危害。作者构建了自动化多智能体红队框架 SkillCascade，并发布包含 213 个验证用例、覆盖 10 个领域的 SkillCascade-Bench 基准。在 OpenClaw、Claude Code、CodeX 等主流智能体系统与八种 LLM 上，级联攻击平均以 89.4% 的成功率诱发有害行为，同时绕过现有单技能扫描器、联合技能扫描器与运行时防御，揭示组件级完整性与系统级安全之间的缺口。
正文节选
Stealth Apart, Harm Together: Skill Cascading Attacks on Skill-Based Agent Systems Abstract A skill is a modular package of natural-language instructions, executable scripts, and reference resources that an agent can load at runtime to extend its capabilities for a specific task. Skill-based agent systems therefore enable flexible reuse of third-party capabilities, but the openness of this skill ecosystem also opens up a new attack surface. Prior work has focused on vulnerabilities within indivi