SkillConsist：通过双向图对齐检测代理技能不一致性
原标题：SkillConsist: Detecting Inconsistencies in Agent Skills via Bidirectional Graph Alignment
AI 摘要
arXiv 论文提出 SkillConsist 方法，用于检测 LLM 代理技能中的声明与实现不一致问题。该方法通过 LLM 分离声明和实现内容，构建行为图并进行双向图对齐和差异比较，以识别冲突。在包含 633 个技能的基准测试中，SkillConsist 在包级检测上达到 87.93% 的 F1 分数，比最佳基线提升 20.43 个百分点。
正文节选
Computer Science > Machine Learning Title:SkillConsist: Detecting Inconsistencies in Agent Skills via Bidirectional Graph Alignment View PDF HTML (experimental) Abstract:Agent Skills provide reusable capabilities to LLM agents. Agent Skill inconsistencies can expose undisclosed dangerous behavior or cause wrong Skill selection. Recent Agent Skill research has increasingly examined Agent Skill consistency detection. Existing methods evaluate behaviors or security-property graphs again