[论文] SkCC: Portable and Secure Skill Compilation for Cross-Framework LLM Agents

论文概要

研究领域: Agent / Skill 编译 / 跨平台 作者: Yihao Wang, Yuheng Ji, Mingyu Cao 等（中山大学团队） 发布时间: 2026-05-05 arXiv: 2605.03353 代码: 开源编译器代码附于论文

中文摘要

LLM 智能体越来越依赖可复用技能（如 SKILL markdown 文件）执行复杂任务，但这些工件缺乏可移植性：智能体框架对提示格式高度敏感，导致同一技能在不同框架上性能差异巨大。大多数技能以格式无关的 Markdown 编写一次，却需要为每个框架进行代价高昂的改写，同时安全问题也 largely unaddressed，实践中存在广泛漏洞。为此，我们提出 SkCC，一个面向 LLM 智能体的编译器，将经典编译设计引入智能体技能开发。SkCC 以 SkIR 为中心——一种强类型中间表示，将技能语义与框架特定格式解耦，从而支持跨智能体框架的可移植部署。在此 IR 之上，静态优化器强制执行安全约束，在部署前拦截漏洞。SkCC 实现为四阶段流水线，将 m 个技能和 n 个框架的适配复杂度从 O(m×n) 降到 O(m+n)。SkillsBench 实验表明，SkCC 在 Claude Code 上任务通过率从 21.1% 提升到 33.3%（+12.2pp），在 Kimi CLI 上从 35.1% 提升到 48.7%（+13.5pp）。编译延迟 <10ms，主动安全触发率 94.8%，运行时 Token 节省 10%~46%。

原文摘要

LLM agents increasingly rely on reusable skills (e.g., SKILL markdown files) to execute complex tasks, yet these artifacts lack portability: agent frameworks are highly sensitive to prompt formatting, leading to a large performance variation for the same skill. Nevertheless, most skills are authored once as format-agnostic Markdown, necessitating costly per-framework rewrites and also leaving security largely unaddressed, with widespread vulnerabilities in practice. To address this, we present SkCC, a compiler for LLM agents that introduces classical compilation design into agent skill development. SkCC centers on SkIR, a strongly-typed intermediate representation that decouples skill semantics from framework-specific formatting, thus enabling portable deployment across agent frameworks. Atop of this IR, a static Optimizer enforces security constraints, blocking vulnerabilities before deployment. Implemented as a four-phase pipeline, SkCC effectively reduces adaptation complexity from O(m × n) to O(m + n) across m skills and n frameworks. Experiments on SkillsBench demonstrate that SkCC delivers consistent and substantial gains over original counterparts, with pass rate increases from 21.1% to 33.3% on Claude Code and from 35.1% to 48.7% on Kimi CLI. Further, the design achieves sub-10ms compilation latency, 94.8% proactive security trigger rate, and 10-46% runtime token savings across frameworks.

--- *自动采集于 2026-07-03*

#论文 #arXiv #Agent #Skill #编译器 #中山大学 #开源 #小凯