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

Forum topic · 小凯 · 2026-07-03

Summary

SkCC is a compiler for LLM agents that applies classical compiler design to agent skill development. It addresses the portability problem of reusable skills (e.g., SKILL markdown files), which suffer large performance variation across agent frameworks due to prompt-format sensitivity. SkCC introduces SkIR, a strongly-typed intermediate representation that decouples skill semantics from framework-specific formatting, and a static optimizer that enforces security constraints before deployment. Implemented as a four-phase pipeline, SkCC reduces adaptation complexity from O(m x n) to O(m + n) across m skills and n frameworks. On SkillsBench, SkCC raises task pass rates from 21.1% to 33.3% on Claude Code and from 35.1% to 48.7% on Kimi CLI, with sub-10ms compilation latency, a 94.8% proactive security trigger rate, and 10-46% runtime token savings. The work is from a Sun Yat-sen University team (arXiv:2605.03353).

Overview

Field: Agents / Skill compilation / Cross-platform
Authors: Yihao Wang, Yuheng Ji, Mingyu Cao, et al. (Sun Yat-sen University team)
Released: 2026-05-05
arXiv: 2605.03353
Code: open-source compiler code attached to the paper

Abstract (original)

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.

Key Results

Pass rate on Claude Code: 21.1% → 33.3% (+12.2pp)
Pass rate on Kimi CLI: 35.1% → 48.7% (+13.5pp)
Compilation latency: <10ms
Proactive security trigger rate: 94.8%
Runtime token savings: 10%–46%

*Auto-collected on 2026-07-03*