SkVM: Revisiting Language VM for Skills across Heterogenous LLMs and Harnesses

TL;DR

SkVM introduces a compiler and runtime system that enhances skill portability and efficiency across diverse LLMs and agent harnesses by leveraging capability profiling and adaptive optimization.

Contribution

It proposes SkVM, a novel system that treats skills as code, enabling portable, efficient execution across heterogeneous LLMs through capability-based compilation and runtime optimization.

Findings

Improves task completion rates across multiple models and environments.

Reduces token consumption by up to 40%.

Achieves up to 3.2x speedup and 19-50x latency reduction.

Abstract

LLM agents increasingly adopt skills as a reusable unit of composition. While skills are shared across diverse agent platforms, current systems treat them as raw context, causing the same skill to behave inconsistently for different agents. This fragility undermines skill portability and execution efficiency. To address this challenge, we analyze 118,000 skills and draw inspiration from traditional compiler design. We treat skills as code and LLMs as heterogeneous processors. To make portability actionable, we decompose a skill's requirements into a set of primitive capabilities, and measure how well each model-harness pair supports them. Based on these capability profiles, we propose SkVM, a compilation and runtime system designed for portable and efficient skill execution. At compile time, SkVM performs capability-based compilation, environment binding, and concurrency extraction. At runtime, SkVM applies JIT code solidification and adaptive recompilation for performance optimization. We evaluate SkVM across eight LLMs of varying scales and three agent harnesses, covering SkillsBench and representative skill tasks. Results demonstrate that SkVM significantly improves task completion rates across different models and environments while reducing token consumption by up to 40%. In terms of performance, SkVM achieves up to 3.2x speedup with enhanced parallelism, and 19-50x latency reduction through code solidification.