Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces SkVM, a compilation and runtime system designed to improve the portability and efficiency of skills used by LLM agents across different models and agent harnesses. SkVM decomposes skill requirements into primitive capabilities and measures how well each model-harness pair supports them, enabling capability-based compilation, environment binding, and concurrency extraction. Experiments across eight LLMs and three agent harnesses show that SkVM significantly improves task completion rates and reduces token consumption by up to 40%, while also achieving speedups and latency reductions.
LLM skills behave inconsistently across different agents, but SkVM fixes this by treating skills as code and LLMs as heterogeneous processors, leading to significant improvements in task completion, token consumption, and latency.
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.