EffiSkill: Agent Skill Based Automated Code Efficiency Optimization

2026-03-29

AI Summary

Existing code optimization approaches are often limited to instance-specific rewriting and lack generalization capabilities. This work proposes EffiSkill, a novel framework that formulates code efficiency optimization as reusable agent skills. By automatically mining generic operators and meta-skills from a large corpus of slow/fast program pairs, EffiSkill constructs a portable optimization toolkit capable of diagnosing new programs, retrieving and composing relevant skills, and generating optimized code candidates—all without requiring execution feedback. The approach enables execution-agnostic automated optimization and facilitates cross-task knowledge transfer. Evaluated on the EffiBench-X benchmark, EffiSkill achieves a 3.69–12.52 percentage point improvement in optimization success rate over the strongest baseline, demonstrating the effectiveness of its skill-reuse mechanism.

Abstract

Code efficiency is a fundamental aspect of software quality, yet how to harness large language models (LLMs) to optimize programs remains challenging. Prior approaches have sought for one-shot rewriting, retrieved exemplars, or prompt-based search, but they do not explicitly distill reusable optimization knowledge, which limits generalization beyond individual instances. In this paper, we present EffiSkill, a framework for code-efficiency optimization that builds a portable optimization toolbox for LLM-based agents. The key idea is to model recurring slow-to-fast transformations as reusable agent skills that capture both concrete transformation mechanisms and higher-level optimization strategies. EffiSkill adopts a two-stage design: Stage I mines Operator and Meta Skills from large-scale slow/fast program pairs to build a skill library; Stage II applies this library to unseen programs through execution-free diagnosis, skill retrieval, plan composition, and candidate generation, without runtime feedback. Results on EffiBench-X show that EffiSkill achieves higher optimization success rates, improving over the strongest baseline by 3.69 to 12.52 percentage points across model and language settings. These findings suggest that mechanism-level skill reuse provides a useful foundation for execution-free code optimization, and that the resulting skill library can serve as a reusable resource for broader agent workflows.

Problem

• code efficiency
• large language models
• optimization knowledge
• reusable skills
• program optimization

Innovation

• agent skill
• code efficiency optimization
• reusable transformation
• execution-free diagnosis
• skill library