Skillinfinity improves agent skills by optimizing how optimization is done
Skill$^{\infty}$: Recursive Self-Improvement for Agent Skill Optimization
Artificial IntelligenceComputation and Language
Summary
Improving how machines improve themselves is a big goal in artificial intelligence. The authors created Skill∞, a system where an AI not only learns new skills but also gets better at improving those skills by changing its own improvement methods. This means the AI can improve its ability to learn and optimize more effectively over time, without relying on preset rules. When tested on a wide range of tasks, Skill∞ performed better than previous methods and even human-curated skills.
What this means in practice
- •For software developers: Use Skill∞ to create AI assistants that can autonomously improve their own task-solving skills over time without manual reprogramming.
- •For automation engineers: Deploy Skill∞ to optimize automation workflows by enabling systems to refine both their tasks and the way they improve those tasks.
Authors
Wangcheng Tao, Zhanglu Yan, Han Wu, Weng-Fai Wong
Abstract
Recursive self-improvement (RSI), in which a system also improves the way it improves, is a long-standing aspiration on the path to AGI. Language agents increasingly store procedural knowledge in reusable skills. Optimizing skills offers a direct way to achieve RSI, because the optimization procedure itself can be encoded as a skill. Existing skill optimizers, however, follow a hand-designed procedure. Even methods that evolve a meta-skill confine it to a fixed pipeline and relearn it for each task. We present Skill$^{\infty}$, which makes skill optimization recursively self-improving: a meta agent follows a meta-skill to optimize task skills and optimizes the meta-skill itself using the same procedure. The optimization is agent-led. The meta agent diagnoses execution traces, proposes candidates, requests independent reviews, and allocates the search budget. The meta-skill defines these steps, allowing self-improvement to rewrite them. This meta-skill is also decoupled from task skills, refined on multiple skill optimization tasks and then reused to optimize task skills across diverse domains. As in deployment, skills at both levels are standard packages loaded through native skill discovery. On 83 SkillsBench tasks, Skill$^{\infty}$ reaches a pass rate of 60.7%, compared with 56.4% for human-curated skills and 54.5% for SkillEvolver.