停止构建智能体，开始构建技能

The future of artificial intelligence lies not in a multitude of specialized bots, but in a universal agent equipped with a comprehensive library of procedural expertise. The prevailing trend in AI development often involves creating isolated agents for specific tasks, which leads to increased system complexity, high maintenance costs, and significant challenges in cross-domain knowledge sharing. This fragmented approach to agent development ultimately limits AI's generality and adaptability. Instead, the article advocates for a paradigm shift towards building a core, universal agent that can leverage an extensible library of 'skills' or 'procedural expertise.' These skills are envisioned as modular, reusable functional units that the universal agent can dynamically invoke and combine to address diverse tasks. This strategic pivot from 'building isolated agents' to 'building composable skills' promises to foster more powerful and versatile AI systems. Such a framework would enable AI to better navigate and respond to the complexities of a dynamic world, promoting efficiency and flexibility over specialized, siloed functionalities.