In a world where we often rely on virtual agents for various tasks, the gap in their expertise remains a noticeable hinderance in performing real-world work. Anthropic, fueled by the insights of Barry Zhang and Mahesh Murag, proposes an innovative shift from merely developing agents to constructing ‘Skills’. This talk, titled “Don’t Build Agents, Build Skills Instead,” hosted on AI Engineer, delves deep into how skills serve as a transportation vehicle for procedural knowledge, enabling agents to dynamically expand their domain expertise. The simplicity of skills—packaged as organized folders with scripts—allows for easy creation and integration. Numerous entities within the tech community have already observed a burgeoning ecosystem of these skills, spanning foundational skills for financial or scientific tasks to more tailored, enterprise-specific offerings. Barry and Mahesh effectively argue for the role these skills play in bridging expertise deficits in intelligent agents, suggesting that skills could evolve as vital components in future agent architectures. While the presentation is insightful in showcasing the strengths of skills, particularly when illustrating the rapid growth and adoption within various sectors, it does, however, gloss over some challenges. For instance, the discussion would benefit from addressing how skills maintain relevance amidst rapidly changing data sets and technologies. In closing, the notion is clear: By leveraging skills, agents can amass actionable expertise, transforming what they can offer across different verticals. The video by AI Engineer was published on December 8, 2025, with 115,000 views, and the engaging discussion therein is set to reshape how we think about AI development in practice.