Researchers have developed a framework called ACES (Agentic Continuous Evaluation of Skills) to evaluate the effectiveness of skills and product capability packages in enterprise agent programs. ACES runs paired live trials with and without a target skill, normalizes trajectories into a standard format, and grades six default runtime metrics. The framework has been tested on 145 real skills from internal enterprise repositories and public catalogs, showing that traditional scan-only gates are insufficient to measure the value of skills. Instead, ACES provides a more comprehensive evaluation of skills, including their added value for specific tasks and workflows. --- Why it matters: This matters because it addresses a critical challenge in AI development: evaluating the effectiveness of skills and capabilities in real-world applications. By providing a standardized framework for continuous evaluation, ACES enables developers to make data-driven decisions about skill deployment and improvement. Source: https://arxiv.org/abs/2608.20614