UV-SAM is a framework that adapts the Segment Anything Model (SAM) for urban village identification using satellite images. Urban villages are informal residential areas with inadequate infrastructure and poor living conditions, closely related to the Sustainable Development Goals (SDGs). Traditional methods for identifying urban villages are time-consuming and labor-intensive, but recent studies have used computer vision techniques to detect them more efficiently. However, existing methods either focus on simple classification or fail to provide accurate boundary information. UV-SAM addresses this by leveraging a small semantic segmentation model to generate mixed prompts, which are then fed into SAM for fine-grained boundary identification. Experimental results on two Chinese datasets show that UV-SAM outperforms existing baselines and provides insights into the development trends of urban villages over time. The framework also demonstrates the effectiveness of SAM for segmenting boundaries and highlights the potential of vision foundation models for sustainable cities. UV-SAM introduces a generalist-specialist framework that automatically generates four types of prompts for urban village identification. The framework uses a lightweight semantic segmentation model to generate prompts for SAM, which then identifies urban village boundaries. The model achieves significant performance improvements compared to state-of-the-art models and provides valuable insights into the spatial distribution and temporal trends of urban villages. The study also shows that urban villages are decreasing in number and area over time, which has implications for urban planning and governance. The dataset and code for UV-SAM are available at the provided link.UV-SAM is a framework that adapts the Segment Anything Model (SAM) for urban village identification using satellite images. Urban villages are informal residential areas with inadequate infrastructure and poor living conditions, closely related to the Sustainable Development Goals (SDGs). Traditional methods for identifying urban villages are time-consuming and labor-intensive, but recent studies have used computer vision techniques to detect them more efficiently. However, existing methods either focus on simple classification or fail to provide accurate boundary information. UV-SAM addresses this by leveraging a small semantic segmentation model to generate mixed prompts, which are then fed into SAM for fine-grained boundary identification. Experimental results on two Chinese datasets show that UV-SAM outperforms existing baselines and provides insights into the development trends of urban villages over time. The framework also demonstrates the effectiveness of SAM for segmenting boundaries and highlights the potential of vision foundation models for sustainable cities. UV-SAM introduces a generalist-specialist framework that automatically generates four types of prompts for urban village identification. The framework uses a lightweight semantic segmentation model to generate prompts for SAM, which then identifies urban village boundaries. The model achieves significant performance improvements compared to state-of-the-art models and provides valuable insights into the spatial distribution and temporal trends of urban villages. The study also shows that urban villages are decreasing in number and area over time, which has implications for urban planning and governance. The dataset and code for UV-SAM are available at the provided link.