Breast cancer has the highest incidence among all cancers in women, and the prognosis of breast cancer is strongly linked to the stage of the disease. As one of the components found in liquid biopsy samples, exosomes are membranous vesicles that are actively secreted by living cells. Therefore, the key genes in exosomes may serve as biomarkers for predicting the prognosis of breast cancer patients. In this study, 128 blood samples collected from breast cancer patients at Fudan University Shanghai Cancer Center between June 2018 and March 2019 were subjected to transcriptome sequencing, and the resulting dataset was used as the training dataset. A LASSO regression model was employed for screening prognostic genes. Additionally, 131 patient samples from February 2020 to February 2022 were collected to establish the validation dataset. The corresponding phenotypes and mechanisms of the key genes were confirmed by CCK8 cell proliferation, colony formation, EdU cell proliferation, flow cytometry, transwell cell migration, scratch assay, animal study and RNA-seq assays. Eleven differentially expressed genes tended to increase from the benign stage to the late stage of breast cancer. Five genes were further identified by LASSO regression analysis to establish a prognostic model. The time-dependent receiver operating characteristic (ROC) curves revealed area under the curve (AUC) values of 0.858 for the 1-year follow-up and 0.772 for the 2-year follow-up. The time-dependent ROC curve of the validation dataset indicated an AUC value of 0.840 for the 1-year follow-up. JCAD, a gene closely associated with prognosis, was selected for further investigation. The experimental results demonstrated that JCAD may activate the Wnt/β-catenin pathway by increasing FZD1 expression, thereby promoting the EMT process and breast cancer progression. Exosomal JCAD, as a prognostic marker, plays an important role in the diagnosis and treatment of breast cancer.