Whether leisure screen time (LST) is causally related to frailty remains uncertain. We used Mendelian randomization (MR) to examine the association of genetically proxied LST with frailty risk and to explore potential mediation through cardiometabolic factors. Summary-level genome-wide association study data for LST, frailty, and candidate mediators were obtained from large consortia. Two-sample MR was conducted using inverse-variance weighted (IVW) as the primary estimator, with MR-Egger, weighted median, weighted mode, MR-PRESSO, and radial MR used as complementary analyses. A two-step MR framework was used to assess mediation through body mass index (BMI), hypertension (HTN), coronary artery disease (CAD), and type 2 diabetes (T2D). Higher genetically predicted LST was associated with a higher risk of frailty after radial MR outlier correction (IVW OR = 1.157, 95% CI: 1.118-1.197, P = 5.42e-17), with broadly consistent directions across sensitivity estimators. Although heterogeneity was observed in the initial analysis, residual heterogeneity and directional pleiotropy were not evident after radial MR correction. In two-step MR analyses, BMI, HTN, CAD, and T2D showed separate, modest mediation signals, accounting for 11.4%, 12.7%, 7.2%, and 7.8% of the total effect, respectively. Genetic evidence was consistent with a potential causal association between higher LST and increased frailty risk. BMI, HTN, CAD, and T2D may each partly mediate this association, although the mediation estimates were modest, pathway-specific, and non-additive. Further studies with refined exposure phenotyping and interventional designs are needed to determine whether reducing screen-based sedentary behavior can help prevent frailty.