Skin malignancy is an important cause of mortality in the United Kingdom and is rising in incidence every year. Most skin cancer presents in primary care, and an important determinant of outcome is initial recognition and management of the lesion. Here we present an observational study of interobserver agreement using data from a population-based randomised controlled trial of minor surgery. Trial participants comprised patients presenting in primary care and needing minor surgery in whom recruiting doctors felt to be able to offer treatment themselves or to be able to refer to a colleague in primary care. They are thus relatively unselected. The skin procedures undertaken in the randomised controlled trial generated 491 lesions with a traceable histology report: 36 lesions (7%) from 33 individuals were malignant or pre-malignant. Chance-corrected agreement (κ) between general practitioner (GP) diagnosis of malignancy and histology was 0.45 (0.36–0.54) for lesions and 0.41 (0.32–0.51) for individuals affected with malignancy. Sensitivity of GPs for the detection of malignant lesions was 66.7% (95% confidence interval (CI), 50.3–79.8) for lesions and 63.6% (95% CI, 46.7–77.8) for individuals affected with malignancy. The safety of patients is of paramount importance and it is unsafe to leave the diagnosis and treatment of potential skin malignancy in the hands of doctors who have limited training and experience. However, the capacity to undertake all of the minor surgical demand works demanded in hospitals does not exist. If the capacity to undertake it is present in primary care, then the increased costs associated with enhanced training for general medical practitioners (GPs) must be borne.