Stain augmentation represents a key concept in deep learning to achieve model generalizability. However, creating realistic and sufficiently varied augmentations often requires the modeling of distribution and constraints of stain variation in a representative reference dataset. Access to such datasets is typically lacking, thus hampering the ability of research groups to achieve comprehensive augmentations. Rather than sharing datasets, which is infeasible due to privacy regulations, the current project thus explored the possibility of producing stain augmentation templates, which can be freely shared and reused to create augmentations even in the absence of a reference dataset. For that purpose, a large collection of Periodic acid-Schiff-stained images was gathered, spanning 14 different datasets with extensive stain variation. The images were utilized to extract (i) a stain style-related augmentation template using the RandStainNA tool and (ii) a stain vector-related template using a bespoke method based on cluster-based sampling of stain matrices. Downstream application of both templates demonstrated varied and realistic augmentations. By sharing these augmentation templates and code for reusing them, we hope to provide computational pathologist with the ability to achieve reliable and realistic augmentations without requiring access to extensive reference datasets. We believe that such a strategy would aid in establishing community standards for augmentation and would help in creating more generalizable deep learning models.