School buildings being a critical social infrastructure, assessment of their seismic behaviour is of utmost importance in ensuring safe schooling facilities in locations of high seismicity. This study presents two important aspects in analysing any existing building stock for seismic behaviour: the development of an appropriate taxonomy system and an appropriate analytical method to conduct fragility assessment. A detailed desk study of existing schools’ databases and tailored field investigation in Guwahati, Assam, situated in India’s highest seismic zone, reveal that the majority of school buildings can be categorised within the confined masonry (CM) typology. This study discusses first, the addition to the World Bank promoted Global Library of School Infrastructure taxonomy of the specific category relating to CM as to include the buildings under study, which are non-engineered CM buildings with flexible roofs. Identifying the density of confinement and quality of connections as critical parameters for the seismic response of these buildings, varying seismic design levels are defined in relation to these indicators. Secondly, the paper presents an approach for carrying out nonlinear static pushover analysis of these buildings with flexible diaphragms and elaborates on the criteria adopted for determining the performance drift limits in buildings with varying levels of seismic design. Numerical analysis for the capacity assessment of selected index buildings is carried out using a commercial software that enables nonlinear extreme loading analysis. Different failure mechanisms as a function of the level of confinement are identified and the performance range for three damage states for three index buildings is obtained by using the N2 method. The study shows the influence of both choices of performance indicators and intensity measure on the resulting fragility functions. Given the consistency of the educational building stock in Guwahati, the results can be used for investment on retrofit decision making at regional level.