Most mapped landslide inventories lack movement-type and trigger information, limiting their value for process-based research and hazard assessment. Here, we present Landsifier 2, an open-source Python library that enriches polygon inventories using three-dimensional geometry, topological descriptors, and configurable classification. Building on our existing trigger identification, this revamped library adds movement-type prediction, customizable classes, and geospatial outputs with class probabilities. Applications to independently labelled and previously undocumented inventories demonstrate its flexibility. Landsifier 2 provides a reproducible means of transforming landslide inventories into informative datasets for comparative geomorphology and hazard assessment.