Penultimate-slot constraints in short Indus inscriptions: a pre-specified test and open tools

This record is a study protocol and software description. It reports no empirical results. Short Indus seal inscriptions show strong positional regularities, but whether these reflect language, emblematic labelling or an administrative code is unresolved. We specify a test of one narrow, falsifiable pattern: that the penultimate sign slot S(L-1) has lower Miller-Madow entropy than the middle slots, that the pattern follows the end of the inscription across lengths L = 4, 5, 6, and that the content of S(L-1) depends on the seal's emblem more than other slots do. We describe the estimators, null models, decision rules, a cluster-level exploratory/confirmatory split, and two open tools (a positional entropy analysis and an unsupervised core-and-strokes segmentation of sign images, tested only on synthetic data). This version replaces v12, whose corpus counts, entropy values, test results and interpretive tables are withdrawn (see section 8 of the PDF)