We study how local topological structure predicts future fine-scale change in Life-like cellular automata. Across Game of Life and HighLife on periodic grids, we compare candidate fine-scale predictors of subsequent component-level change and identify a local topological fragility hierarchy. The strongest predictors are local component counts and embedded isolated-cell motifs, with temporally stable negative response slopes across tested conditions and horizons. We separate prestate predictive information from outcome leakage, identify analytic and empirical mechanism carriers, and test transfer of the standardized mechanism direction across grid size, rule, density, and horizon. The results show that description-dependent change can be made quantitatively precise within a fully controlled discrete dynamical system, while remaining explicitly scoped to the tested Life-like rules, densities, and periodic finite grids.