This deposit contains the supplementary materials and dataset metadata for the paper "Echoes of Abandonment: A Predictive Framework for Code Quality and Sustainability in Open-Source Repositories." The study addresses the systemic fragility of open-source software supply chains by proposing a longitudinal predictive framework that fuses socio-technical repository telemetry with structural code evolution metrics. We mine a curated dataset of 2,800 GitHub repositories, each with a minimum 36-month commit history, and engineer a 15-dimensional feature space spanning process variables (commit velocity, contributor distribution entropy, pull request resolution lag, bus factor) and product variables (code churn, cyclomatic complexity, code smell density, documentation ratio). A Temporal Fusion Transformer model is trained to forecast project abandonment risk and future code quality degradation up to six months in advance, achieving an F1-score of 0.84 on the primary classification task. Feature importance analysis reveals that contributor churn entropy and prolonged pull request stagnation are stronger leading indicators of project decay than static code complexity metrics. The deposit includes the curated repository list with anonymized identifiers, the full feature extraction and model training pipeline, pre-computed feature matrices for all 2,800 projects, and the trained model weights for reproducibility. These resources support the replication of our experiments and enable further research into data-driven open-source ecosystem health monitoring.