What problem does it solve? Setting up battery health prediction models from scratch is repetitive and error-prone, especially when every model must follow the same project conventions for versioning, serialization, and architecture. This Skill scaffolds a ready-to-use model class that matches the project's approved designs. ## Core Features & Use Cases - CNN-LSTM Template: Generates a PyTorch SOH predictor class taking (batch, 30, 3) telemetry windows of voltage, current, and temperature and outputting a State of Health percentage. - Isolation Forest Template: Generates a scikit-learn anomaly detector wrapper with fit, score, and classify methods mapping anomaly scores to Normal, Degrading, or Failed battery states. - Versioned Persistence: Both templates include save and load methods that embed version metadata and assert on version mismatch at load time, preventing silent inference errors. - Use Case: Run the scaffold with the argument SOHPredictor to create models/soh_predictor.py with the full CNN-LSTM architecture, or AnomalyDetector to create models/anomaly_detector.py with the Isolation Forest wrapper. ## Quick Start Ask the assistant to scaffold a new battery model named SOHPredictor or AnomalyDetector and it will create the corresponding Python file under the models directory.
