| name | engineering-features-for-machine-learning |
| description | Create, encode, transform, and select features before model fitting.
Use when the user needs feature engineering decisions or implementation, not final training ownership or leakage auditing.
|
| allowed-tools | Read, Write, Edit, Grep, Glob, Bash(cmd:*) |
| version | 1.0.0 |
| author | Jeremy Longshore <jeremy@intentsolutions.io> |
| license | MIT |
Feature Engineering Toolkit
Use this skill when the main question is how to improve or restructure the input features.
Overview
This skill covers feature creation, encoding, scaling coordination, and feature selection before the model is finalized.
When to Use This Skill
- Creating derived variables, interaction terms, bins, encodings, or date-based features
- Selecting or pruning features before training
- Reworking feature representations to fit model assumptions or data geometry
Not For / Boundaries
- Full training runs and benchmark ownership: use
training-machine-learning-models
- Post-hoc interpretation of a trained model: use
feature-importance-analyzer
- Leak checking across the preprocessing order: use
ml-data-leakage-guard
Typical Outputs
- Candidate feature set changes
- Implementation notes for encoders, scalers, and selectors
- Rationale for what to keep, drop, or combine
Related Skills
data-normalization-tool for scaling-only questionsml-data-leakage-guard before accepting the engineered pipeline
