feature-engineering

name	feature-engineering
description	Feature engineering for data scientists — delegates to ds-feature-engineer agent. Use when building scikit-learn Pipelines, encoding categoricals, imputing missing values, scaling features, or creating lag/rolling features for time series.

Feature Engineering Command

Build preprocessing pipelines, encode features, and engineer inputs for ML models

Usage

/ds-feature-engineering <dataset-or-description>

Examples

/ds-feature-engineering data/train.csv
/ds-feature-engineering "Encode categoricals and impute nulls for churn model"
/ds-feature-engineering "Build ColumnTransformer for mixed numeric/categorical dataset"
/ds-feature-engineering "Create lag and rolling features for daily sales forecasting"

---

What This Command Does

1. Invokes the ds-feature-engineer agent
2. Audits column types, cardinality, and null rates
3. Loads KB patterns from scikit-learn and pandas domains
4. Generates:
   • ColumnTransformer + Pipeline definition preventing data leakage
   • Encoder selection (OrdinalEncoder vs OneHotEncoder vs TargetEncoder)
   • Imputer selection (SimpleImputer vs KNNImputer)
   • Scaler selection (StandardScaler vs RobustScaler)
   • Feature selection step (SelectKBest or RFE)

Agent Delegation

Agent	Role
ds-feature-engineer	Primary — Pipeline, encoding, imputation, scaling, selection
ds-time-series-analyst	Escalation — when lag/rolling features are needed
ds-eda-analyst	Escalation — when column audit is needed first

KB Domains Used

• scikit-learn — Pipeline, ColumnTransformer, encoders, imputers, scalers
• pandas — Feature creation with DataFrame operations
• time-series — Lag features, rolling statistics, date/time features

Output

The agent generates a complete scikit-learn Pipeline with ColumnTransformer, ready to fit on training data without leakage.