code-review

name	code-review
description	Review data science and ML code for correctness, performance, and best practices. Use when receiving feedback, completing features, or before merging.

Code Review

Activate this skill when reviewing or receiving reviews on data/ML code.

When to Use

• Reviewing data pipeline implementations
• Checking ML training code for correctness
• Validating data preprocessing logic
• Ensuring reproducibility of experiments
• Reviewing notebook-to-production conversions

Checklist

Data Quality

• Input validation and schema checks
• Missing value handling documented
• Data leakage prevention (train/test separation)

ML Correctness

• Random seeds set for reproducibility
• Metrics computed on correct splits
• Hyperparameters documented and justified
• Baseline comparison included

Code Quality

• Functions are testable and focused
• Magic numbers replaced with named constants
• Error handling for data edge cases
• Type hints on public functions

Performance

• Vectorized operations over loops
• Appropriate data types (category, float32)
• Memory-efficient processing for large datasets

Rules

• Prioritize correctness over style
• Flag potential data leakage immediately
• Verify metrics are computed correctly
• Check for hardcoded paths or credentials
• Ensure experiments are reproducible