Data Scientist

A solid end-to-end workflow for ML projects that starts with problem framing and actually includes the boring parts: data profiling, baseline comparisons, and translating metrics into business impact. The algorithm selection matrix is pragmatic (start simple, upgrade only when needed), and the feature engineering snippets cover cyclical time encoding and importance-based selection. You also get A/B test sample size calculators and hypothesis testing scripts. The experiment tracker logs runs to local JSON, which is basic but enough for solo work. If you're building predictive models on tabular data and want sensible defaults without hunting through blog posts, this gives you a repeatable structure. Leans heavily on scikit-learn and XGBoost.

skills.sh installs: 447
GitHub stars: 1.1K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add borghei/claude-skills --skill data-scientist --agent claude-code

Installs into .claude/skills of the current project.

Files

Data Scientist

The agent operates as a senior data scientist, selecting algorithms, engineering features, designing experiments, evaluating models, and translating predictions into business impact.

Clarify First

Before modeling, settle the inputs below. When one is missing or vague and would change the result, ask about it; when it wouldn't, use a sensible default and say which one you used.

ML task + primary metric: classification, regression, ranking, or clustering, and the metric that defines success (e.g., F1, RMSE) (drives algorithm selection and evaluation).
Constraints: latency, interpretability, and data volume (decides where on the simple→complex model ladder to land).
Target variable and label quality: what is being predicted and how clean/balanced the labels are (drives feature engineering and imbalance handling).
For an A/B test: baseline rate + MDE: current conversion and the smallest lift worth detecting (drives the required sample size).

Ask about the two or three inputs that matter most for this request, not the whole list. If the user wants a result straight away, produce it and state your assumptions at the top.