scikit-learn-consistency

An agent skill that makes generated scikit-learn code consistent: one canonical, modern (1.x) idiom set — Pipeline-first and leak-free — instead of a mix of legacy and current APIs.

What it does

scikit-learn is stable and well known to models, but generated code drifts between API eras — manual fit-everything-then-split preprocessing next to Pipelines, the removed get_feature_names next to get_feature_names_out, OneHotEncoder(sparse=...) next to sparse_output=. This skill pins the canonical idioms (split first, every learned step inside a Pipeline/ColumnTransformer, CV over the whole pipeline, explicit metrics and random_state) and catalogs the pitfalls — above all data leakage — that produce silently inflated scores.

Install

skills.sh

npx skills add guidogl/scikit-learn-consistency

Claude Code plugin marketplace

/plugin marketplace add guidogl/scikit-learn-consistency

SKILL.md — the skill instructions: canonical idioms, pitfalls, version notes, workflow.
references/scikit-learn-patterns.md — deep reference: legacy→modern migration map, the leakage taxonomy, estimator contract, Pipeline/ColumnTransformer, CV/tuning, metrics, randomness, persistence.

License

Apache-2.0 — see LICENSE.txt.

scikit-learn-consistency

Popularity

What's Inside

README