pandas-consistency

An agent skill that makes generated pandas code consistent: one canonical, modern (2.x) idiom set instead of a mix of legacy and current APIs.

What it does

pandas is stable and well known to models, but generated code drifts between API eras — chained indexing next to .loc, deprecated df.append and inplace=True next to method chains. This skill pins the canonical idioms (single-indexer writes, vectorized chained pipelines, explicit dtypes and merge validation, copy-on-write-safe patterns) and catalogs the pitfalls that produce silently wrong results, so every snippet written or reviewed follows the same rules.

Install

skills.sh

npx skills add guidogl/pandas-consistency

Claude Code plugin marketplace

/plugin marketplace add guidogl/pandas-consistency

Contents

• SKILL.md — the skill instructions: canonical idioms, pitfalls, version notes, workflow.
• references/pandas-patterns.md — deep reference: legacy→modern migration map, copy-on-write semantics, dtype/datetime/merge/groupby/I-O patterns, performance ladder.

License

Apache-2.0 — see LICENSE.txt.