Pandas Pro steers Claude through DataFrame work in five steps: inspect dtypes, memory and nulls, design a vectorized transformation, implement it with method chaining, validate shapes and null counts, then optimize memory.
The skill carries code patterns for groupby aggregation, merges with validate and indicator, forward-fill and interpolation, resampling, pivot tables and categorical downcasting.
What You Get
- A five-step workflow with inspection and assertion snippets for dtypes, memory, nulls and row counts
- Code patterns for vectorized assignment, safe subsetting with copy, named-aggregation groupby, validated merges, missing values, resampling, pivots and memory reduction
- Reference files for DataFrame operations, data cleaning, aggregation, merging and performance optimization
- Rules against iterrows, chained indexing and the deprecated ix and append methods
- An output template with commented code, performance notes and validation checks
Common Use Cases
- Joining two tables on several keys and counting the unmatched rows
- Filling gaps in a price series with forward-fill and interpolation
- Resampling event timestamps into daily totals
- Building a pivot table of revenue by region and product line
- Cutting the memory use of a large DataFrame with categorical and downcast types
What Makes It Unique
The skill is built around checks. Merges use validate and indicator to expose key problems, every transform ends with assertions on shape, columns and nulls, and missing values are handled explicitly and never dropped silently. Its before and after snippets show the vectorized form next to the loop it replaces.
Under The Hood
The skill is instructions and reference markdown that Claude loads by topic; it ships no scripts. Following it, Claude writes and edits Python analysis code in your project and may run it against your data files to print dtypes, memory use and validation results. It does not call outside services.