Menu

Individual skill

pandas-expert

Writes correct, vectorized pandas for cleaning, joining, reshaping, and aggregating tabular data, with validated joins and deliberate dtype and missing-data handling. Use when someone asks "why did my merge duplicate rows", "how do I clean this CSV in pandas", "my groupby numbers look wrong", "this apply is too slow", or is transforming a DataFrame for analysis or a pipeline. Do NOT use for first-pass profiling of an unfamiliar dataset - use eda-playbook instead; for data too large for one machine use spark-jobs; for answering business questions directly in SQL use sql-to-insights.

skills/pandas-expert/SKILL.md

Read pinned source

Source commit 97a13607d669157a4a16d5ba79b77f208319f7b9

What was checked

Matches the observed repository HEAD
Yes
Scanner rules are current
Yes
Repository scan complete
No
Skill scan complete
No
Observed behavior
Instructions only
Last checked
Nov 6, 2026

These are observations at the recorded commit, not a guarantee of safety. Review the source before running it.

Install this skill

Choose an agent in your Stack or copy a command below. Commands fetch the pinned commit and copy only this skill directory. Existing destinations are refused. Copying a command does not install anything.

An install command is unavailable: current rules, complete repository and skill scans, an eligible behavior level and a permitted license are required.

Repository context

Repository stars (not skill ratings)
24
Repository last updated
Oct 13, 2026
License
MIT

Observed agent integration: No specific integration evidence. Standard SKILL.md installation does not imply verified compatibility.

Files and excerpt

File metadata and permitted, redacted excerpts come from the recorded scan. Source links are pinned to that commit.

Browse files
Read excerpt

# Pandas Expert

Most pandas bugs are silent: a join that fans out rows, a dtype that upcast to object, a NaN that vanished from a sum. The output looks plausible and is wrong. This skill produces transformations that validate themselves - every join asserted, every coercion counted, every aggregate reconciled - so errors surface at the line that caused them instead of in a dashboard three weeks later.

## Operating procedure

Follow the steps in order. Types must be fixed before cleaning, cleaning before joining, joining before aggregating - each later step silently produces wrong numbers if an earlier one was skipped.

### Step 1: gather inputs

Before writing any transformation, establish:

- The grain of each table: what one row represents. If the user cannot state it, derive it with `df.duplicated(subset=candidate_keys).sum()` and label the result a guess until confirmed.
- The expected output: shape, grain, and one or two spot-check numbers the user already trusts (last month's total, a known customer's count).