Data Analysis Skills — SQL, Pandas, Excel Analysis, Visualization (2026)

Data Analysis Skills sit between raw data and business decisions. They accept structured input (CSV files, database rows, Excel workbooks, JSON payloads) and produce answers — a query result, a cleaned dataset, a summary statistic, a chart, a forecast. Under the hood most of these Skills wrap well-known Python libraries (pandas, numpy, scipy, statsmodels, plotly, matplotlib) or delegate to a database engine directly.

The value the Skill layer adds is discipline — refusing to invent columns that aren't in the data, checking dtypes before applying operations that require specific types, flagging when a query result seems surprisingly large or small, and asking clarifying questions when the input is ambiguous ("which date column did you mean?"). Plain LLM output for data work is notoriously prone to hallucinated columns and quietly incorrect aggregations; Skills lock the discipline in.

What lives here vs. elsewhere: if the Skill's output is a data artifact (a query, a cleaned frame, a chart, a summary stat), it belongs here. If it produces a document that contains data (a PDF report with charts), that lives in Document Creation. If it fetches data from a live web source, that lives in Web & API. If it's an autonomous multi-step research workflow, see Research & Analysis.

• You have a CSV, Excel file, or database and want AI to answer specific questions about the data without hand-writing the queries.
• You need to clean messy data (date formats, encoding, dedupe, categorical values) before loading it into a downstream system.
• You're building a recurring report — same shape, different data — and want an AI to keep it updated without you rewriting the pipeline every week.
• You want AI-generated statistical work (t-tests, regressions, forecasts) with the assumption checks a human analyst would do.
• You need visualizations that are correct rather than beautiful-but-misleading — the Skill layer keeps chart choices honest.