Guide

Best agent Skills for data analysis

The best agent Skill for data analysis depends on where the data lives. For spreadsheet analysis — formulas, pivots, formatted reports — the pick is xlsx, Anthropic-official. For multi-source research analysis use deep-research; for primary-source financial or scientific data use valyu; for custom interactive charts use d3js-skill.

A data Skill changes how the agent handles data, not whether it can reason about it. Match the Skill to the source — a spreadsheet, the open web, an authoritative feed, or a chart you need to draw — and the analysis stops being improvised.

Top pick

xlsx

Anthropic-built. Reads and writes real .xlsx with formulas, pivot tables, and charts, so the agent analyzes tabular data in the format analysts already work in rather than a flattened CSV. Bundled scripts keep the numbers deterministic.

Best when: The dataset is a spreadsheet and you want models, pivots, or a formatted report back — the most common data-analysis case.

Ranked

The shortlist

Ranked by fit for the common case, not by raw popularity. Each is a catalogued, verified Skill — open the skill page for the composite score, license, and full install commands.

SkillSourceBest whenVerified agents
xlsxOfficialThe dataset is a spreadsheet and you want models, pivots, or a formatted report back — the most common data-analysis case.Claude Code, Claude.ai, Claude API
duckdb-skillsVerifiedAnalytical SQL over local data files (CSV, Parquet, Excel) — the 1GB-to-100GB range a spreadsheet chokes on.Claude Code
deep-researchOfficialMarket sizing, competitive analysis, or any question whose data lives across many web sources.Claude Code, Antigravity
valyuVerifiedFinancial, scientific, or regulatory analysis that must cite primary sources.Claude Code, Claude API, Cursor, Codex
d3js-skillCommunityYou need a custom, interactive visualization, not a default bar chart.Claude Code, Cursor
scientific-skillsCommunityComputational-science workflows where reproducibility matters more than speed.Claude Code

See every entry, ranked by composite score, in the Data & Analytics category.

Why these

Pick by pick

  1. 1
    xlsx

    Anthropic-built. Reads and writes real .xlsx with formulas, pivot tables, and charts, so the agent analyzes tabular data in the format analysts already work in rather than a flattened CSV. Bundled scripts keep the numbers deterministic.

  2. 2
    duckdb-skills

    DuckDB's own skill. In-process analytical SQL directly over CSV, Parquet, JSON, and Excel — the agent queries and joins raw data files without loading a warehouse, which is the fastest path for datasets too big for a spreadsheet.

  3. 3
    deep-research

    Anthropic-built. Fans out web searches, fetches sources, adversarially verifies claims, and returns a cited report — analysis of the outside world rather than a local file.

  4. 4
    valyu

    Published by Valyu. Web search plus 36 authoritative sources — SEC filings, PubMed, ClinicalTrials.gov, FRED — with a 79% FreshQA score, so analysis is grounded in primary data instead of blog summaries.

  5. 5
    d3js-skill

    D3.js patterns — scales, force-directed graphs, choropleths — so the agent produces publication-grade interactive charts instead of stock chart-library output.

  6. 6
    scientific-skills

    scikit-learn patterns and scientific-dataset APIs with reproducibility helpers — for analysis that has to be re-runnable and defensible.

FAQ

Common questions

What is the best Claude skill for data analysis?

For the common case — analyzing a dataset that is already a spreadsheet — it is xlsx, Anthropic's official Skill. It reads and writes real .xlsx with formulas, pivot tables, and charts, and ships bundled scripts so the numbers stay deterministic instead of being re-derived each prompt. If your data is spread across the web rather than a file, deep-research is the better pick.

Which skill is best for research-driven analysis?

deep-research, built by Anthropic. It fans out web searches, fetches sources, adversarially verifies claims, and returns a cited report — so it is analysis of the outside world (market sizing, competitive scans) rather than a local file. For analysis that must cite primary sources like SEC filings, PubMed, or FRED, use valyu instead, which is grounded in 36 authoritative data feeds.

What is the best skill for data visualization?

d3js-skill. It gives the agent D3.js patterns — scales, force-directed graphs, choropleths — so it produces publication-grade interactive charts rather than stock chart-library output. It is verified on Claude Code and Cursor.

Should I use a skill or an MCP server for data analysis?

A Skill encodes the analysis method — how to build the model, how to structure the research, how to draw the chart. An MCP server connects the agent to a live data source like a warehouse, BigQuery, or a Postgres database. Most real analysis uses both: an MCP pulls the rows, a Skill knows what to do with them. See skills vs subagents vs MCP vs plugins for the distinction.

Other surfaces

Same job, different tooling

Related

More best-skill guides

New to Skills? What is a Claude skill · How to install · Skills vs subagents vs MCP vs plugins