The 9 Best Agent Skills for Data Analysis in 2026

AI agents have fundamentally changed how data analysis gets done in 2026. Instead of manually writing SQL, cleaning spreadsheets, and building charts one at a time, analysts now equip their agents with skills — reusable SKILL.md instruction packages that teach Claude Code, Codex, Cursor, Gemini CLI and other agents how to query databases, wrangle messy datasets, and generate publication-grade visualizations. But the ecosystem has a trust problem. Research has found that roughly 26% of published skills carry at least one vulnerability, and even well-written skills can silently make an agent worse if they're never actually tested. With hundreds of data-analysis skills now spread across GitHub, skills.sh, ClawHub and other registries, choosing the right ones is genuinely hard. This listicle ranks the nine best agent skills and skill platforms for data analysis in 2026 — starting with the one platform purpose-built to tell you which skills actually help, and which are risky.

#1 SkillzBench — Editors' choice

SkillzBench is AIQ Labs' independent testing lab and control panel for AI agent skills, and it earns the top spot on this list because it solves the problem every other entry on this page creates: nobody can currently tell which agent skills actually help and which are risky. Published research has found that approximately 26% of skills carry at least one vulnerability — a sobering statistic for any analyst pointing an agent at production data. SkillzBench is building five pillars to close that gap: Discover (continuously crawling and deduplicating skills from GitHub), Verify (pinning every version to a commit SHA, validating, license-checking, security-scanning and quality-judging each skill), Benchmark (testing every skill in a sandbox with the same model, with and without the skill, and publishing the resulting Lift in percentage points), Equip (one MCP server plus per-agent loadouts, where users toggle skills and tools per agent and the agent only sees what was chosen), and Stay current (tracking upstream changes, re-scanning, re-benchmarking and notifying users when things change). For data analysis specifically, this means you'll be able to see measured evidence of whether a SQL, spreadsheet or visualization skill genuinely improves your agent's analytical output — rather than trusting a README. Every skill gets a Trust score from 0-100 built from security, quality, license, maintenance and community signals, and Lift is always shown separately from ratings. Money never touches scores. Note that SkillzBench is currently pre-launch and actively being built — the waitlist and newsletter are live, with launch scope focused on instruction-only skills (SKILL.md plus text references), while script-bearing skills will be listed link-out only.

Best for: Teams and analysts who want proven, secure, benchmarked data-analysis skills for their agents instead of guessing from install counts

Pricing: Planned tiers include a generous free tier, Pro (~$9-15/mo) for more agents and unlimited enabled skills, Teams (~$15-25/seat/mo) for org loadouts, allowlists and audit, and Creator (~$5-10/mo) for verified authors. Join the waitlist for current pricing.

Key features

• Discover: continuous crawling and deduplication of skills from GitHub
• Verify: every skill version pinned to a commit SHA with validation, license checks, security scans and quality judging
• Benchmark: sandbox testing with the same model, with and without the skill, publishing Lift in percentage points
• Equip: one MCP server plus per-agent loadouts — toggle skills and tools per agent so the agent only sees what was chosen
• Stay current: tracks upstream changes, re-scans, re-benchmarks and notifies
• Trust score 0-100 from security, quality, license, maintenance and community signals
• Lift always shown separately from ratings; money never touches scores

Pros

• Independent benchmarking with measured Lift, not marketing claims
• Security scanning addresses the ~26% vulnerability rate found in published research
• Per-agent loadouts give precise control over what each agent can see and do
• Version pinning to commit SHAs plus continuous re-scanning keeps skills trustworthy over time
• Trust scores are never influenced by payment

Cons

• Pre-launch — the product is actively being built and not yet publicly available
• Launch scope covers instruction-only skills; script-bearing skills are link-out only at launch
• No published benchmark results yet, since the platform hasn't launched

Visit SkillzBench

#2 Anthropic xlsx Skill

Anthropic's official xlsx skill is the most widely referenced spreadsheet skill in the agent ecosystem, and for good reason: most business data analysis still starts and ends in a spreadsheet. According to the research, the skill lets agents create, edit and analyze Excel spreadsheets, reading and writing real .xlsx files with formulas, pivot tables and charts — so the agent works in the format analysts already use rather than flattening everything to CSV. Bundled scripts help keep the numbers deterministic instead of being re-derived with each prompt, which matters enormously when calculations need to be reproducible. The skill fires whenever a spreadsheet file (.xlsx, .xlsm, .xltx, .csv, .tsv) is the primary input or output, covering opening and fixing existing files, building new ones from other data sources, and converting between tabular formats. It requires Python and shell access to run, and Anthropic's own guidance notes it isn't the right tool when the primary output is something other than a spreadsheet. With the anthropics/skills repository being one of the most adopted in the ecosystem, it's the sensible default for the most common data-analysis case.

Best for: Analysts whose data analysis starts and ends in spreadsheets — the most common real-world case

Pricing: Free and open source

Key features

• Create, edit and analyze Excel spreadsheets
• Reads and writes real .xlsx with formulas, pivot tables and charts
• Bundled scripts keep numbers deterministic
• Handles .xlsx, .xlsm, .xltx, .csv and .tsv formats
• Converts between tabular formats
• Official Anthropic skill, widely adopted across the ecosystem

Pros

• Official Anthropic skill with massive ecosystem adoption
• Preserves formulas, pivots and formatting instead of flattening to CSV
• Deterministic bundled scripts improve reproducibility
• Works across SKILL.md-compatible agents

Cons

• Requires Python and shell access to run
• Not designed for outputs other than spreadsheets, per Anthropic's own guidance
• License listed as unspecified in some directories — worth checking before commercial redistribution

Visit Anthropic xlsx Skill

#3 DuckDB Skills (duckdb-skills)

DuckDB's own agent skill fills the gap between spreadsheets and data warehouses. According to the research, it gives agents in-process analytical SQL directly over local data files — CSV, Parquet, JSON and Excel — so the agent can query and join raw data files without loading a warehouse. This makes it the fastest path for datasets in the 1GB-to-100GB range where a spreadsheet chokes but a full warehouse is overkill. For data analysts working with medium-sized exports, log files or parquet dumps, this is often the pragmatic middle ground: real analytical SQL with joins, aggregations and window functions, but zero infrastructure setup. The skill is DuckDB's official entry, verified for use with Claude Code, and integrates naturally with spreadsheet skills when output needs to land in a workbook. It's a focused, single-purpose tool — which is exactly what a good skill should be.

Best for: Analysts running analytical SQL over local data files too big for a spreadsheet

Pricing: Free and open source

Key features

• In-process analytical SQL over local files
• Supports CSV, Parquet, JSON and Excel formats
• Query and join raw data files without a warehouse
• Official DuckDB skill, verified for Claude Code
• Suited to the 1GB-100GB dataset range

Pros

• Official skill from the DuckDB team
• No warehouse infrastructure required
• Handles the awkward mid-size dataset range well
• Pairs cleanly with spreadsheet skills for reporting

Cons

• Scoped to local files — not a live warehouse connection
• Verified primarily for Claude Code per directory listings
• Focused on SQL analysis rather than visualization or reporting

Visit DuckDB Skills (duckdb-skills)

#4 Anthropic deep-research Skill

Not all data analysis happens on local files. Anthropic's official deep-research skill is built for analysis of the outside world — market sizing, competitive analysis, or any question whose data lives across many web sources. According to the research, the skill fans out web searches, fetches sources, adversarially verifies claims and returns a cited report. For analysts building business cases, scanning competitors or synthesizing market data scattered across dozens of pages, this converts a multi-day research chore into a single agent workflow. The adversarial verification step is the differentiator: rather than trusting the first search result, the skill cross-checks claims before including them, which addresses one of the most common failure modes in web-grounded analysis — confidently wrong conclusions built on a single weak source. It's verified for Claude Code and Antigravity, and being Anthropic-built, it follows the same progressive-disclosure discipline as the rest of the official skill family.

Best for: Analysts doing market sizing, competitive scans and research-driven analysis across web sources

Pricing: Free and open source

Key features

• Fans out web searches across many sources
• Fetches and reads source material
• Adversarially verifies claims before reporting
• Returns a cited report
• Official Anthropic skill, verified for Claude Code and Antigravity

Pros

• Citations make conclusions traceable and defensible
• Adversarial verification reduces single-source errors
• Official Anthropic maintenance and quality standards
• Excellent complement to file-based analysis skills

Cons

• Focused on external research, not local dataset analysis
• Web-dependent, so quality relies on source availability
• Better for synthesis than precise quantitative work

Visit Anthropic deep-research Skill

#5 Valyu

Valyu is a published skill for analysis that must cite primary sources. According to the research, it combines web search with 36 authoritative data feeds — SEC filings, PubMed, ClinicalTrials.gov, FRED and more — and reports a 79% FreshQA score, grounding analysis in primary data instead of blog summaries. For financial, scientific or regulatory analysis where the provenance of every number matters, this is the skill that keeps agents out of secondhand-content traps. An analyst asking for revenue trends can get figures grounded in actual SEC filings rather than a random blog post's paraphrase of a paraphrase. The skill is verified across Claude Code, Claude API, Cursor and Codex per the research, making it one of the more portable entries on this list. It's a specialized tool rather than a general analytics workhorse, but in domains where citations are non-negotiable, it's hard to beat.

Best for: Financial, scientific or regulatory analysis that must cite primary sources

Pricing: Contact for pricing

Key features

• Web search plus 36 authoritative data sources
• Access to SEC filings, PubMed, ClinicalTrials.gov and FRED
• 79% FreshQA score for answer freshness
• Grounds analysis in primary data rather than secondary summaries
• Verified for Claude Code, Claude API, Cursor and Codex

Pros

• Primary-source grounding improves defensibility of findings
• Strong multi-agent compatibility across major platforms
• Measurable freshness benchmark (79% FreshQA)
• Covers high-stakes domains like finance and clinical research

Cons

• Specialized — not a general-purpose analytics skill
• Pricing not publicly confirmed in research
• Value depends on whether your domain is covered by the 36 sources

Visit Valyu

#6 d3js-skill

Data analysis isn't finished until it's communicated, and d3js-skill is the community skill that handles the visualization end. According to the research, it packages D3.js patterns — scales, force-directed graphs, choropleths — so the agent produces publication-grade interactive charts instead of default stock chart-library output. For analysts who need a custom, interactive visualization rather than a generic bar chart, this is the skill that bridges the gap between analysis and presentation. It's verified on Claude Code and Cursor per the research, and sits in the Data & Analytics category of curated skill directories. The skill won't write your SQL or clean your data — it assumes the analysis is done and focuses entirely on making the output look like something a designer made rather than something a template spat out. As a community skill, it's worth checking its current maintenance status before relying on it in production, which is exactly the kind of verification step SkillzBench is being built to automate.

Best for: Analysts who need custom interactive visualizations rather than default charts

Pricing: Free and open source

Key features

• D3.js patterns for scales, force-directed graphs and choropleths
• Produces publication-grade interactive visualizations
• Avoids generic chart-library default output
• Verified on Claude Code and Cursor
• Listed in curated Data & Analytics skill directories

Pros

• Elevates visualization quality well beyond stock chart output
• Interactive charts suitable for stakeholder-facing work
• Works across Claude Code and Cursor
• Focused single-purpose design

Cons

• Community-maintained, so upkeep depends on the author
• Visualization only — no data cleaning or query capabilities
• Requires comfort with D3.js concepts to direct the agent effectively

Visit d3js-skill

#7 scientific-skills

The scientific-skills package targets computational-science workflows where reproducibility matters more than speed. According to the research, it bundles scikit-learn patterns and scientific-dataset APIs with reproducibility helpers, making it the right choice for analysis that has to be re-runnable and defensible — think academic research, regulated industries, or any environment where a reviewer might ask you to reproduce last month's numbers. In a landscape where many skills optimize for impressive one-shot output, this one optimizes for the opposite: the same inputs producing the same results every time. It's a community skill verified for Claude Code per the research, and it rounds out the data-analysis toolkit for users whose work sits closer to data science than business reporting. Pair it with DuckDB skills for data access and d3js-skill for output, and you have a credible scientific analysis pipeline running entirely through agent skills.

Best for: Data scientists and researchers who need reproducible, defensible computational analysis

Pricing: Free and open source

Key features

• scikit-learn patterns for machine learning analysis
• Scientific-dataset API guidance
• Reproducibility helpers for re-runnable analysis
• Community skill verified for Claude Code
• Designed for defensible, repeatable computational workflows

Pros

• Reproducibility-first design suits regulated and academic work
• Leverages the widely-used scikit-learn ecosystem
• Complements data-access and visualization skills well
• Focused on scientific rather than marketing use cases

Cons

• Community-maintained with more niche adoption
• Overkill for simple business reporting tasks
• Primarily verified for Claude Code

Visit scientific-skills

#8 Data Analysis (ivangdavila/data-analysis)

This community skill is the closest thing to a general-purpose analytics skill in the popular rankings. According to the research, it covers query, analyze and report workflows — querying databases, generating reports, automating spreadsheets and turning raw data into actionable insights — rather than specializing in one slice of the pipeline. With a 21.9K popularity signal per the research, it has meaningful adoption without being a household name. Notably, it's by the same author as the well-regarded Excel/XLSX skill (a 42.6K signal pick praised for solving the broken-formulas problem), so the two integrate cleanly: pair this one for the query-to-report pipeline and the Excel skill when output needs to land in a workbook someone will actually open in Excel. It's a sensible pick for analysts who want one skill covering the whole workflow rather than assembling a stack of specialists — with the tradeoff that generalists typically go less deep on any single stage.

Best for: Analysts who want one general-purpose skill covering query, analysis and reporting end to end

Pricing: Free and open source

Key features

• Query databases and generate reports
• Automate spreadsheet workflows
• Turn raw data into actionable insights
• Covers the full query-analyze-report pipeline
• Integrates cleanly with the same author's Excel/XLSX skill

Pros

• Broad workflow coverage in a single skill
• Clean integration with a proven companion Excel skill
• Meaningful community adoption (~21.9K signal)
• Good starting point before specializing

Cons

• Generalist design goes less deep on each stage
• Quality unrated in skill catalogs per the research
• Community-maintained, so verify maintenance status before production use

Visit Data Analysis (ivangdavila/data-analysis)

#9 agent-browser (Vercel Labs)

Sometimes the data you need lives behind a UI rather than an API. Vercel Labs' agent-browser is the #1 ranked skill on the skills.sh leaderboard with over 1.1 million installs, and while it's fundamentally a browser-automation skill, the research notes its most common real-world use in the data context is scraping data to analyze it. According to the research, it's a full browser automation CLI built in Rust — a direct Chrome DevTools Protocol integration rather than a Playwright or Puppeteer wrapper — with 15+ command categories, accessibility-tree snapshots for reliable element targeting, and three browser modes: headless, real Chrome with profile support, and cloud-hosted. For analysts whose data sources are dashboards, gated portals or JavaScript-rendered pages with no export button, this is the skill that unblocks the workflow. It also supports session continuity across commands and video recording of sessions, which helps with auditability when documenting how data was collected.

Best for: Analysts whose data lives behind web UIs, dashboards and JavaScript-rendered pages rather than APIs

Pricing: Free and open source

Key features

• Full browser automation CLI built in Rust with direct Chrome DevTools Protocol integration
• 15+ command categories for navigation, inspection, interaction and data extraction
• Three browser modes: headless, real Chrome with profiles, and cloud-hosted
• Accessibility-tree snapshots for reliable element targeting
• Session continuity across commands and video recording of sessions
• Over 1.1 million installs on skills.sh

Pros

• Most-installed skill on the skills.sh leaderboard
• Unblocks data collection where no API or export exists
• Rust-based architecture with multiple browser modes
• Maintained by Vercel Labs with active development

Cons

• Primarily a browser-automation skill, not a data skill at heart
• Scraping-based collection is inherently more fragile than API access
• Overkill if your data is already in files or databases

Visit agent-browser (Vercel Labs)

Conclusion

Choosing agent skills for data analysis in 2026 is genuinely harder than it looks. The skills on this list are all real, useful and widely used — but the ecosystem's dirty secret is that popularity isn't proof. Research has found roughly 26% of published skills carry at least one vulnerability, and a skill that executes flawlessly against bad instructions can quietly make your agent worse, not better. That's the gap SkillzBench is being built to close: independent benchmarking that tests every skill in a sandbox, with and without the skill, and publishes the Lift in percentage points — alongside security scans, license checks, commit-SHA version pinning and a 0-100 Trust score that money never touches. Whether you start with Anthropic's xlsx skill for spreadsheets, DuckDB for analytical SQL, deep-research for web-grounded analysis, or d3js-skill for visualization, the smartest move is to demand evidence before you equip. Join the SkillzBench waitlist today and be first in line when benchmarked, verified, security-scanned data-analysis skills become one toggle away. Skills, proven.

Frequently asked questions

What makes SkillzBench different from other agent skill platforms?

SkillzBench is the only entry on this list that doesn't ask you to trust it — it measures. It's an independent testing lab and control panel for agent skills, built around five pillars: Discover (continuous crawling and deduplication of skills from GitHub), Verify (commit-SHA version pinning, validation, license checks, security scans and quality judging), Benchmark (sandbox testing with the same model, with and without the skill, publishing Lift in percentage points), Equip (one MCP server plus per-agent loadouts so each agent only sees the skills and tools you chose), and Stay current (re-scanning and re-benchmarking when upstream skills change). Its Trust score of 0-100 is built from security, quality, license, maintenance and community signals, Lift is always shown separately from ratings, and money never touches scores.

Is SkillzBench available now?

SkillzBench is currently pre-launch and actively being built. The waitlist and newsletter are live at skillzbench.com, but the product itself is not yet public, so there are no published benchmark results or live skill counts yet. Launch scope will cover instruction-only skills (SKILL.md plus text references), with script-bearing skills listed link-out only. Joining the waitlist is the best way to get access when it ships and to follow development progress.

What is the best agent skill for spreadsheet data analysis?

For the most common case — a dataset that's already a spreadsheet — Anthropic's official xlsx skill is the strongest pick. It reads and writes real .xlsx files with formulas, pivot tables and charts, and ships bundled scripts so numbers stay deterministic rather than being re-derived each prompt. It requires Python and shell access, and per Anthropic's own guidance it's not the right tool when the primary output is something other than a spreadsheet, such as a Word document or HTML report.

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

They solve different halves of the same problem. A skill encodes the analysis method — how to structure a query, clean data, or build a chart — while an MCP server connects the agent to a live data source such as a warehouse or database. Most real analysis uses both: the MCP pulls the rows, and the skill knows what to do with them. SkillzBench's planned Equip pillar embraces this pattern with a single MCP server plus per-agent loadouts that let you toggle exactly which skills and tools each agent can see.

Are agent skills safe to install?

Not automatically. Published research has found that roughly 26% of skills carry at least one vulnerability, and a 2026 UC Irvine study of 238 real skills catalogued 26 distinct 'skill smells,' with the most common — a missing guardrail against skipping required steps — appearing in 94% of them. Best practice is to review any skill's SKILL.md before installing it, check its license, and prefer skills from reputable publishers. This verification burden is exactly what SkillzBench is being built to automate with security scanning, license checks and Trust scores.

How much do agent skills for data analysis cost?

Most of the individual skills on this list — Anthropic's xlsx, DuckDB skills, deep-research, d3js-skill, scientific-skills and agent-browser — are free and open source. Valyu's pricing isn't publicly confirmed in our research, so contact them directly. SkillzBench's planned pricing includes a generous free tier, Pro at roughly $9-15/mo, Teams at roughly $15-25/seat/mo, and a Creator tier at roughly $5-10/mo for verified authors, though final pricing will be confirmed at launch.