The 7 Best Claude Skills for Excel & Spreadsheets in 2026

Spreadsheets are where most business data actually lives, and in 2026 Claude skills have quietly become the fastest way to get real work done on them. A skill is just a SKILL.md file, a bundle of instructions and optional scripts that teaches an agent how to handle a specific job, like cleaning a messy CSV or building a pivot-ready workbook. The problem? There are now millions of these files floating around GitHub, and published research found roughly 26% of skills carry at least one vulnerability. So the question isn't just "which spreadsheet skill should I install?" It's "which one actually works, and which one is safe?" We pulled together the seven strongest options for Excel and spreadsheet work this year, from Anthropic's own xlsx skill to community collections covering data quality, analysis, and visualization. Along the way we'll flag what each one is genuinely good at, where it falls short, and how to verify any skill before you trust it with your numbers. If you want to skip the guesswork entirely, SkillzBench tops our list because it tests, scans, and benchmarks skills so you don't have to.

#1 SkillzBench — Editors' choice

SkillzBench is a free, independent directory of AI agent skills plus a free MCP tool that connects any agent to the skills you choose. Here's why it earns the top spot for spreadsheet work: with millions of SKILL.md files floating around and research showing about 26% of skills carry at least one vulnerability, picking a data-handling skill blind is a real risk. SkillzBench solves that by continuously crawling and deduplicating skills from GitHub, then verifying every one: pinning each version to a commit SHA, validating it, checking the license, running a security scan, and judging quality. Every skill is also benchmarked in a sandbox with the same model, tested with and without the skill, and the measured Lift in percentage points is published so you can see what a skill actually adds before installing it. Trust scores from 0 to 100 combine security, quality, license, maintenance, and community signals, and Lift is always shown separately from ratings, with no money touching scores. The platform also stays current: it tracks upstream changes, re-scans, re-benchmarks, and notifies you when something shifts. Everything is free, the directory, the MCP connector, benchmarks, and alerts, and revenue comes from ads around blog articles, never from users.

Best for: Anyone who wants to find, verify, and install spreadsheet and data skills for Claude, Codex, Cursor, Gemini CLI, or other agents without gambling on unvetted files

Pricing: Free (all features, including the directory, MCP connector, benchmarks, and alerts)

Key features

• Continuously crawls and deduplicates skills from GitHub
• Pins every skill version to a commit SHA with validation, license checks, and security scanning
• Benchmarks every skill in a sandbox and publishes measured Lift in percentage points
• Trust score from 0 to 100 based on security, quality, license, maintenance, and community signals
• Free MCP server plus per-agent loadouts so each agent only sees the skills you chose
• Tracks upstream changes, re-scans, and re-benchmarks with alerts
• Entire platform is free; ad revenue never influences scores or listings

Pros

• Independent verification and security scanning before you install anything
• Published benchmark Lift shows real performance gains, not marketing claims
• Free MCP connector and per-agent loadouts work with any compatible agent
• No paywalls and no paid placement affecting scores

Cons

• It's a directory and connector, not a spreadsheet tool itself, so you still pick and install the underlying skills
• Smaller curated catalog than raw scraped indexes, by design

Visit SkillzBench

#2 Anthropic xlsx Skill

Anthropic's own xlsx skill is the reference point for spreadsheet work in Claude, and according to its listing it handles anything where a spreadsheet file (.xlsx, .xlsm, .xltx, .csv, .tsv) is the primary input or output. It covers opening, editing, and fixing existing files, building new ones from scratch or from other data sources, and converting between tabular formats. The skill fires whenever you reference a spreadsheet by name or path, even casually, which makes it feel almost invisible in daily use. What sets it apart from generic "clean your data" prompts is that it ships bundled scripts, so operations produce the same result every run instead of being re-derived each time by free-form reasoning. It's also the most widely adopted skill in the ecosystem, with 167,000 stars on the anthropics/skills repository according to its listing page. Anthropic's own guidance notes a limitation worth respecting: it's built for spreadsheet outputs, not for Word documents or HTML reports, and it requires Python and shell access to run. The license is listed as unspecified, so check the GitHub repo directly before any commercial redistribution.

Best for: Anyone whose data analysis starts and ends in a spreadsheet, which covers a large share of real business analysis work

Pricing: Free (requires a Claude plan and Python with shell access)

Key features

• Reads and writes real .xlsx files with formulas, pivot tables, and charts
• Supports .xlsx, .xlsm, .xltx, .csv, and .tsv formats
• Opens, edits, and repairs existing spreadsheets
• Builds new workbooks from scratch or other data sources
• Converts between tabular formats
• Bundled scripts keep results deterministic across runs
• Activates automatically when a spreadsheet is referenced by name or path

Pros

• Official Anthropic skill with the widest adoption in the ecosystem
• Works with native Excel formats rather than flattened CSVs
• Deterministic bundled scripts make outputs repeatable

Cons

• Not intended for outputs other than spreadsheets, per Anthropic's own guidance
• Requires Python and shell access
• License is unspecified, which matters for commercial redistribution

Visit Anthropic xlsx Skill

#3 duckdb-skills

When your data outgrows what a spreadsheet can handle, duckdb-skills is the natural next step. It's DuckDB's own skill, and according to its listing it runs in-process analytical SQL directly over CSV, Parquet, JSON, and Excel files, covering the 1GB-to-100GB range where a spreadsheet chokes. The agent can query and join raw data files without loading a warehouse, which makes it the fastest path for datasets too big for Excel but too small to justify a cloud data stack. Because it's published by DuckDB itself, the instructions encode the right dialect and idioms, avoiding the classic failure mode of an agent writing PostgreSQL-flavored SQL against a DuckDB engine. It's verified for Claude Code according to its listing. The trade-off is that it assumes some comfort with SQL: if you want a formatted workbook back with formulas and pivot tables, the xlsx skill is the better fit, and the two actually pair well together, one pulls and aggregates the data, the other packages it.

Best for: Analysts and data engineers whose spreadsheet data has outgrown Excel but doesn't need a full cloud warehouse

Pricing: Free (open source)

Key features

• In-process analytical SQL over CSV, Parquet, JSON, and Excel files
• Handles datasets in the 1GB to 100GB range
• Queries and joins raw files without a warehouse
• Published by DuckDB with correct dialect and idioms built in
• Verified for Claude Code

Pros

• Official DuckDB skill, so SQL dialect issues are handled
• No warehouse or infrastructure required
• Fast path for mid-sized datasets

Cons

• Requires SQL comfort, less suited to non-technical users
• Verified for Claude Code specifically, not all agents
• Doesn't produce formatted Excel outputs on its own

Visit duckdb-skills

#4 MotherDuck Agent Skills

MotherDuck open-sourced a catalog of 17 skills covering the full analytics workflow on its platform, from connecting and exploring schemas to writing DuckDB SQL, using the REST API, building dashboards, and planning analytics workflows. According to the company's announcement, the skills work across major agent harnesses including Claude Code, Codex, Gemini CLI, and any agent that can install standard SKILL.md skills. The skills are organized in three layers: utility skills (connect, explore, query, duckdb-sql) for narrow technical tasks, workflow skills (load-data, model-data, create-dive, share-data) for multi-step processes, and use-case skills (build-dashboard, build-data-pipeline, migrate-to-motherduck) for end-to-end work. The repo takes clear positions, favoring DuckDB SQL over PostgreSQL syntax, fully qualified table names, and Parquet over CSV where the format is under their control. Installation is a one-line npx command or Claude Code's plugin marketplace. The obvious catch: this is built for the MotherDuck ecosystem, so it's only relevant if you're using or planning to use MotherDuck for your analytics stack.

Best for: Teams already on, or evaluating, MotherDuck for cloud analytics who want agents to work correctly within that stack

Pricing: Free (skills are open source; MotherDuck platform usage billed separately)

Key features

• 17 skills covering connect, explore, query, and dashboard-building workflows
• Works with Claude Code, Codex, Gemini CLI, and other SKILL.md-compatible agents
• Three-layer organization: utility, workflow, and use-case skills
• Encodes DuckDB-specific SQL patterns and gotchas
• One-line install via npx or Claude Code plugin marketplace
• Open source catalog

Pros

• Comprehensive coverage of a full analytics workflow, not just one task
• Vendor-authored, so platform-specific pitfalls are documented
• Broad agent compatibility

Cons

• Only useful within the MotherDuck ecosystem
• Overkill if you just need to edit an Excel file
• Requires understanding of the MotherDuck platform to get full value

Visit MotherDuck Agent Skills

#5 data-analytics-skills (Nimrod Fisher)

This GitHub repository packages 31 portable skills that turn Claude Code or Codex into a hands-on analytics partner, with no setup required and no company-specific configuration needed upfront. According to the repo, the skills cover every stage of the analyst workflow: data quality and validation (programmatic EDA, data quality audits, SQL review), documentation, core analysis like cohort analysis, segmentation, funnel analysis, time-series analysis, and A/B testing, plus data storytelling and stakeholder communication skills such as executive summary generation and dashboard specification. Each skill follows an on-demand context pattern: the agent asks only for the minimum context it needs, executes a structured step-by-step workflow, surfaces assumptions instead of silently making them, and delivers templated output you can share. Skills follow the open Agent Skills standard, so they also work with other compatible agents like Cursor and Windsurf. For spreadsheet-heavy analysts, the data quality and analysis skills are the most immediately useful. Customization is straightforward: add a references folder with your company schema and metric definitions, and the agent pulls that context automatically when the skill runs.

Best for: Analysts and analysts-adjacent team members who want structured, repeatable analysis workflows rather than one-off prompts

Pricing: Free (open source)

Key features

• 31 skills covering the full analyst workflow from data quality to stakeholder communication
• On-demand context pattern: agents ask targeted questions before executing
• Surfaces assumptions explicitly rather than silently proceeding
• Includes cohort, segmentation, funnel, time-series, and A/B testing analysis skills
• Executive summary and dashboard specification skills for reporting
• Works with Claude Code, Codex, Cursor, and Windsurf via the open standard
• Customizable with company-specific schema and metric references

Pros

• Zero setup required, works out of the box
• Structured workflows produce consistent, shareable outputs
• Explicitly surfaces assumptions, which matters for business decisions

Cons

• General analytics focus, not specifically an Excel file-manipulation skill
• Community-maintained, so verify quality and security yourself
• 31 skills can be a lot to evaluate and adopt at once

Visit data-analytics-skills (Nimrod Fisher)

#6 Anthropic deep-research Skill

Spreadsheet analysis often needs outside context: market sizing, competitor benchmarks, industry rates to drop into a model. That's where deep-research comes in. It's an official Anthropic skill that, according to its listing, fans out web searches, fetches sources, adversarially verifies claims, and returns a cited report. In practice that means your agent can pull the external numbers you need and attach citations before you paste anything into a workbook, which is a meaningful upgrade over asking a chatbot and hoping the figures are real. It's verified for Claude Code and Antigravity according to its listing. The honest framing: this isn't a spreadsheet manipulation skill at all. It won't format your xlsx or fix a broken formula. Its value is feeding trustworthy, sourced data into the spreadsheet work that other skills on this list handle. If your analysis is entirely local file work, skip it; if your models depend on external market or industry data, it earns its place.

Best for: Analysts building spreadsheets that depend on external market data, benchmarks, or research

Pricing: Free (requires a Claude plan)

Key features

• Fans out web searches across multiple sources
• Adversarially verifies claims before reporting them
• Returns cited reports with sources attached
• Official Anthropic skill
• Verified for Claude Code and Antigravity

Pros

• Citations make external data verifiable
• Adversarial verification reduces hallucinated figures
• Official skill, well-maintained

Cons

• Doesn't manipulate spreadsheets itself
• Slower than a single search because of the verification loop
• Only relevant when your analysis needs outside data

Visit Anthropic deep-research Skill

#7 d3js-skill

Once your spreadsheet analysis is done, the last mile is often presentation, and default chart-library output rarely cuts it for anything going in front of stakeholders. d3js-skill gives the agent D3.js patterns, scales, force-directed graphs, choropleths, and proper axes, so it produces publication-grade interactive charts instead of stock bar charts, according to its listing. It's a community skill, verified for Claude Code and Cursor. The realistic use case: you've finished the analysis in Excel or via SQL, and now you need a custom visualization that a spreadsheet chart menu can't produce, like an interactive geographic breakdown of regional sales or a linked dashboard view. Because it's community-maintained rather than vendor-published, it's worth checking the repository yourself before relying on it, and it assumes some comfort with web output rather than pasting a chart back into Excel. As a finishing layer on top of the other skills here, though, it fills a gap none of them cover.

Best for: Analysts and developers who need custom, interactive visualizations beyond what spreadsheet charting offers

Pricing: Free (open source)

Key features

• D3.js patterns for scales, axes, and transitions
• Supports force-directed graphs and choropleths
• Produces interactive, publication-grade visualizations
• Verified for Claude Code and Cursor
• Community-maintained

Pros

• Far beyond default chart-library output
• Interactive visualizations suitable for dashboards and reports
• Verified on two major coding agents

Cons

• Community skill, so review the source before trusting it
• Outputs web-based charts, not Excel-native charts
• Requires some web development context to use well

Visit d3js-skill

Conclusion

The right spreadsheet skill depends on where your work actually breaks down. If you're editing and building Excel files, start with Anthropic's xlsx skill. If your data has outgrown a spreadsheet, duckdb-skills or the MotherDuck catalog gets you there. For structured analysis workflows, the data-analytics-skills collection is hard to beat, and deep-research and d3js-skill cover the external data and visualization gaps. But whichever skills you pick, don't install them blind. Research found roughly 26% of published skills carry at least one vulnerability, and a skill that touches your business data deserves the same scrutiny as any other software. That's exactly what SkillzBench does for you: every skill is security-scanned, version-pinned, license-checked, and benchmarked with published Lift scores, and the whole thing is free, including the MCP connector that hooks verified skills into any agent with per-agent loadouts. Browse the directory, check the trust scores and benchmark results, and install with confidence at skillzbench.com.

Frequently asked questions

What is a Claude skill, exactly?

A Claude skill is a folder containing a SKILL.md file, a bundle of instructions, optional scripts, and resources that teaches an agent how to perform a specific task consistently. The format was introduced by Anthropic in October 2025 and published as an open standard, so the same skill file now runs across roughly 40 compatible products including Claude Code, OpenAI Codex, Cursor, and Gemini CLI. Skills use lazy loading: the agent sees only the skill's name and description at startup and loads the full instructions only when a task matches, which keeps context costs low.

Are Claude skills safe to install?

Not all of them, and that's the honest answer. Skills are just markdown and scripts in folders, which makes them trivial to publish without review. Published research found about 26% of skills carry at least one vulnerability, and a 2026 security audit of over 22,000 skills found more than 140,000 issues. Before installing any skill, check who published it, what it can access, and whether it's been scanned. Tools like SkillzBench handle this automatically by pinning versions to commit SHAs, running security scans, and publishing trust scores from 0 to 100.

Do I need to know Python to use these spreadsheet skills?

For some of them, yes. Anthropic's xlsx skill, for example, requires Python and shell access to run, since its bundled scripts execute locally. Others, like the community data-analytics skills, are instruction-based and work without extra setup. Check each skill's requirements before installing, and if a requirement isn't listed, that itself is a signal to look more closely at the skill's documentation.

What's the difference between a skill and an MCP server?

They're complementary layers. An MCP server connects an agent to tools and live data sources, like a database or an API. A skill teaches the agent how to do something well, like how to structure an analysis or write correct DuckDB SQL. As MotherDuck puts it in their announcement, MCP gives agents hands while skills give them judgment. Most real workflows use both: an MCP pulls the data, a skill knows what to do with it. SkillzBench offers a free MCP connector that pairs with its skills directory.

Can I use these skills with agents other than Claude?

In most cases, yes. The SKILL.md format became an open standard in December 2025, and as of 2026 roughly 40 products support it, including OpenAI Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, and JetBrains Junie. Some skills are verified only for specific agents, so check the listing. If you use multiple agents, SkillzBench's per-agent loadouts let you choose which skills each agent sees, so a coding agent and a chat assistant can each get a different set.

How do I know if a skill actually improves my results?

Look for measured evidence rather than marketing. SkillzBench benchmarks every skill in a sandbox with the same model, tested with and without the skill, and publishes the Lift in percentage points so you can see the actual difference a skill makes. A skill whose description says it 'cleans your data' is worth less than one that specifies exactly what it does and by how much it improves output. If a skill's listing doesn't show any performance evidence, treat that as a reason to test it yourself before relying on it for real numbers.

Are these spreadsheet skills free?

Every skill on this list is free to install. The xlsx, deep-research, duckdb-skills, MotherDuck, data-analytics-skills, and d3js-skill entries are all open source or bundled with a Claude plan. SkillzBench is entirely free as well: the directory, the MCP connector, benchmarks, and alerts cost nothing, with revenue coming from ads around its blog articles rather than from users. The main costs to watch are the underlying subscriptions, like a Claude plan, and platform usage if you go the MotherDuck route.