XLSX (Anthropic, official)

tl;dr: one of the four document skills Anthropic ships under anthropics/skills (with docx, pdf, pptx) — SKILL.md is the routing layer, and the heavy lifting is openpyxl for formula-and-format work, pandas for bulk in/out, and markitdown for a quick text preview. Every output with formulas must end with python scripts/recalc.py — openpyxl writes formulas as strings with no cached values, so pandas and data_only=True reads return None until LibreOffice rewrites the file.

What it actually does

Per the frontmatter, the skill activates for any spreadsheet mention by path or even casually — “the xlsx in my downloads.” It covers create, read, edit, fix, format, chart, find-and-replace, clean messy tabular data, and convert between .xlsx / .xlsm / .xltx / .csv / .tsv. It explicitly does not trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration.

Task routing (from SKILL.md)

openpyxl, pandas, and markitdown are preinstalled — write the script and import directly. Only pip install if an import actually fails or the markitdown command is missing.

Recalculate (mandatory when the file contains formulas)

openpyxl writes formulas as strings with no cached values, so until you recalc, every formula cell reads back as None to pandas, to load_workbook(data_only=True), and to most previewers.

python scripts/recalc.py output.xlsx [timeout_seconds]   # default 30

LibreOffice computes every formula, the file is rewritten in place, and you get JSON: status (success | errors_found), total_formulas, total_errors, and an error_summary naming up to 100 cells per error type (locations_truncated says how many it withheld — trust total_errors, not the length of the list). Fix what it names and run it again.

Two gotchas the skill is explicit about:

1. JSON with an error key instead of status means nothing was recalculated — and only that case exits non-zero. errors_found exits 0, so never treat a clean exit as a clean workbook.
2. A green recalc proves your formulas evaluate, not that they are right. An off-by-one range or a reference to the wrong row yields a clean, error-free file with wrong numbers. Write 2–3 formulas first and check they pull the values you expect, before building out a grid.

Formulas that survive verification

LibreOffice implements fewer functions than Excel, and one it cannot evaluate becomes a literal #NAME? baked into the file you deliver.

• Prefer Excel-2007-era functions — SUMIFS, INDEX, MATCH, IFERROR, SUMPRODUCT — which need no prefix.
• Six post-2007 functions work, but only with an _xlfn. prefix (openpyxl writes your formula into the XML verbatim and Excel stores post-2007 names prefixed; its UI hides the prefix): _xlfn.TEXTJOIN, _xlfn.CONCAT, _xlfn.IFS, _xlfn.SWITCH, _xlfn.MAXIFS, _xlfn.MINIFS. Written bare, each yields #NAME?.
• Never use XLOOKUP, XMATCH, SORT, FILTER, UNIQUE, or SEQUENCE. The runtime’s LibreOffice cannot evaluate them under any prefix. Newer builds do evaluate them, but they are spilling array functions and an openpyxl-written file has no spill metadata, so only the top-left cell of the range gets a value — and recalc.py reports total_errors: 0 on the truncated result. Use INDEX/MATCH for lookups; sort, filter, and dedupe in Python before writing the cells.
• A formula LibreOffice could not parse is written back lowercased — a quick tell beside a #NAME?.

openpyxl gotchas (from SKILL.md)

• Reading a model takes two loads. data_only=True yields cached values with the formulas gone; the default yields formula strings with no values. One pass cannot give you both.
• data_only=True is destructive if you save. That workbook has no formulas left, so saving replaces every one with a literal — permanently.
• data_only=True on a file openpyxl just wrote returns None everywhere — run recalc.py first. (A formula whose result is "" also reads back as None.)
• Merged cells: write the top-left anchor only. Every other cell in the range is a MergedCell whose .value is read-only.
• .xlsm loses its macros unless you pass keep_vba=True to load_workbook.
• A sheet name containing a space must be quoted in a cross-sheet reference: ='Assumptions Inputs'!$B$5. Unquoted, it evaluates to #VALUE!.
• A workbook that links to another file loses those links if you re-save it with openpyxl and then recalculate. The cell’s cached value is the only thing holding its data, openpyxl strips it on save, and LibreOffice then writes #NAME? for the missing external. recalc.py refuses in that state — copy those cells’ values out of the original before saving over them (--force overrides, and accepts the loss).

Financial-model conventions (from SKILL.md)

Unless the user says otherwise, or the existing file already does something else:

Color: blue text (0,0,255) for hardcoded inputs and scenario levers · black for formulas · green (0,128,0) for links to another sheet · red (255,0,0) for links to another file · yellow fill (255,255,0) for key assumptions and cells the user should fill in.

Numbers: currency $#,##0, with the unit named in the header (Revenue ($mm)) · zeros render as -, including in percentages ($#,##0;($#,##0);-) · negatives in parentheses · percentages 0.0%, stored as fractions (0.15 renders 15.0%; storing 15 renders 1500.0%) · valuation multiples 0.0x · years as text ("2024", never 2,024).

Structure: every assumption in its own labeled cell, referenced by the formulas that use it (=B5*(1+$B$6), never =B5*1.05) · formulas consistent across every projection period, since a lone edited cell mid-row is the commonest silent error · guard denominators that can be zero.

Output requirements (from SKILL.md)

• Professional font (Arial, Times New Roman) throughout, unless the user says otherwise.
• Zero formula errors. Never ship while recalc.py reports errors_found. An error you introduced looks exactly like one you inherited — prove it: load the original with data_only=True and look at that cell.
• Use formulas, never hardcoded results. Write sheet['B10'] = '=SUM(B2:B9)', not the Python-computed total. The sheet must recalculate when its inputs change.
• Follow the user’s spec literally. Exact tab names, exact column headers, and the formula they spelled out. A redesign that computes something else fails, however elegant.
• Document every assumption and hardcoded number where the reader will see it — a cell comment, or an adjacent cell at a table’s end. Cite a real source for real numbers; when the number came from the user, say so plainly.
• A workbook you create for someone to fill in needs a short legend naming which cells to edit, and one example row of realistic values showing the expected format. Never add such a row to a file you were asked to edit.
• Editing an existing file: match its conventions exactly. They override every guideline here. Find its designated input cells first — a distinct font color, fill, or shading marks them — write only there, and leave every existing formula untouched.

Install

The skill is one entry in the document-skills plugin of anthropic-agent-skills (per .claude-plugin/marketplace.json on 2026-09-10 — a 1.7 KB marketplace manifest listing five plugins: document-skills, example-skills, claude-api, academy-guide, discernment-nudge). Install requires the marketplace to be registered first.

# Claude Code — one-time marketplace registration
/plugin marketplace add anthropics/skills

# Install the document-skills plugin (includes docx, pdf, pptx, xlsx)
/plugin install document-skills@anthropic-agent-skills

The Anthropic README does not document an npx skills add workflow — that command comes from a third-party installer. Use /plugin install for the official path.

On Claude.ai, the document skills are pre-built for paid plans. On the API, attach via the container.skills parameter:

{
  "model": "claude-opus-5",
  "max_tokens": 4096,
  "container": {
    "skills": [
      { "type": "anthropic", "skill_id": "xlsx", "version": "latest" }
    ]
  },
  "messages": [
    { "role": "user", "content": "Build a Q3 revenue model in /mnt/q3.xlsx with three scenarios" }
  ],
  "tools": [
    { "type": "code_execution_20250825", "name": "code_execution" }
  ]
}

Skills require code_execution_20250825; pinned to the Claude API ID claude-opus-5 in the example. Use the Files API to pull the result back when Claude writes a new .xlsx.

Dependencies (per SKILL.md)

openpyxl, pandas, markitdown (pip, preinstalled — install only if an import fails or the command is missing) · LibreOffice (soffice, auto-configured for sandboxed environments via scripts/office/soffice.py).

Gotchas

• License is Proprietary. LICENSE.txt restricts redistribution and derivative works. You use the skill; you do not fork it.
• scripts/recalc.py is the gate, not a hint. Files without cached formula values are unreadable by every other tool downstream. Skipping it is the most common way to ship a “working” workbook that the user opens to #NAME? errors.
• XLOOKUP and friends silently truncate. They evaluate on newer LibreOffice but with no spill metadata, so the model looks correct to recalc.py and is wrong in Excel. The skill’s instruction is hard: use INDEX/MATCH.
• Two-pass read is mandatory for any model with both formulas and values. There is no single load_workbook that returns both, and saving after data_only=True strips formulas permanently.
• --force on recalc.py accepts external-link loss. External references ([1]Other File syntax) lose their cached values on openpyxl save. Without --force, recalc refuses; with it, you accept #NAME? for every cell that referenced an external file.
• Install command is /plugin install, not npx skills add. Anyone running npx skills add anthropics/skills is using a third-party installer, not Anthropic’s path.
• sandbox_test untested. We read SKILL.md and the marketplace manifest, not exercised the openpyxl pipeline in a sandbox.

Related

• Anthropic DOCX (official skill) — sibling in the document-skills plugin, same unzip → edit → rezip editing pattern
• Anthropic PPTX (official skill) — sibling in the document-skills plugin, same LibreOffice recalc dependency
• Anthropic PDF (official skill) — sibling in the document-skills plugin, same Python-tooling pattern
• Claude skills guide — what the cache.directory skills shelf actually is