Skip to main content
Agent skillFirst-run guide

Docx

Create and edit Microsoft Word (.docx) files with headings, tables, and tracked-change-friendly structure from an agent workflow.

Get started

Get the complete skill folder, including its scripts and reference files.

·

Ready-made practice workflow

Turn a sales spreadsheet into a monthly report

Check two months of figures, build a formula-backed summary, and write an editable one-page review. (3 sample files)

Try guided workflow

When to use this skill

Use the Word skill when someone has to open the file in Microsoft Word or Google Docs conversion — proposals, letters, policies, and anything that still lives as .docx in a company workflow.

When not to use it

Use PDF when the artifact should not be casually edited. Use markdown in-repo for engineering docs that belong in git.

How to install

Where will you use it?

Install the publisher’s plugin

This installs a bundle that can include other skills. Choose this method or the manual alternative below; you only need one installation.

  1. Run these commands in Claude Code:

    Claude Code CLI
    /plugin marketplace add anthropics/skills
    /plugin install document-skills@anthropic-agent-skills
  2. Ask Claude Code to use Docx on a small task and confirm that it loads. Check the setup requirements in the publisher’s guide if it cannot run.
Alternative: install the folder manually

Use these steps if you choose the manual folder method. If you installed the plugin above, skip this alternative.

  1. Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/docx/SKILL.md. For a shared project, commit the folder at .claude/skills/docx/ instead.
  2. Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
  3. For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.

Full install guide for Claude, Cursor, and Codex →

What supports this recommendation?

75/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.

Documentation
77/100
Practical detail
68/100
Source evidence
70/100
Source signals
94/100
Model compatibility evidence
  • claudeuntested

    No model-specific signal or recorded compatibility test was found.

  • chatgptuntested

    No model-specific signal or recorded compatibility test was found.

  • geminiuntested

    No model-specific signal or recorded compatibility test was found.

  • copilotuntested

    No model-specific signal or recorded compatibility test was found.

  • llamauntested

    No model-specific signal or recorded compatibility test was found.

  • perplexityuntested

    No model-specific signal or recorded compatibility test was found.

  • mistraluntested

    No model-specific signal or recorded compatibility test was found.

  • grokuntested

    No model-specific signal or recorded compatibility test was found.

Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.

Source reviewedExample execution recorded

Try one small task

Best first exercise: make an editable one-page report from supplied facts.

See actual outputs and execution limits →

What you need

  • An agent that can write files and run Node.js; upstream uses the docx npm package for new documents.
  • A Word-compatible reader to inspect the resulting file; review page layout after conversion or export.
  • Use the figures and brief in the monthly report workflow below.

Get the sample input and full checklist

Turn verified monthly totals into an editable management brief.

First task prompt
Use the DOCX skill to create monthly-report.docx. Title: February 2026 sales review. Use only these supplied figures: January revenue USD 12,000, cost USD 6,000; February revenue USD 15,000, cost USD 7,800. Include a heading, a two-month comparison table, and three short observations. Calculate gross profit and gross margin. Label suggested actions as proposals and do not invent causes for the changes. Keep the body editable and check the rendered layout.

Expected result

A specification to check against, not output from an AI run.

An editable DOCX containing the four supplied figures, calculated gross profit and margin, and a clearly marked proposed next step.

  • Word or another DOCX reader opens the file without repair warnings.
  • The table shows revenue growth of 25%, gross profit growth of 20%, and a gross-margin decline of 2 percentage points.
  • Headings and the table remain editable, with no clipped text when rendered.

Where this exercise stops

  • Creating a new document and editing an existing DOCX use different upstream approaches.
  • Rendering varies across document readers. File creation alone does not verify pagination.

Recorded example run

These files came from actual execution in a Codex session. They were produced and checked within the same session; this is scoped evidence, not independent review or a measured success rate.

Sales workbook summarized in an editable monthly report

Executed · Automated Codex exercise

Codex agent in this session; Python 3.12.3, openpyxl 3.1.5, LibreOffice 24.2.7.2 420(Build:2); aarch64. Report generation: docx 9.8.1, Node v24.13.0.

The agent read the pinned upstream instructions and applicable helpers, then authored and executed the committed scripts locally. This is not a native Claude skill-loader test or an independent benchmark.

Download the actual outputs

Recorded checks and observations

  • Passed: All six input rows are preserved and the summary retains formulas.
  • Passed: Recalculated January/February revenue is USD 12,000 / 15,000; gross margin is 50% / 48%.
  • Passed: The DOCX contains an editable five-row table and the expected figures; all XML parts parse.
  • Passed: The report renders to one page with expected text. The creating agent inspected the image and observed no clipping or overlap.
Scope, limitations and pinned sources
  • The agent read the pinned upstream instructions and used their prescribed tools. Native Claude skill loading and general cross-agent compatibility were not tested.
  • One small synthetic input per workflow; results are not a success rate or a broad model-compatibility claim.
  • The Codex agent that produced the outputs also ran and interpreted the checks; no independent human review occurred.
  • Spreadsheet recalculation used LibreOffice 24.2.7.2; Microsoft Excel was not run.
  • DOCX structure and one-page LibreOffice rendering were checked; Microsoft Word and Google Docs rendering were not tested.
  • The temporary LibreOffice render emitted Java-launch and liblangtag registry warnings. Conversion returned 0 and the produced PDF was inspected; unsupported language/layout cases remain untested.
  • This session already knew the published acceptance criteria. The CSV answer key was read by verification after PDF extraction, but this is not a blind or held-out evaluation.
Read the execution record (JSON) →
Source and evidence

The upstream instructions informed this guide. The exercise and acceptance checks are an OpxSkills test recipe. Where a run is recorded, its actual outputs, environment and limits appear separately above. A recorded example does not establish general model compatibility or a success rate.

Source reviewed · Upstream revision 8a1541c4a3ff

What this skill does

The DOCX skill produces real Office Open XML, not a paste of text. That matters for headings, lists, and tables that survive a round-trip through Word.

Intended uses

  • Draft a client proposal as .docx
  • Rewrite an internal policy without losing heading structure
  • Turn meeting notes into a letter template

Implementation notes

  • Specify heading levels if the org uses a template.
  • Ask for comments vs body text explicitly.

Try it, then tell us how it went

A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.