How to use the LibreOffice Document CLI agent skill

What the skill lets the agent do

The skill wraps a stateful command-line tool for document editing. It produces real ODF files, which are ZIP archives containing XML, and it is designed for AI agents and power users who need to work without a GUI or a LibreOffice window.

The agent builds a project file and then runs commands to add paragraphs, headings, lists and tables to a Writer document, add sheets and set cell values in Calc, or add slides in Impress. When the work is finished, it exports the document to a file such as a PDF. Existing DOCX, XLSX, PPTX or ODF files can be imported into an editable project.

• Writer: add paragraphs, headings, lists, tables and page breaks, and list or edit content items
• Calc: add, rename and remove sheets, and set or read cell values
• Impress: add, remove and list slides, update slide titles and content, and add elements
• Styles: create, modify, list, apply (Writer only) and remove styles
• Export: list presets, inspect a preset and render the document to a file
• Session: check status, undo, redo and view history (up to 50 levels)

Install the skill from the ZIP or the Claude apps

You can add the skill to Claude Code by downloading the ZIP from this page and unzipping it into your user skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder. In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

The skill package includes a SKILL.md file. The underlying CLI is installed separately as a Python package, as shown in the code block, and LibreOffice must already be on your system.

• Requirements: Python 3.10 or newer
• Requirements: LibreOffice installed on the system
• Requirements: the pip package cli-anything-libreoffice

Check that the command-line tool works

Before asking an agent to use the skill, confirm the command runs in your terminal. The source lists a help command and a project-creation command you can use for this check.

Ask the agent for document work in plain language

Once the skill is installed, you do not need to type the commands yourself. Describe the result you want and name the file type or output. The agent is expected to pick the matching commands, and it works best when you give it absolute paths for inputs and outputs, because the source asks agents to use absolute paths for all file operations.

Examples of requests you might make:

• Create a Writer document with a title, three headings and a table of monthly figures, then export it as a PDF to /home/me/reports/summary.pdf.
• Open the existing file /home/me/docs/contract.docx, import it into an editable project, and add a paragraph at the end.
• Make a Calc workbook with a sheet called Sales and fill in the cell values from this list.
• Build a five-slide Impress presentation about our Q3 results and save the slides as a PDF.

How agents should use the commands

The source gives agents a short set of rules. Using them yourself is a good way to understand what the agent is doing and to read its output.

• Always use the --json flag for parseable output
• Check return codes: 0 means success, non-zero means an error
• Parse stderr for error messages on failure
• Use absolute paths for all file operations
• Import existing Office files to JSON first, for example with document import input.docx -o work.json
• Verify that outputs exist after export operations

Safety notes before you run it

The skill runs scripts and reads and writes local files. The package is installed from Python, and the code was not reviewed as part of this assessment, so check what you are installing. The tool does not use the network according to this listing.

Be careful with sensitive documents on a shared machine, since the tool works directly on files on that machine. Keep copies of originals, and prefer a test folder while you are learning. The export command has an --overwrite option, so know which output path you are targeting.

• The skill runs scripts (yes) and uses the network (no), per this listing
• Work on copies of important documents
• Review the package code yourself if the documents are confidential

Tips for better results

Keep one project file per document so the agent's changes are easy to trace. Use the session history when you want to back out a change, since undo and redo cover up to 50 levels. The skill supports the import of DOCX, XLSX, PPTX and ODF files, so start from an existing file when you need one.

Be realistic about layout. The skill covers a subset of commands and, per this listing, offers no direct formatting options beyond styles. If you need exact visual design, use a desktop office program.

• Use one project JSON file per document
• Use undo and redo to correct mistakes
• Start from an existing file with import when you can
• Use styles for formatting rather than expecting fine-grained controls