Word documents

Use this Skill for .docx work. Preserve the user's content and formatting intent, and make the smallest change that satisfies the request.

Start here

Call skill_view for this Skill before using its resources. It returns baseDirectory; every bundled helper below lives under <baseDirectory>/scripts.

Default to a new output file. Replace the input only when the user explicitly asks for an in-place edit. Before changing a document, inspect it with docx_read.py and keep a recoverable original.

Core dependency:

python3 -m pip install python-docx lxml

Do not install dependencies silently. Check availability first and explain any missing dependency that blocks the task.

Helpers

• docx_read.py: extract paragraphs, headings, tables, links, comments, headers, footers, and document metadata.
• docx_create.py: create a document from structured JSON.
• docx_edit.py: perform targeted text, paragraph, table, image, header, and footer edits.
• docx_template.py: inspect placeholders and populate a template.
• docx_comments.py: add, list, or remove comments.
• docx_revisions.py: inspect, accept, or reject tracked revisions.
• docx_validate.py: validate the resulting package and report structural problems.

Run helpers with terminal_exec, passing python3 as the command and arguments as an array. Use --help when a helper's exact arguments are unclear.

python3 <baseDirectory>/scripts/docx_read.py input.docx --json
python3 <baseDirectory>/scripts/docx_create.py spec.json -o output.docx
python3 <baseDirectory>/scripts/docx_edit.py input.docx edits.json -o output.docx
python3 <baseDirectory>/scripts/docx_validate.py output.docx

For tracked changes and comments, read references/revisions-and-comments.md through skill_view before editing them.

Workflow