feat: Add docx skill for reading and editing Word documents

[stevenobiajulu]

What feature would you like to see?

A docx skill that enables Kimi to read and edit .docx files using Python stdlib only (zipfile + xml.etree.ElementTree).

The skill-creator guide already uses a docx skill as its canonical example (Pattern 3: Conditional details, and the description field example), but no such skill exists yet.

v1 scope:

• Read: extract text with paragraph indices
• Search: find text across paragraphs (with quote normalization)
• Replace: text replacement preserving run formatting (w:rPr)
• Save: repack to new .docx with namespace validation

Implementation notes:

• Single SKILL.md (~210 lines), inline-script pattern (like worktree-status)
• Zero dependencies — stdlib only
• .docx is in _NON_TEXT_SUFFIXES, so the skill provides its own reading via inline Python
• Deterministic formatting (no AI judgment on w:rPr)
• Saves to new file only, never overwrites

I have a working branch ready if this aligns with the roadmap.

Additional information

Happy to follow whichever file pattern you prefer — bare SKILL.md (matching existing skills) or the scripts/ subdirectory structure from the guide.