benbrastmckie/nvim/skill-docx-edit

Claude Code and agent skill by benbrastmckie/nvim

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

In-place DOCX editing routing to docx-edit-agent

Install

npx vskill@latest install benbrastmckie/nvim/skill-docx-edit

SKILL.md

DOCX Edit Skill

Thin wrapper that routes in-place DOCX editing requests to the docx-edit-agent, which edits documents using SuperDoc MCP with optional Word AppleScript integration for zero-friction workflow.

Context Pointers

Reference (do not load eagerly):

  • Path: .claude/context/formats/subagent-return.md
  • Purpose: Return validation
  • Load at: Subagent execution only

Note: This skill is a thin wrapper. Context is loaded by the delegated agent, not this skill.

Trigger Conditions

This skill activates when:

Direct Invocation
  • User explicitly runs /edit command with a .docx file
  • User requests document editing in conversation
Implicit Invocation (during task implementation)

When an implementing agent encounters any of these patterns:

Plan step language patterns:

  • "Edit [file].docx to [change]"
  • "Replace [text] with [text] in [file].docx"
  • "Update the Word document [file]"
  • "Add tracked changes to [file].docx"
  • "Modify [file].docx with [instruction]"
  • "Create a new document [file].docx"

File extension detection:

  • Target file has extension: .docx
  • Operation type: edit, replace, update, modify, create

Task description keywords:

  • "DOCX editing"
  • "Word document editing"
  • "tracked changes"
  • "in-place document edit"
When NOT to Trigger

Do not invoke for:

  • Non-DOCX files (.xlsx, .pptx, .pdf, etc.)
  • Document conversion (use skill-filetypes for /convert)
  • PDF annotation extraction (use skill-scrape for /scrape)
  • Spreadsheet table extraction (use skill-spreadsheet for /table)
  • Reading document content without editing (use markitdown or /convert)

Execution

1. Input Validation

Validate required inputs:

  • file_path - Must be provided; for edit/batch must exist; for create mode may not exist yet
  • instruction - Natural language editing instruction (required)
  • mode - One of: "edit", "batch", "create" (inferred from file_path and flags)
bash
# For edit mode: validate source exists and is .docx
if [ "$mode" = "edit" ]; then
if [ ! -f "$file_path" ]; then
return error "DOCX file not found: $file_path"
fi
if [[ "${file_path##*.}" != "docx" ]]; then
return error "Source must be a .docx file: $file_path"
fi
fi
# For batch mode: validate directory exists
if [ "$mode" = "batch" ]; then
if [ ! -d "$file_path" ]; then
return error "Directory not found: $file_path"
fi
fi
# For create mode: validate .docx extension
if [ "$mode" = "create" ]; then
if [[ "${file_path##*.}" != "docx" ]]; then
return error "New file must have .docx extension: $file_path"
fi
fi
2. Context Preparation

Prepare delegation context:

json
{
"file_path": "/absolute/path/to/document.docx",
"instruction": "Replace ACME Corp with NewCo Inc. using tracked changes",
"mode": "edit",
"metadata": {
"session_id": "sess_{timestamp}_{random}",
"delegation_depth": 1,
"delegation_path": ["orchestrator", "edit", "skill-docx-edit"]
}
}
3. Invoke Agent

CRITICAL: You MUST use the Agent tool to spawn the docx-edit agent.

Required Tool Invocation:

Tool: Agent (NOT Skill, NOT Plan)
Parameters:
- subagent_type: "docx-edit-agent"
- prompt: [Include file_path, instruction, mode, metadata]
- description: "Edit {file_path}: {instruction}"

DO NOT use Skill(docx-edit-agent) - this will FAIL. Agents live in .claude/agents/ or extension agent directories, not .claude/skills/. The Skill tool can only invoke skills from .claude/skills/.

The agent will:

  • Check if Word has the file open (macOS only)
  • Save partner's unsaved work if needed (macOS only)
  • Perform SuperDoc editing operations
  • Reload the document in Word if it was open (macOS only)
  • Confirm edit results
  • Return standardized JSON result
4. Return Validation

Validate return matches subagent-return.md schema:

  • Status is one of: edited, created, partial, failed
  • Summary is non-empty and <100 tokens
  • Artifacts array present with output file path
  • Metadata contains session_id, agent_type, delegation info

Status value rules:

  • edited - File successfully modified in place
  • created - New file successfully created
  • partial - Some edits applied, others failed (batch mode)
  • failed - No edits applied, error occurred
  • NEVER use "completed" (triggers Claude stop behavior)
5. Return Propagation

Return validated result to caller without modification.


Return Format

See .claude/context/formats/subagent-return.md for full specification.

Expected successful return:

json
{
"status": "edited",
"summary": "Replaced 3 instances of 'ACME Corp' with 'NewCo Inc.' in contract.docx with tracked changes (author: Claude)",
"artifacts": [
{
"type": "implementation",
"path": "/absolute/path/to/contract.docx",
"summary": "Edited DOCX with tracked changes"
}
],
"metadata": {
"session_id": "sess_...",
"agent_type": "docx-edit-agent",
"delegation_depth": 2,
"delegation_path": ["orchestrator", "edit", "skill-docx-edit", "docx-edit-agent"],
"tool_used": "superdoc",
"changes_count": 3,
"tracked_changes": true,
"word_was_open": true,
"word_reloaded": true
},
"next_steps": "Review tracked changes in Word"
}

Error Handling

Input Validation Errors

Return immediately with failed status if file not found, not a .docx, or instruction is empty.

Agent Errors

Pass through the agent's error return verbatim.

Tool Not Available

Return failed status with installation instructions for SuperDoc MCP or python-docx.


MUST NOT

  • Run the postflight step (git commit, status update) -- that is the command's responsibility
  • Modify the return from docx-edit-agent before propagating
  • Load context files eagerly -- only reference them when needed
  • Use the Skill tool to invoke docx-edit-agent (use Agent tool instead)

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
444
Repository forks
462
7-day trend score
30

Stars and forks are for the whole GitHub repository, not this skill alone. Refreshed 22 Aug 2026, 2 months ago.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/benbrastmckie/nvim/skill-docx-edit/badge)](https://verified-skill.com/skills/benbrastmckie/nvim/skill-docx-edit)

Something wrong with this skill or its page? Tell us.