Word Template Generator Skill

⚠️ Execution Environment Disclaimer

This skill provides guidance and generates structured data for Word document processing.

Important Limitations:

• Claude is a text-based AI model and cannot directly execute Python code
• Claude cannot directly manipulate .docx binary files - these require specialized libraries
• All Python code examples in this skill are reference implementations for an external runtime
• Actual Word document processing requires:
  • A Python runtime environment
  • Dependencies:

    python-docx

    ,

    docxtpl

    ,

    openpyxl

    ,

    pandas

    ,

    pillow

What Claude DOES:

• Extract and structure data from user input/materials into validated JSON context
• Analyze template requirements and placeholder mapping
• Generate structured output that a runtime tool can consume

What an External Tool DOES:

• Execute Python code to read/write .docx files
• Use

  python-docx

  for document manipulation
• Use

  docxtpl

  for Jinja2 template rendering
• Handle binary file operations and image processing

Overview

This skill guides the processing of Word document templates through a clear separation of concerns:

• Claude's Role: Generate validated context JSON from user input and materials
• Tool's Role: Execute docx rendering using the provided context

The workflow covers:

• Analyzing Word template structure and placeholders
• Extracting and structuring data from materials
• Generating validated JSON context for template rendering
• Providing reference implementations for runtime execution

Trigger Conditions

Trigger this skill when:

• User provides a Word template file and asks to fill/generate a document
• User provides a "materials folder" and asks to generate documents based on a template
• User says "help me generate a Word document" and mentions a template
• User asks to "generate Word based on template" + mentions material location
• User mentions "Word template", "docx template", "fill template"
• User asks to batch generate Word documents
• User says "use these materials to generate document" and provides a template
• User asks to "process Word template", "parse docx template", or "extract placeholders from Word"

Architecture & Responsibilities

┌─────────────────┐      ┌──────────────────┐      ┌─────────────────┐
│   User Input    │ ──▶ │   Claude (LLM)   │ ──▶ │  Context JSON   │
│  + Template     │      │  - Extract data  │      │  - Structured   │
│  + Materials    │      │  - Validate      │      │  - Validated    │
└─────────────────┘      └──────────────────┘      └─────────────────┘
                                                              │
                                                              ▼
┌─────────────────┐      ┌──────────────────┐      ┌─────────────────┐
│ Generated .docx │ ◀── │   Runtime Tool   │ ◀── │  Template .docx │
│   - Rendered    │      │  - python-docx   │      │  - Jinja2       │
│   - Formatted   │      │  - docxtpl       │      │  - Placeholders │
└─────────────────┘      └──────────────────┘      └─────────────────┘

Claude's Responsibilities (This Skill)

1. Parse Template Requirements: Analyze placeholders and their types
2. Extract Data: Process user input and materials folders
3. Validate Structure: Ensure data matches JSON Schema requirements
4. Generate Context JSON: Produce structured, validated output

Runtime Tool's Responsibilities

1. Read .docx Files: Use

   python-docx

   to access template content
2. Render Templates: Use

   docxtpl

   with Jinja2 syntax
3. Process Images: Handle image embedding and formatting
4. Save Output: Generate the final .docx file

JSON Schema Reference

The context JSON passed to the runtime tool MUST conform to this schema:

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "title": "Word Template Context",
  "description": "Context data for Word template rendering",
  "type": "object",
  "properties": {
    "title": {
      "type": "string",
      "description": "Document title or heading"
    },
    "content": {
      "type": "string",
      "description": "Main content text"
    },
    "items": {
      "type": "array",
      "description": "List of items for table/loop rendering",
      "items": {
        "type": "object",
        "properties": {
          "name": {
            "type": "string",
            "description": "Item name or identifier"
          },
          "value": {
            "oneOf": [
              {"type": "string"},
              {"type": "number"},
              {"type": "boolean"}
            ],
            "description": "Item value"
          },
          "description": {
            "type": "string",
            "description": "Optional item description"
          },
          "price": {
            "type": "number",
            "description": "Price for numeric items"
          },
          "qty": {
            "type": "number",
            "description": "Quantity"
          }
        },
        "required": ["name"]
      }
    },
    "image": {
      "type": "object",
      "description": "Image placeholder data",
      "properties": {
        "path": {
          "type": "string",
          "description": "Absolute or relative path to image file"
        },
        "width": {
          "type": "number",
          "description": "Image width in inches (optional)"
        },
        "height": {
          "type": "number",
          "description": "Image height in inches (optional)"
        }
      },
      "required": ["path"]
    },
    "metadata": {
      "type": "object",
      "description": "Document metadata",
      "properties": {
        "date": {
          "type": "string",
          "format": "date",
          "description": "Document date (ISO 8601 format)"
        },
        "author": {
          "type": "string",
          "description": "Document author"
        },
        "version": {
          "type": "string",
          "description": "Document version"
        },
        "contract_no": {
          "type": "string",
          "description": "Contract or reference number"
        }
      }
    },
    "party_a": {
      "type": "string",
      "description": "First party name (for contracts)"
    },
    "party_b": {
      "type": "string",
      "description": "Second party name (for contracts)"
    },
    "is_vip": {
      "type": "boolean",
      "description": "VIP status flag for conditional rendering"
    },
    "total_amount": {
      "type": "number",
      "description": "Calculated total amount"
    }
  },
  "additionalProperties": true
}

Supported Data Types

string

"Hello World"

number

123.45

boolean

true

array

[{"name": "A"}, {"name": "B"}]

object

{"path": "/img/logo.png"}

string

"/path/to/image.png"

object

{"path": "/img.png", "width": 2.5}

Claude Output Format Examples

Example 1: Simple Contract Context

Claude generates this JSON:

{
  "contract_no": "HT-2026-001",
  "party_a": "Zhang San Company",
  "party_b": "Li Si Enterprise",
  "date": "2026-02-22",
  "title": "Service Agreement"
}

Example 2: Invoice with Table Data

Claude generates this JSON:

{
  "invoice_no": "INV-001",
  "items": [
    {"name": "Product A", "qty": 2, "price": 100, "total": 200},
    {"name": "Product B", "qty": 3, "price": 50, "total": 150}
  ],
  "total_amount": 350
}

Example 3: Report with Mixed Content

Claude generates this JSON:

{
  "title": "Quarterly Sales Report",
  "project_info": "Analysis of Q1 2026 sales performance across all regions...",
  "data_items": [
    {"name": "Region A", "value": 125000},
    {"name": "Region B", "value": 98000},
    {"name": "Region C", "value": 156000}
  ],
  "chart_image": {
    "path": "/Users/user/materials/chart.png",
    "width": 6,
    "height": 3
  },
  "metadata": {
    "date": "2026-02-22",
    "author": "Sales Department",
    "version": "1.0"
  }
}

Workflow

Phase 1: Template Analysis

1. Receive the Word template file path from the user
2. Claude's Task: Guide the user to identify placeholders
3. Identify placeholder types:
   • Simple variables:

     {{name}}

   • Lists/loops:

     {% for item in items %}...{% endfor %}

   • Conditions:

     {% if condition %}...{% endif %}

   • Table rows:

     {%tr for item in items %}...{%tr endfor %}

Note: Actual parsing requires runtime execution. Claude guides the process and generates expected context structure.

1. Receive the Word template file path from the user
2. Use python-docx to read document content
3. Extract placeholders (

   {{variable_name}}

   )
4. Identify placeholder types:
   • Simple variables:

     {{name}}

   • Lists/loops:

     {% for item in items %}...{% endfor %}

   • Conditions:

     {% if condition %}...{% endif %}

Phase 2: Materials Folder Processing (Core Feature)

Claude's Task: Guide the user through materials organization and generate context JSON.

1. Materials Folder Structure Analysis

   • Scan and categorize files by type
   • Identify potential placeholder-to-file mappings
   • Flag missing or ambiguous mappings

2. Smart Data Extraction Guidance

   • Match placeholder names with filenames (e.g.,

     {{contract_info}}

     matches

     contract_info.txt

     )
   • Structure data according to JSON Schema
   • Validate data types and required fields

3. Data Mapping Strategy

   Placeholder Naming Convention (Recommended):
   - {{name}} -> name.txt or read from name.xlsx
   - {{items}} -> items.xlsx or items.csv (table data → JSON array)
   - {{logo}} -> logo.png (image path or object)
   - {{content}} -> content.docx (embed other Word content)
   

Claude's Output: Structured context JSON conforming to the schema

Phase 3: Context Generation & Validation

Claude's Task: Produce validated JSON context.

1. Collect Data from:

   • User input (direct text/values)
   • Material files (read and structured)
   • Default values where appropriate

2. Validate Against Schema:

   • Check required fields are present
   • Verify data types match expectations
   • Ensure arrays have correct structure

3. Generate Final JSON Output:

   • Follow the JSON Schema structure
   • Include all required fields
   • Handle optional fields appropriately
   • Provide validation status to user

Example Claude Output:

{
  "title": "Quarterly Sales Report",
  "project_info": "Analysis of Q1 2026 performance...",
  "data_items": [
    {"name": "Region A", "value": 125000},
    {"name": "Region B", "value": 98000}
  ],
  "chart_image": {"path": "/materials/chart.png"}
}

Phase 4: Document Generation (Runtime)

This phase requires an external runtime tool.

Runtime Tool's Task: Execute template rendering.

1. Load template using

   docxtpl

2. Parse the Claude-generated context JSON
3. Render template with Jinja2
4. Handle image embedding
5. Save the generated .docx file

See Runtime Reference section for implementation details.

Supported Placeholder Syntax

Simple Variables

Dear {{name}}:
Your order {{order_id}} has been confirmed.

List Loops

{% for item in items %}
- {{item.name}}: {{item.price}}
{% endfor %}

Conditional Logic

{% if is_vip %}
VIP Member Exclusive Offer
{% endif %}

Table Row Loops

| Product | Quantity | Unit Price |
|---------|----------|------------|
{%tr for product in products %}
| {{product.name}} | {{product.qty}} | {{product.price}} |
{%tr endfor %}

Runtime Reference Code

IMPORTANT: The following code is for runtime environment execution only. Claude cannot execute this code directly. It should be used by an external tool.

Read Template and Extract Placeholders

import re
from docx import Document

def extract_placeholders(template_path):
    """Extract placeholders from template - RUNTIME CODE"""
    doc = Document(template_path)
    text = '\n'.join([p.text for p in doc.paragraphs])
    # Extract text from tables
    for table in doc.tables:
        for row in table.rows:
            for cell in row.cells:
                text += cell.text + '\n'

    # Find {{...}} placeholders
    placeholders = re.findall(r'\{\{([^}]+)\}\}', text)
    return list(set([p.strip() for p in placeholders]))

Fill Template and Generate Document

from docxtpl import DocxTemplate

def generate_document(template_path, output_path, context):
    """Fill template with data and generate new document - RUNTIME CODE"""
    doc = DocxTemplate(template_path)
    doc.render(context)
    doc.save(output_path)

Complete Workflow (with Materials Folder Support)

from docx import Document
from docxtpl import DocxTemplate
from pathlib import Path
import re
import pandas as pd

def process_template_with_materials(template_path, materials_folder, output_path):
    """Fill template using data from materials folder - RUNTIME CODE"""

    # 1. Extract template placeholders
    doc = Document(template_path)
    full_text = get_full_text(doc)
    placeholders = extract_placeholders(full_text)

    # 2. Scan materials folder
    materials = scan_materials_folder(materials_folder)

    # 3. Smart data extraction
    context = {}
    for placeholder in placeholders:
        # Try to match by filename
        matched_data = find_data_for_placeholder(placeholder, materials)
        if matched_data is not None:
            context[placeholder] = matched_data

    # 4. Check for missing data
    missing = [p for p in placeholders if p not in context]
    if missing:
        print(f"Missing data for placeholders: {', '.join(missing)}")
        print("Please provide these data manually")

    return context

def scan_materials_folder(folder_path):
    """Scan materials folder - RUNTIME CODE"""
    materials = {}
    folder = Path(folder_path)

    for file in folder.iterdir():
        if file.is_file():
            # Use filename (without extension) as key
            key = file.stem
            materials[key] = {
                'path': str(file),
                'type': file.suffix.lower(),
                'file': file
            }
    return materials

def find_data_for_placeholder(placeholder, materials):
    """Find and extract data based on placeholder name - RUNTIME CODE"""
    # Direct filename match
    if placeholder in materials:
        mat = materials[placeholder]
        return extract_data_from_file(mat['path'], mat['type'])

    # Fuzzy match (e.g., contract_info matches contract_info_data.txt)
    for key, mat in materials.items():
        if placeholder.lower() in key.lower() or key.lower() in placeholder.lower():
            return extract_data_from_file(mat['path'], mat['type'])

    return None

def extract_data_from_file(file_path, file_type):
    """Extract data from file - RUNTIME CODE"""
    if file_type in ['.txt', '.md']:
        with open(file_path, 'r', encoding='utf-8') as f:
            return f.read()

    elif file_type == '.docx':
        doc = Document(file_path)
        return '\n'.join([p.text for p in doc.paragraphs])

    elif file_type in ['.xlsx', '.csv']:
        # Read Excel/CSV, return list of dictionaries
        df = pd.read_excel(file_path) if file_type == '.xlsx' else pd.read_csv(file_path)
        return df.to_dict('records')

    elif file_type in ['.png', '.jpg', '.jpeg']:
        # Return image path for template use
        return file_path

    return None

def generate_document(template_path, context, output_path):
    """Generate final document - RUNTIME CODE"""
    tpl = DocxTemplate(template_path)
    tpl.render(context)
    tpl.save(output_path)
    print(f"Document generated: {output_path}")

Usage Examples

Example 1: Simple Contract Generation

Template (contract_template.docx):

Contract No: {{contract_no}}

Party A: {{party_a}}
Party B: {{party_b}}

Date: {{date}}

User Input:

Generate a contract using this template:
/Users/user/contract_template.docx

Party A: Zhang San Company
Party B: Li Si Enterprise
Date: 2026-02-22
Contract No: HT-2026-001

Claude Generates (Context JSON):

{
  "contract_no": "HT-2026-001",
  "party_a": "Zhang San Company",
  "party_b": "Li Si Enterprise",
  "date": "2026-02-22"
}

Runtime Tool Output:

/Users/user/contract_20260222.docx

Example 2: Invoice Generation (with Tables)

Template (invoice_template.docx):

Invoice No: {{invoice_no}}

| Product | Quantity | Unit Price | Total |
|---------|----------|------------|-------|
{%tr for item in items %}
| {{item.name}} | {{item.qty}} | {{item.price}} | {{item.total}} |
{%tr endfor %}

Total: {{total_amount}}

User Input:

Generate invoice:
Template: /Users/user/invoice_template.docx
Invoice No: INV-001
Product List:
- Product A x 2 @ 100 yuan
- Product B x 3 @ 50 yuan

Claude Generates (Context JSON):

{
  "invoice_no": "INV-001",
  "items": [
    {"name": "Product A", "qty": 2, "price": 100, "total": 200},
    {"name": "Product B", "qty": 3, "price": 50, "total": 150}
  ],
  "total_amount": 350
}

Example 3: Auto-Generate with Materials Folder (Recommended)

Template (report_template.docx):

# {{title}}

## Project Information
{{project_info}}

## Data Analysis
{% for item in data_items %}
- {{item.name}}: {{item.value}}
{% endfor %}

## Chart
{{chart_image}}

Materials Folder Structure:

/Users/user/materials/
├── title.txt           # Content: Quarterly Sales Report
├── project_info.txt    # Content: Project details...
├── data_items.xlsx     # Table: name, value columns
└── chart.png           # Image file

User Input:

Generate document using this template and materials:
Template: /Users/user/report_template.docx
Materials: /Users/user/materials/

Claude's Processing:

1. Analyze template placeholders:

   {{title}}

   ,

   {{project_info}}

   ,

   {{data_items}}

   ,

   {{chart_image}}

2. Map materials to placeholders:

   • title.txt

     →

     {{title}}

   • project_info.txt

     →

     {{project_info}}

   • data_items.xlsx

     →

     {{data_items}}

     (parse as array)

   • chart.png

     →

     {{chart_image}}

     (image path object)
3. Validate structure against JSON Schema

Claude Generates (Context JSON):

{
  "title": "Quarterly Sales Report",
  "project_info": "Analysis of Q1 2026 sales performance across all regions showing 15% growth...",
  "data_items": [
    {"name": "Region A", "value": 125000},
    {"name": "Region B", "value": 98000},
    {"name": "Region C", "value": 156000}
  ],
  "chart_image": {
    "path": "/Users/user/materials/chart.png"
  }
}

Runtime Tool Output:

/Users/user/report_output.docx

Limitations

Claude (This Skill)

• Cannot directly execute Python code or manipulate .docx binaries
• Cannot read binary file contents directly
• Relies on user-provided file descriptions or structured inputs
• JSON generation depends on clear user input

Runtime Tool Requirements

• Only supports .docx format (not .doc)
• Complex formatting may require manual adjustment
• Limited support for nested tables
• Images require special handling (path reference or InlineImage)
• File naming in materials folder should have clear correspondence with placeholders
• Excel files need standardized table structure

Validation Criteria

Claude's Responsibilities (Validated by This Skill)

1. ✅ Analyzes template structure and identifies placeholders
2. ✅ Generates context JSON conforming to JSON Schema
3. ✅ Validates data types (string, number, boolean, array, object)
4. ✅ Maps materials to placeholders intelligently
5. ✅ Handles user input and converts to structured data
6. ✅ Provides clear validation status and error messages
7. ✅ Separates concerns between data generation and rendering

Runtime Tool's Responsibilities (External)

1. ✅ Reads Word template files using

   python-docx

2. ✅ Extracts placeholders from .docx content
3. ✅ Scans and reads files from materials folder
4. ✅ Renders templates using

   docxtpl

   with Jinja2
5. ✅ Handles image embedding and formatting
6. ✅ Generates formatted Word documents
7. ✅ Processes complex syntax (loops, conditions, tables)

Installation of Dependencies

For Runtime Environment Only - Claude cannot execute these commands.

To run the reference code in an actual Python environment:

pip install python-docx docxtpl openpyxl pandas pillow

Dependency Overview:

• python-docx

  : Reading .docx files

• docxtpl

  : Jinja2 template rendering for Word

• openpyxl

  : Excel file support

• pandas

  : Data manipulation and Excel/CSV parsing

• pillow

  : Image processing

Summary

This skill provides:

• ✅ Clear separation between Claude (data generation) and runtime tools (document rendering)
• ✅ JSON Schema definitions for validated context output
• ✅ Reference implementations for runtime environments
• ✅ Complete workflow guidance from template to generated document

Claude's Output: Validated JSON context conforming to the defined schema Runtime Tool's Role: Execute template rendering with the provided context