Document Generator

Document Generator Process

Core process skill for generating professional business documents. Defines the end-to-end workflow from user request to delivered file.

Generation Workflow

Step 1: CLASSIFY

Determine the document type from the user's request:

Step 1.5: CHECK USER PREFERENCES (MANDATORY)

You MUST check preferences before any generation. Scripts will reject requests if preferences are missing.

cat ~/.document-generator/preferences.json 2>/dev/null

• File missing → STOP. Run the onboarding interview (see user-preferences skill) before proceeding.
• File exists → load style preset, default company, language, and currency to pre-fill fields

Step 2: GATHER

Collect required data from the user. Each document type has specific required and optional fields.

Refer to the document-templates skill for complete field checklists per document type. It defines exactly which fields are required, which are optional, and the expected format for each.

If user preferences are loaded, pre-fill company info, currency, and language from the stored profile. Only ask for fields that aren't already known.

Step 3: SELECT FORMAT & ENGINE

Engine selection for DOCX:

• docx-js (default): Always works, full feature support (images, complex tables, headers/footers)
• pandoc: Produces DOCX that visually matches PDF output (uses same HTML templates). Set "engine": "pandoc" in input JSON. Requires pandoc installed (which pandoc).

Step 4: BUILD JSON Input

1. Read the template file from <plugin_dir>/templates/{type}_template.json
2. Merge user data into the template structure
3. Set outputPath to: {cwd}/{type}_{sanitized_title}_{YYYY-MM-DD}.{ext}
4. Write the complete JSON input to a temp file: {cwd}/.doc_input.json

Input JSON structure:

{
  "type": "proposal|invoice|estimate|report|contract|nda|act",
  "engine": "playwright|pdfkit",
  "outputPath": "./proposal_acme_2026-03-17.docx",
  "template": { "...from template file..." },
  "data": { "...user provided data..." }
}

Step 5: GENERATE

Run the appropriate script via Bash:

cd <plugin_dir> && node scripts/generate_docx.js /absolute/path/to/.doc_input.json

Check dependencies first (use the centralized checker):

cd <plugin_dir> && node scripts/check_deps.js

If ready: false, show what's missing and ask user permission to install.

Script output (JSON to stdout):

• Success: { "success": true, "outputPath": "/abs/path/to/file.docx", "size": 12345 }
• Failure: { "success": false, "error": "description" }

Step 6: DELIVER

1. Parse the script's JSON output
2. Confirm the file exists and report its path and size
3. Offer follow-up actions: "Want me to convert this to PDF?" or "Want to make changes?"

Format Conversion

For converting between formats, use the pandoc wrapper:

<plugin_dir>/scripts/convert.sh input.md output.pdf

Supported conversions: MD->PDF, MD->DOCX, MD->HTML, DOCX->PDF, DOCX->MD, HTML->PDF, HTML->DOCX, MD->PPTX

Error Handling

File Naming Convention

Pattern: {type}_{title}_{date}.{ext}

Sanitize title: lowercase, replace spaces with underscores, remove special characters, max 50 chars.

Examples:

• proposal_acme_consulting_2026-03-17.docx
• invoice_inv-001_2026-03-17.pdf
• report_q1_analysis_2026-03-17.docx
• presentation_product_launch_2026-03-17.pptx
• contract_service_agreement_2026-03-17.docx
• act_akt-001_2026-03-17.pdf