document-generator-docs

name	document-generator-docs
description	Use the Document Generator app CLI or MCP tool to create DOCX, PPTX, PDF, and XLSX files in workspace generated storage, or extract text from workspace documents.

Document Generator Docs

Use this skill when a user asks an agent to create a document file or read text from a workspace document.

App Scripts

The app exposes one official CLI command plus MCP tools for generation, extraction, references, and view state. Use those surfaces for real work.

Implementation scripts behind those surfaces:

• CLI entrypoint: <repo>/apps/document-generator/cli/app_cli.py
• MCP entrypoint: <repo>/apps/document-generator/mcp/server.py
• shared service: <repo>/apps/document-generator/backend/service.py
• text extractor: <repo>/apps/document-generator/backend/extractors.py
• DOCX writer: <repo>/apps/document-generator/backend/generators/docx_generator.py
• PPTX writer: <repo>/apps/document-generator/backend/generators/pptx_generator.py
• PDF writer: <repo>/apps/document-generator/backend/generators/pdf_generator.py
• XLSX writer: <repo>/apps/document-generator/backend/generators/xlsx_generator.py

Do not call generator scripts directly. They are implementation details used by the CLI, MCP, and backend service.

Output Formats

Supported formats:

• docx
• pptx
• pdf
• xlsx

Text extraction supports the same modern document formats:

• pdf
• docx
• pptx
• xlsx

Do not request xls; this app intentionally supports only xlsx for spreadsheets.

CLI Procedure

Use command id:

app.document-generator.document-generator

Call the command with action validate_spec before generation when the document structure is non-trivial. Then call it again with action generate_document.

The generated file is written under storage/generated/. Report the returned workspace_relative_path to the user.

The CLI arguments are JSON. In a Maverick CLI tool invocation, pass the JSON object as the command arguments. If you need a shell smoke test inside the repository, invoke the core CLI service rather than the private generator scripts.

DOCX

Use DOCX for narrative documents, letters, briefs, proposals, company profiles, reports, and documents that need headings and paragraphs.

Validate:

{
  "action": "validate_spec",
  "spec": {
    "format": "docx",
    "title": "Quarterly Brief",
    "output_filename": "quarterly-brief.docx",
    "sections": [
      {"heading": "Summary", "text": "Key points for the quarter."},
      {"heading": "Next steps", "text": "Actions and owners."}
    ],
    "tables": [
      {"rows": [["Metric", "Value"], ["Revenue", "Growing"]]}
    ]
  }
}

Generate:

{
  "action": "generate_document",
  "spec": {
    "format": "docx",
    "title": "Quarterly Brief",
    "output_filename": "quarterly-brief.docx",
    "sections": [
      {"heading": "Summary", "text": "Key points for the quarter."},
      {"heading": "Next steps", "text": "Actions and owners."}
    ],
    "tables": [
      {"rows": [["Metric", "Value"], ["Revenue", "Growing"]]}
    ]
  }
}

PPTX

Use PPTX for slide decks, pitch material, meeting presentations, product walkthroughs, and executive summaries.

Generate:

{
  "action": "generate_document",
  "spec": {
    "format": "pptx",
    "title": "Product Update",
    "output_filename": "product-update.pptx",
    "slides": [
      {"title": "Overview", "bullets": ["Progress this month", "Key risks", "Next priorities"]},
      {"title": "Roadmap", "bullets": ["Launch beta", "Collect feedback", "Prepare release"]}
    ]
  }
}

PDF

Use PDF for fixed-layout summaries, externally shared briefs, one-page statements, and read-only deliverables.

Generate:

{
  "action": "generate_document",
  "spec": {
    "format": "pdf",
    "title": "Executive Summary",
    "output_filename": "executive-summary.pdf",
    "sections": [
      {"heading": "Context", "text": "The project is entering the validation phase."},
      {"heading": "Recommendation", "text": "Proceed with the next milestone after stakeholder review."}
    ]
  }
}

XLSX

Use XLSX for spreadsheets, structured data, tables, trackers, exports, financial models, and simple datasets.

Generate:

{
  "action": "generate_document",
  "spec": {
    "format": "xlsx",
    "title": "Budget Tracker",
    "output_filename": "budget-tracker.xlsx",
    "sheets": [
      {
        "name": "Budget",
        "rows": [
          ["Category", "Planned", "Actual"],
          ["Design", 5000, 4200],
          ["Development", 12000, 11800]
        ]
      }
    ]
  }
}

MCP Procedure

Use app.document-generator.maverick_document_generator with the same arguments as the CLI.

For document text extraction, prefer:

• CLI command app.document-generator.document-generator with action extract_text
• MCP tool app.document-generator.document_generator_extract_text

Extraction inputs should identify a workspace file under storage/uploaded/ or storage/generated/:

{
  "action": "extract_text",
  "workspace_relative_path": "storage/uploaded/file-id/example.pdf",
  "max_chars": 50000
}

Expected extraction shape:

{
  "status_code": 200,
  "document": {
    "format": "pdf",
    "filename": "example.pdf",
    "workspace_relative_path": "storage/uploaded/file-id/example.pdf"
  },
  "text": "Extracted document text...",
  "text_length": 1234,
  "truncated": false
}

For discovery or follow-up references, use:

• app.document-generator.document_generator_reference_manifest
• app.document-generator.document_generator_reference_search
• app.document-generator.document_generator_reference_resolve
• app.document-generator.document_generator_reference_summarize

Result Handling

After generation:

1. Check status_code.
2. Read document.workspace_relative_path.
3. Confirm the file exists under storage/generated/.
4. Return the workspace-relative path to the user.

Expected success shape:

{
  "status_code": 200,
  "document": {
    "format": "docx",
    "filename": "quarterly-brief.docx",
    "workspace_relative_path": "storage/generated/quarterly-brief.docx"
  }
}