SuperAPI AI Directory

document-spreadsheet-tooling-readiness

Office & productivity

View repository

Updated 10-16 08:03

About this skill

Document Spreadsheet Tooling Readiness

Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.

Use When

  • Before promising or creating .docx, .pdf, .xlsx, application registers, scoring matrices, budgets, dashboards, price schedules, CV packs, reports, or annexes.
  • When moving an engine to a new machine and confirming whether document/spreadsheet tooling is installed.

Do Not Use When

  • The user only asks for plain Markdown or a conceptual outline and no generated file is claimed.

Required Inputs

  • Desired output formats, local tooling access, plugin/connector availability, and any required templates.

Workflow

  • Check plugins/connectors first, then Python packages, then binaries.
  • Run a one-file smoke test before production export.
  • Choose a route and state fallback honestly.
  • Never claim a file exists until it was generated and opened or validated.

Quality Standards

  • File-output claims are factual claims.
  • Fallbacks are explicit, not silent.
  • PDF conversion route must be known before promising PDF.

Anti-Patterns

  • Saying a Word, PDF, or Excel file was generated when only Markdown exists.
  • Assuming LibreOffice, Pandoc, or a plugin exists without checking.
  • Shipping a workbook without reopening or formula/control checks.

Outputs

  • Toolchain check, smoke-test result, route decision, fallback plan, and validation note.

References

  • references/toolchain-checks.md: Commands and package/binary checklist.
  • references/fallback-routes.md: Output routes and fallback decision tree.

Core Workflow

  1. Confirm requested formats and whether a generated file is actually required.
  2. Prefer built-in document/spreadsheet plugins where available.
  3. Check Python packages: openpyxl, XlsxWriter, pandas, python-docx, docxtpl, docxcompose, pypandoc, markdown, beautifulsoup4, lxml, pillow, PyMuPDF, pypdf, pdfplumber, reportlab.
  4. Check binaries: pandoc, soffice/LibreOffice, wkhtmltopdf where needed, and tesseract for OCR tasks.
  5. Run a minimal DOCX and XLSX smoke test before production output on a new machine.
  6. Pick the generation route and record it in the deliverable notes.
  7. Open, parse, or otherwise validate the generated file before telling the user it was created.