office trained

which guide owns which document format, and how to verify an office artifact instead of trusting the write.

mental model

Route by source and deliverable format: docx, pdf, pptx, or xlsx/csv/tsv. Cross-format work chains the guides in data order. Resolve bundled scripts relative to their guide's directory.

examples

An in-memory PDF creation and extraction check:

from io import BytesIO
from reportlab.pdfgen import canvas
from pypdf import PdfReader
out = BytesIO()
page = canvas.Canvas(out)
page.drawString(72, 720, "probe")
page.save()
out.seek(0)
assert PdfReader(out).pages[0].extract_text().strip() == "probe"

best practices

Preserve an existing document's style and semantic structure. Render and inspect every page or slide when layout matters. For spreadsheets, keep formulas where inputs should remain editable; recalculate and inspect cached values only with an available calculation engine. pypdf docs, LibreOffice headless options.

strengths

OOXML documents (docx, pptx) support targeted edits; PDF tools support extraction, forms, and fixed-layout delivery; spreadsheet tools support formulas and tabular analysis. OOXML standard, ReportLab guide.

weaknesses / pain points

A valid OOXML ZIP can still render badly. PDF text extraction can lose reading order, while scanned PDFs need OCR. Spreadsheet formulas written by openpyxl do not supply recalculated cached values. openpyxl formulas.

gotchas

The bundled OOXML scripts and schemas live inside this guru, not the user's project. LibreOffice is often absent from a machine; check command -v soffice before promising PDF conversion or formula recalc.

known bugs

No version-specific upstream defect is known. A missing local LibreOffice installation is an environment fact, not an upstream bug.

troubleshooting

If a script path is missing, resolve it from the selected guide's directory.
If formula values are absent, run a calculator if one is installed and compare formula cells with cached values; otherwise state the verification gap.
For overflow or clipping, render the artifact, inspect it, revise, and render again.

practiced cases

ReportLab 5.0.0 creates a one-page PDF in memory and pypdf 6.14.2 extracts probe from it; lxml 6.0.3 compiles both bundled DOCX and PPTX wml.xsd schemas. These probes do not verify a user document, page layout, spreadsheet calculation, or form filling.

Read the office skill.