Document Toolkit (PDF · DOCX · XLSX)

Handle documents with the right Python library, not regex over raw bytes. Install what you need:

pip install pypdf pdfplumber python-docx openpyxl

Pick by file type, then by task. Work on a copy; never overwrite the user's original without confirming.

PDF

Extract text/tables — pdfplumber is best for layout-aware text and tables:

import pdfplumber
with pdfplumber.open("in.pdf") as pdf:
    text = "\n".join(p.extract_text() or "" for p in pdf.pages)
    tables = pdf.pages[0].extract_tables()

Merge / split / rotate — pypdf:

from pypdf import PdfReader, PdfWriter
w = PdfWriter()
for f in ["a.pdf", "b.pdf"]:
    for page in PdfReader(f).pages:
        w.add_page(page)
with open("merged.pdf", "wb") as fh:
    w.write(fh)

Scanned PDFs need OCR first (ocrmypdf in.pdf out.pdf) before text extraction. Fill forms: read field names with reader.get_fields(), then writer.update_page_form_field_values(...).

Word (.docx)

from docx import Document
doc = Document("in.docx")                      # read
full = "\n".join(p.text for p in doc.paragraphs)
for table in doc.tables:
    for row in table.rows:
        print([c.text for c in row.cells])

out = Document()                               # create
out.add_heading("Report", level=1)
out.add_paragraph("Body text.")
t = out.add_table(rows=1, cols=2); t.style = "Light Grid"
out.save("out.docx")

python-docx covers headings, paragraphs, runs (bold/italic), tables, images (add_picture), and styles. It does not render to PDF — convert with libreoffice --headless --convert-to pdf out.docx if needed.

Excel (.xlsx)

from openpyxl import load_workbook, Workbook
wb = load_workbook("in.xlsx", data_only=True)  # data_only → values, not formulas
ws = wb.active
for row in ws.iter_rows(values_only=True):
    print(row)

wb2 = Workbook(); ws2 = wb2.active             # create
ws2.append(["Name", "Score"]); ws2.append(["A", 91])
ws2["C1"] = "=SUM(B2:B100)"                     # formulas as strings
wb2.save("out.xlsx")

openpyxl handles sheets, formulas, styles, number formats, charts, and freezing panes. For heavy numeric work, read with pandas.read_excel and write with df.to_excel.

Rules

• Library over shell. Don't grep/sed a binary doc — use the library.
• Work on a copy; confirm before overwriting an original.
• Verify the output — open the produced file (extract its text / list its rows) and confirm it's what was asked, not just that the script ran.
• For big batches, write a small script and run it once, rather than many ad-hoc one-liners.

Credit

Approach and library choices follow the pdf / docx / xlsx skills in anthropics/skills.