PDF

What This Skill Helps With

Use this skill when you need to extract text, capture tables, fill forms, split or merge files, or diagnose why a PDF is difficult to parse.

Ask for this skill with prompts like

Use the exact skill name $pdf in your prompt.

• Use $pdf to extract every table from this PDF into CSV.
• Use $pdf to split this file into one PDF per page.
• Use $pdf to merge these three PDFs into a single packet.
• Use $pdf to fill this form and save a completed copy.

Tools and Assumptions

This skill works best when the agent can use some of these tools:

• pypdf for splitting, merging, rotating, and basic metadata work
• pdfplumber for text and table extraction
• OCR tooling for scanned PDFs when there is no machine-readable text

If extraction returns almost nothing, assume the PDF may be image-based and switch to OCR.

Recommended Workflow

1. Check whether the PDF is text-based or scanned

• Try normal text extraction first.
• If the result is empty or unreadable, the document is probably scanned.
• For scanned PDFs, use OCR or a vision-capable workflow instead of repeating text extraction blindly.

2. Extract text or tables

• Use pdfplumber when layout and tables matter.
• Use pypdf for simpler metadata or page-level operations.
• Export tables into CSV, JSON, or markdown so the result is usable.

Example:

import pdfplumber

with pdfplumber.open("input.pdf") as pdf:
    first_page = pdf.pages[0]
    text = first_page.extract_text()
    tables = first_page.extract_tables()

3. Split, merge, or rotate pages

• Use pypdf for page operations.
• Preserve the original file until the new output is verified.

4. Fill forms

• First detect whether the PDF has actual form fields.
• If it does, fill them programmatically.
• If it does not, explain that the file is visually form-like but not machine-fillable.

Important Limits

• Table extraction quality depends on how the PDF was authored.
• Scanned PDFs often need OCR before they become searchable.
• Some PDFs look like forms but do not contain real form fields.
• Complex or encrypted PDFs may require a more manual workflow.

Common Mistakes

• Assuming every PDF contains selectable text.
• Treating image scans like text PDFs.
• Extracting tables without checking for merged cells or header rows.
• Overwriting the original before confirming page order and orientation.