OCR and Documents

• You have a PDF (digital or scanned) and need its text extracted for summarization, archival, or further analysis.
• You need to OCR a scanned invoice, form, or photo of a document and pull out structured fields.
• You want to convert a DOCX file into plain text or markdown for use by another skill or pipeline.
• You are building a preprocessing step that feeds clean document text into search indexing or an LLM workflow.

• Your input is a PowerPoint/PPTX file — route those to a dedicated presentation skill instead.
• Your scanned documents are very low resolution, heavily skewed, or in an unsupported language; OCR quality will be poor and results unreliable.
• You need a no-dependency, zero-setup solution — this skill requires a Python environment and at least one document/OCR library.

Uses PyMuPDF or pdfminer to pull text directly from born-digital PDFs without OCR overhead, preserving layout as closely as possible.

Routes scanned PDFs and document photos through Tesseract or a compatible OCR service to recover text from image-based content.

Reads Microsoft Word files via python-docx, extracting paragraph and heading structure into clean text or markdown output.

Normalizes extracted content into either plain text or structured markdown, making output immediately consumable by summarization, indexing, or other downstream skills.

Automatically selects the appropriate extractor — direct text layer, OCR, or format-specific parser — based on the document type detected at runtime.

OCR a scanned invoice or filled form and return structured fields such as vendor name, amount, and due date for downstream processing or storage.

Extract text from a multi-page PDF and hand each page's content to a summarization skill, enabling document-wide summaries without manual copy-paste.

Parse a Word document and convert its content to markdown, making it ready for a knowledge base, static site, or further editing in a text-based workflow.

Batch-extract text from a collection of mixed PDFs and DOCX files to produce clean, normalized text chunks suitable for vector or full-text search indexing.