heavy-file-ingestion

Convert PDFs and DOCX files into markdown using local deterministic converters like pdfplumber and python-docx. The process extracts structured text and generates markdown artifacts alongside an index to minimize token consumption during agent processing.

Feeding large spreadsheets wastes context because raw XLSX and CSV files contain verbose formatting that consumes excessive tokens. Converting them to per-sheet CSV artifacts with a lightweight index prevents wasting costly model context before any reasoning step begins.

Yes, you can extract text from PPTX slide decks for automated summarization using the python-pptx library. The converter transforms presentation content into markdown, allowing agents to efficiently read and summarize slides without processing raw binary formats.

markitdown is an optional tool for converting heavy files for agents. The Skill relies primarily on deterministic local converters like pdfplumber, python-docx, python-pptx, and openpyxl, using markitdown as an optional escalation path when needed.

The best way to index large vendor invoices is converting the source PDF into markdown or CSV, then generating an index.json with counts, preview lines, and quality flags. This allows the main agent to decide whether to read extracted artifacts or escalate to a stronger model.

You should avoid automated file conversion when the source file is already lightweight plain text, or when precise layout rendering is critical. The converter produces quality flags in the index to indicate when extraction may be insufficient and escalation to a stronger model is warranted.