doc-to-markdownlisted

# Reading and Converting Documents ## Instructions ### 1. Decide whether to convert at all Conversion is a cost, not a default. Read the file directly when the question is bounded and the runtime can open the format — direct reading preserves figures, equations, and layout that no text conversion keeps, and lets you work a page range at a time. Where the runtime cannot open PDFs directly, convert, but convert only the pages you need. **Read directly** when the document is under ~20 pages and the task is one pass over it (summarize, answer a question, check a claim, review a draft), when the payload is visual (figures, plots, scanned tables, slides), or when you need a page number you can cite. **Convert** when the document is longer than one read comfortably spans, when the text will be searched or quoted repeatedly, when it joins a corpus or source library, when many documents get the same treatment, or when a downstream tool (grep, a classifier, a script) needs plain text. `$citation-check` and `$fact-check` both require converted Markdown, one file per source. State which path you took in one line. Silently converting a two-page memo wastes a step; silently reading page 1 of a 300-page book and answering as if you read it all is worse. ### 2. Route by what the document actually is, not by its extension Probe first — the extension says nothing about structure: ```bash pdfinfo doc.pdf | grep -E '^(Pages|Encrypted)' pages=$(pdfinfo doc.pdf | awk '/^Pages:/{print $2}'