# raw/ — Verbatim Machine Output Only

**NEVER modify files in this folder.** Everything here is the untouched output
of a mechanical extraction step, kept so every cleaned source file and wiki
claim can be traced back to what the machine actually produced.

Conventions:

- Subfolders start with an underscore and are named after the extraction tool:
  - `_pdf/` — original PDFs (gitignored; only text extractions are committed)
  - `_pdftotext/` — verbatim `pdftotext` output (text-layer PDFs)
  - `_ocr/` — verbatim OCR output (image-based PDFs)
  - Add more as needed for your source formats (`_html/`, `_transcripts/`, …)
- Cleaned, human/LLM-readable versions live in `source/`, never here.
- See `meta/guides/pdf-extraction.md` for the PDF extraction workflows
  (delete that guide if your domain's sources are not PDFs).