/doc2md

Local document-to-markdown converter. Zero LLM tokens spent on conversion.

Default behavior

No arguments:

1. Processes .raw-docs/ → saves output to docs/
2. Only converts new or changed files (MD5 cache)
3. Audio files (.mp3, .wav) → tells user to use /whisper

With arguments: follows the explicit command.

Usage

/doc2md                            # .raw-docs/ → docs/ (default)
/doc2md <file.pdf>                 # single file
/doc2md <folder>                   # entire folder (recursive)
/doc2md --url <url>                # YouTube, Wikipedia, HTML
/doc2md <folder> --watch           # auto-convert on new files
/doc2md <folder> --out <dir>       # custom output folder
/doc2md <file> --force             # skip cache
/doc2md <folder> --list            # show cache status

Supported formats

.pdf .docx .pptx .xlsx .xls .csv .epub .ipynb .msg .zip

URLs: YouTube (full transcript), Wikipedia, generic HTML.

Steps when invoked

1. Check dependency

python -c "from markitdown import MarkItDown; print('ok')"

If missing: pip install markitdown[all]

2. Run the converter

python ~/.claude/skills/doc2md/scripts/convert.py <input> [flags]

Returns JSON. Parse and display as a results table.

3. Report to user

✓ report.pdf  → report.md  (2,309 words, ~1,732 tokens saved)
⚠️ memo.pdf   (cached, 587 words)
✗ audio.mp3   → use /whisper

graphify integration

/doc2md           # 0 tokens: convert docs to .md locally
/graphify docs/   # reads .md instead of PDF: 60-80% fewer tokens