doc-processing

name	doc-processing
description	Extract and convert documents (PDF, DOCX, etc.) into the sources/docs knowledge base. Use when processing a document from a URL or local file into a structured source folder with metadata and markdown content.
user-invocable	false

Document Extraction

Extract a document into sources/docs/ with structured metadata and markdown content.

For folder layout, slug convention, and meta.md template see references/structure.md.

Process

1. Obtain document and create folder

Determine slug (YYYY-MM-short-descriptive-slug), create folder structure.

If URL — download into _raw-source/:

uv run .claude/skills/doc-processing/scripts/doc-download.py '<URL>' 'sources/docs/<slug>/_raw-source'

If local file — copy into _raw-source/:

mkdir -p sources/docs/<slug>/_raw-source
cp <local-path> sources/docs/<slug>/_raw-source/

2. Convert to markdown

uv run .claude/skills/doc-processing/scripts/doc-convert.py 'sources/docs/<slug>/_raw-source/<filename>' 'sources/docs/<slug>/content.md'

3. Create meta.md

Write initial meta.md using the template from references/structure.md. Fill in known fields: title, URL, author, retrieved date, format.

4. Enrich metadata

Read content.md and enrich meta.md with:

• Tags: 3-8 keywords derived from the document's topics
• Summary: Exactly 2 sentences — what it covers and why it matters
• Published: Extract from content if available

Follow the field rules in references/structure.md.

5. Update index

Append entry to sources/index.md (see column format in sources/README.md).

Script Interfaces

Script	Input	Output (stdout)
doc-download.py	<url> <output-folder>	file:, size_kb:, filename:
doc-convert.py	<input-file> <output-file>	file:, size_kb:, lines:, chars:

Both use uv run with inline dependencies — no global installs needed.