Ingest

The /ingest skill is the front door. It detects source type and delegates to specialized handlers.

Usage

/ingest https://some-article.com --vault my-research
/ingest path/to/paper.pdf --vault my-research
/ingest https://youtube.com/watch?v=abc --vault my-research

What happens

1. Detect source type (URL, PDF, YouTube, tweet, gist, Office, text)
2. Extract content via the appropriate handler
3. Save raw source to raw/ with descriptive filename
4. Create source-note page in wiki/sources/
5. Create/update entity pages in wiki/entities/ for people, orgs, tools mentioned
6. Create/update concept pages in wiki/concepts/ for key ideas
7. Update index.md and append to log.md
8. Auto-commit the vault

A single source typically creates 3-7 wiki pages.

7 Source Types

Source	Handler	Dependencies
Web articles	ingest-web	None (curl)
PDFs	ingest-pdf	poppler (auto)
Office docs	ingest-office	pandoc (auto)
YouTube	ingest-youtube	yt-dlp (auto)
Tweets	ingest-tweet	None (FXTwitter API)
GitHub Gists	ingest-gist	None (raw URL)
Text / notes	ingest-text	None