Handles the messy reality of data that doesn't fit anywhere else. ChatGPT exports, Slack threads, Discord logs, meeting transcripts, CSV dumps, browser bookmarks, even folders of screenshots. It figures out the format on the fly, extracts what matters, and writes wiki pages instead of dumping raw text into your vault. The interesting part is the content trust boundary: it treats all source data as untrusted input, so instructions buried in chat logs or JSON payloads get distilled into content, not executed. Good for when you've got a pile of unstructured text and you want knowledge pages, not file clutter. Works with vision models for image sources.
npx skills add https://github.com/ar9av/obsidian-wiki --skill data-ingest