LLM Wiki Knowledge System

Problem

Personal knowledge often lives in scattered articles, notes, PDFs, and clippings. An LLM can read a single document, but it usually lacks a durable maintenance loop, so ideas duplicate, conflict, or lose provenance over time. This skill turns raw material into a navigable Obsidian knowledge base.

How it works

• Init: after you provide the vault path, it creates SCHEMA.md, raw/, wiki/, index.md, and log.md.
• Ingest: reads an article, file, or directory, extracts viewpoints, entities, concepts, cases, and actions, then writes summary pages and updates entity/concept pages.
• Links: uses [[wikilinks]] so related pages can be followed during later queries.
• Query: locates relevant pages from the index, reads multi-level context, and answers with citations; useful answers can become analysis pages.
• Lint: checks conflicts, orphan pages, missing pages, stale content, weak cross-references, and index consistency.

Boundaries

It depends on an Obsidian vault path and requires explicit first-time setup. raw/ is treated as read-only, so source material is not modified. It fits personal or team note organization, knowledge synthesis, and Q&A, but it is not a search engine, database, or automatic content production system.