llm-wiki

The llm-wiki skill provides a structured framework for building and maintaining an AI-powered knowledge base within Obsidian. It addresses the challenge of organizing, synthesizing, and preserving knowledge in a way that is persistent and compounding, rather than re-derived on each query. By applying a three-layer architecture, it separates raw source material from distilled, curated content, ensuring that the knowledge base remains authoritative and navigable. This skill is designed as a theoretical foundation, guiding users on best practices for knowledge management and wiki construction without performing direct ingestion or querying operations itself.

The main features of llm-wiki include a three-layer knowledge architecture consisting of raw sources, the LLM-maintained wiki, and the schema that governs wiki structure. Raw sources are immutable documents and images, serving as the original authoritative content. The wiki layer compiles these sources into interconnected markdown files with clear provenance, Obsidian-compatible wikilinks, and structured frontmatter metadata. The schema layer defines categories, templates, and operational rules, ensuring consistency across the knowledge base. Wiki organization is divided into categories for types of knowledge and projects for project-specific content, supporting both global and scoped knowledge management.

This skill is suitable for knowledge workers, researchers, developers, and anyone looking to establish a robust personal or organizational wiki. Typical use cases include setting up a new knowledge base, designing the structure of an Obsidian vault, deciding on knowledge categorization, and creating guidelines for maintaining distilled knowledge over time. Users benefit from guidance on organizing complex information, ensuring traceable provenance, and creating a system that grows and compounds as new knowledge is added.