llm-wiki defines the knowledge distillation pattern that the other obsidian-wiki skills build on. Raw sources stay unchanged, the wiki is a compiled and cross-referenced set of Obsidian markdown pages, and the schema says how the agent maintains them.
It covers vault folders and projects, special files, the page template, a paper deep-dive template, provenance markers, typed relationships, confidence and lifecycle fields, importance tiers, cheap retrieval rules and how config is resolved across vaults.
What You Get
- A page template with frontmatter for summary, provenance, confidence, lifecycle, tier and relationships
- Inline provenance markers that separate extracted, inferred and ambiguous claims
- A retrieval ladder that reads the cheapest source first and escalates only when needed
- A config resolution protocol covering multiple vaults, per-request vault overrides and a writing profile
Common Use Cases
- Decide how to organize a new vault into categories and projects
- Learn the conventions every other wiki skill assumes
- Add a paper to the wiki using the deep-dive template
- Set up several vaults and route a request to the right one
What Makes It Unique
It is a schema skill, and the rest of the repository points back to it for config resolution, link format, provenance, confidence formulas and the retrieval ladder. It makes trust visible with inline markers, a lifecycle where only humans promote a page, and importance tiers that limit how much each ingest rereads.
Under The Hood
The skill is reference material for the agent, with no scripts of its own. It tells the agent how to resolve config files, how to write index, log and hot files through the obsidian-wiki memory command, how to read and write manifest entries and page frontmatter, and when to refresh a QMD search index. It also shows shell snippets for finding config and computing a vault state folder. Applying it means reading and writing markdown inside your vault. It makes no web requests.