llm-wiki

name	llm-wiki
description	Builds and maintains a persistent, LLM-maintained wiki inside an Obsidian vault — the LLM writes/maintains all pages, the human curates sources and asks questions. Ingest a source, query across pages, or lint for stale/contradictory/orphan content. Vault root comes from the overlay (config/knowledge.toml). Use when ingesting an article/paper/transcript into the vault, synthesizing across notes, health-checking the wiki, or building a knowledge structure for a domain. Do not use for code documentation, one-off notes, or non-vault knowledge.
allowed-tools	Read, Write, Edit, Bash, WebFetch

LLM Wiki

A pattern for a persistent, LLM-maintained wiki inside an Obsidian vault. The LLM writes and maintains all wiki pages; the human curates sources, asks questions, and directs the analysis.

Vault is not hardcoded. The vault root lives in the overlay — resolve it first, never assume a path:

VAULT=$(uv run --python 3 python -c 'import tomllib,os;p=os.path.expanduser(os.environ["AXON_PERSONAL_ROOT"])+"/config/knowledge.toml";print(os.path.expanduser(tomllib.load(open(p,"rb"))["vault_root"]))')

All $VAULT/... paths below resolve against that.

When to use

• Ingest a new source (article, paper, book chapter, transcript) into the knowledge base
• Answer a question that needs synthesis across multiple wiki pages
• Health-check / lint the wiki for stale, contradictory, or orphaned content
• Build out a knowledge structure for a new domain

What this skill does

1. Reads the LLM Wiki project note at $VAULT/Projects/LifeOS/LLM Wiki.md for current schema conventions, page formats, and domain config.
2. Reads index.md in the relevant Knowledge subdirectory to see what pages exist.
3. Ingest: reads the source, extracts key info, updates all affected pages (summaries, entities, concepts, cross-refs, index, log).
4. Query: searches the wiki, reads relevant pages, synthesizes an answer with citations, optionally files it back as a page.
5. Lint: scans for contradictions, stale claims, orphan pages, missing cross-references, data gaps.

Wiki structure

Under $VAULT/Knowledge/<domain>/llm-wiki/:

index.md              -- catalog of all pages with links + one-line summaries
log.md                -- append-only chronological record of all operations
sources/source-<title>.md
entities/entity-<name>.md
concepts/concept-<name>.md
comparisons/comparison-<topic>.md
queries/query-<topic>.md

A flat form (index.md, log.md, <topic>.md) is valid when the domain is narrow.

Schema conventions

Every page gets YAML frontmatter following vault conventions:

---
type: note
maturity: seedling | growing | evergreen
summary: "One-line description of what this page covers"
categories:
  - "[[Category Link]]"
related:
  - "[[Related Page]]"
sources:
  - "Source title or URL"
---

Maturity: seedling = new, single source, likely incomplete · growing = multiple sources synthesized, cross-refs in place · evergreen = stable, well-cross-referenced, contradictions resolved.

Operations

Ingest a source

1. Ask the user to place the source in $VAULT/raw/<domain>/ (or accept a URL to fetch → save to $VAULT/raw/<domain>/<title>.md)
2. Read the source; discuss key takeaways with the user
3. Determine pages to create/update: entity pages, concept pages, the source summary, cross-ref updates, index + log
4. Present the planned changes before writing, then write/update all pages and append to log.md

Answer a question (query)

1. Read index.md to find relevant pages → read them
2. Synthesize an answer with citations to specific wiki pages
3. If it adds lasting value, file it as queries/query-<topic>.md and update the index

Lint the wiki

1. Walk all wiki pages; check for contradictions, stale claims, orphan pages, concepts lacking a page, missing cross-refs
2. Surface results with suggested fixes; apply approved fixes

Tips

• Mermaid for architecture/flow/comparison diagrams (Obsidian-native)
• Dataview over YAML frontmatter for dynamic tables
• MOCs link pages into broader domains via categories
• One source at a time for ingest — keeps the human-in-the-loop quality
• Take a backup/dump of affected pages before a schema change