skills.rest
ninthday
ninthdayCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

nice-llm-wiki

Build and maintain a personal LLM-powered knowledge base from immutable source snapshots.

Updated Mar 30, 2026
Search Tags:#markdown#knowledge base#citations#obsidian#lint#ingest#llm wiki
One-click install
npx skills add https://github.com/ninthday/skills-base --skill nice-llm-wiki-ninthday
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: nice-llm-wiki
Source: https://github.com/ninthday/skills-base/tree/main/skills/nice-llm-wiki
Command: npx skills add https://github.com/ninthday/skills-base --skill nice-llm-wiki-ninthday

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Scattered notes and repeated re-reading of source material make personal knowledge hard to compound. This Skill turns an LLM into the maintainer of a structured markdown wiki: sources are snapshotted once into an immutable raw/ directory, then compiled into durable, cross-linked wiki articles that improve over time instead of being re-derived on every question. ## Core Features & Use Cases - Ingest & Compile: Snapshot web pages, files, pasted text, or Obsidian Web Clipper Markdown from a Clippings/ queue into raw/, then create or update Traditional Chinese wiki articles with triage dispositions (New, Update, Disputed, No material). - Grounded Query: Search the wiki and answer questions with citations linking back to wiki pages, with optional archiving of answers as point-in-time pages. - Lint & Evidence Checking: Auto-fix index inconsistencies and broken links, and run scripts/check_evidence.py to verify that every number, date, and quote in articles appears verbatim in the linked raw sources. - Use Case: Save research articles about a technology into Clippings/, ask the agent to process pending clippings, then later ask "what do I know about X" and receive a cited answer synthesized from your compiled wiki. ## Quick Start Ingest this article into my wiki and compile it into knowledge pages: https://example.com/some-article

Frequently Asked Questions about nice-llm-wiki

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a personal knowledge base with an LLM?▼

Use the ingest workflow: give the agent a URL, file, or pasted text, and it snapshots the source into raw/ then compiles durable markdown articles into wiki/. The wiki compounds over time as new sources update and cross-link existing articles.

What is the difference between an LLM wiki and RAG?▼

RAG stores raw chunks and embeddings and synthesizes answers at query time, while an LLM wiki stores curated markdown pages synthesized during ingest and maintenance. The wiki model suits compounding knowledge, summaries, and durable cross-links rather than broad retrieval over large corpora.

How do I process Obsidian Web Clipper clippings into the wiki?▼

Save clips as Markdown under Clippings/, then explicitly ask to list or process pending clippings. Each clipping is snapshotted into raw/, compiled into wiki articles, and deleted only after all updates succeed; nothing runs automatically in the background.

Does this skill work with Claude Code, Cursor, and Codex?▼

Yes, it follows the agentskills.io open standard and installs via npx skills add for Claude Code, Cursor, and OpenCode, or by copying SKILL.md, references/, and scripts/ into the tool's skill directory for Codex CLI and other tools.

How does the wiki prevent hallucinated facts in articles?▼

A grounding invariant requires every number, date, and quote to be located verbatim in the linked raw file before being written. The scripts/check_evidence.py lint pass mechanically greps high-signal literals against raw sources and reports suspects and evidence errors.

What are the limitations of the LLM wiki approach?▼

It deliberately omits vector or graph search, source-hash freshness tracking, numeric confidence scores, and automatic scheduled runs. At 50K-100K tokens of curated wiki, grep and read are considered more reliable than embedding-based retrieval.

Related Skills

Explore More Education & Research Skills→
Srajangpt1Srajangpt1

read-later

Save the current page as a bookmark and offline PDF.

Community
Intermediate
davpatel605-beepdavpatel605-beep

llm-wiki

Build and maintain a persistent markdown knowledge base with YAML frontmatter and cross-linking.

Community
Advanced
agentteamlandagentteamland

wiki

Initialize and maintain cross-referenced wiki pages from ingested sources.

Official
Advanced
danilonovaisvdanilonovaisv
1

AgentDB Vector Search

Enable fast semantic search over document collections using AgentDB.

Community
Advanced