wiki

Ingest sources, query, and lint a markdown-based LLM Wiki knowledge base.

20|2|Updated May 16, 2026
One-click install
npx skills add https://github.com/reddb-io/red-skills --skill wiki-reddb-io
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: wiki
Source: https://github.com/reddb-io/red-skills/tree/main/plugins/memory/skills/core/wiki
Command: npx skills add https://github.com/reddb-io/red-skills --skill wiki-reddb-io

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Knowledge scattered across URLs, PDFs, and notes gets lost or re-derived on every query. This Skill maintains an incremental markdown wiki where every source is absorbed into cross-linked pages, contradictions are surfaced, and answers cite their sources. ## Core Features & Use Cases - Ingest: Fetch a URL or convert a local PDF/markdown file into .red/wiki/raw/, then write source, entity, and concept pages with cross-references and an updated index. - Query: Answer free-form questions by mapping the index, drilling into pages, and synthesizing cited answers as prose, tables, or Mermaid diagrams, optionally filed back as synthesis pages. - Lint: Run health checks for contradictions, stale pages, orphans, stubs, missing concepts, unresolved gaps, and C4 diagram staleness, producing a grouped markdown report. - Use Case: After reading a long architecture article, ask the agent to ingest it into the wiki; it saves the raw source, updates entity pages, flags a contradiction with an older page, and logs the operation. ## Quick Start Ask the agent to ingest a URL into the wiki, for example: ingest https://example.com/architecture-overview into the wiki and highlight the main takeaways.

Frequently Asked Questions about wiki

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

FAQPage Schema
How do I ingest a URL or PDF into a markdown knowledge base?▼

Point the skill at a URL to fetch and save it under .red/wiki/raw/ with a YAML header, or provide a local PDF which is converted with pdftotext and normalized to markdown. The agent then writes source, entity, and concept pages and updates the index.

What is the LLM Wiki pattern for knowledge management?▼

The LLM Wiki pattern, from Andrej Karpathy's gist, replaces RAG with an incrementally maintained markdown wiki. Each new source is absorbed into existing pages with cross-references and syntheses, while the human curates and the LLM does the bookkeeping.

How do I check a markdown wiki for contradictions and orphan pages?▼

Run the lint operation, which scans pages for conflicting claims, stale content older than 90 days, orphans with no inbound links, stubs under 10 lines, missing concepts, and unresolved open questions. It reports findings without auto-fixing.

Does the wiki skill work without running wiki-init first?▼

No. The skill requires .red/wiki/ and .red/agents/wiki.md to exist and .red/wiki/ to be listed in .gitignore. If any precondition fails, it stops and instructs you to run /wiki-init before any operation.

When should I add a C4 architecture diagram to the wiki?▼

Add a C4 model at .red/wiki/C4.md when the system has roughly three or more services or non-trivial cross-system integrations. The wiki never auto-creates it; when a trigger fires and the file is absent, the skill proposes creation.