po4yka/llm-wiki-skills / skills/llm-wiki-orient
llm-wiki-orient
VerifiedExplain the LLM-Wiki pattern, its raw/wiki/schema architecture, how it differs from RAG, GraphRAG and agent memory, and what ready-made solution families exist. Use when the user asks what LLM-Wiki is, whether it is worth adopting, what OpenWiki/wiki memory means, or how the ecosystem fits together.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 4.6 KB |
SKILL.md
View on GitHub---
name: llm-wiki-orient
description: Explain the LLM-Wiki pattern, its raw/wiki/schema architecture, how it differs from RAG, GraphRAG and agent memory, and what ready-made solution families exist. Use when the user asks what LLM-Wiki is, whether it is worth adopting, what OpenWiki/wiki memory means, or how the ecosystem fits together.
license: MIT
compatibility: Designed for Agent Skills-compatible coding agents. For current ecosystem claims, browse and cite fresh sources.
metadata:
author: po4yka
version: "1.0.0"
install_scope: self-contained
---
# LLM-Wiki Orient
## Goal
Give the user a clear, non-hype explanation of LLM-Wiki and route them to the next useful skill.
## When to use
Use when the user asks:
- "What is LLM-Wiki?"
- "How is this different from RAG?"
- "What tools exist?"
- "Is this useful for my notes/team/repo?"
- "Explain OpenWiki/wiki memory/compiled knowledge."
For objections such as "why do I need this?", "what proof exists?", "what benefits will I get?", "how do I keep it alive?", or "do I need Obsidian?", route to or invoke `llm-wiki-faq`.
## Inputs
- The user's question about LLM-Wiki (what it is, how it differs from RAG/GraphRAG, whether to adopt it).
- Optional context frame: personal second brain, team/company knowledge, repository documentation, research corpus, or product build.
- Optional local-first requirement and sensitivity/privacy level.
- Optional target agent: Claude Code, Codex, Cursor, OpenCode or other.
- Whether the user already has documents to migrate or is starting empty.
## Procedure
### 1. Identify the user's frame
Infer or ask for the minimum needed context:
- personal second brain, team/company knowledge, repository documentation, research corpus, or product build;
- local-first requirement;
- target agent: Claude Code, Codex, Cursor, OpenCode or other;
- existing documents or empty start;
- sensitivity/privacy level.
If the user only wants a general explanation, do not interrogate them.
### 2. Explain the pattern
Use this mental model:
```text
raw/ immutable source material
wiki/ LLM-compiled, human-readable Markdown knowledge layer
schema/ AGENTS.md, CLAUDE.md, skills, schemas and conventions
```
Emphasize:
- LLM-Wiki is a compiled knowledge layer, not just a note app.
- Raw sources stay preserved.
- The wiki is reviewable Markdown.
- Skills define procedures over the wiki.
- `index.md` and `log.md` make the wiki navigable and auditable.
- Evidence and adoption questions should be answered with direct/adjacent/local evidence levels, not hype.
### 3. Compare to neighboring patterns
Explain briefly:
| Pattern | Difference |
| --- | --- |
| RAG | Retrieves chunks at query time. LLM-Wiki pre-compiles durable pages and still may use search. |
| GraphRAG | Builds graph-aware retrieval structures. LLM-Wiki makes human-readable Markdown the primary artifact. |
| Agent memory | Often black-box or agent-owned. LLM-Wiki is inspectable, git-versioned and user-owned. |
| CLAUDE.md / AGENTS.md | Schema/instruction layer, not the whole knowledge base. |
| Skills | Operators that perform tasks over the wiki. |
### 4. Present solution families
Use current browsing for volatile facts. The stable categories are:
- repo-documentation agents such as OpenWiki-style tools;
- Obsidian/local-first Markdown workflows;
- hybrid search add-ons such as qmd-style local retrieval;
- graph/RAG systems when machine retrieval is primary;
- custom CLI/plugin/product builds;
- team PR-based wiki maintenance workflows.
### 5. State the trust boundary
Say clearly:
> Automate bookkeeping, not belief.
The agent can maintain links, indexes, summaries, metadata, triage reports and lint reports. Human review should own final synthesis, ambiguous claims and promotion to verified knowledge.
### 6. Route to next skill
Recommend one next skill:
- answer adoption objections or evidence questions: `llm-wiki-faq`;
- choose a solution: `llm-wiki-choose`;
- install/configure: `llm-wiki-setup`;
- migrate existing docs: `llm-wiki-refactor`;
- build custom: `llm-wiki-design`;
- get current news: `llm-wiki-news-radar`;
- operate an existing vault: `wiki-triage`, `wiki-ingest`, `wiki-query`, `wiki-lint`.
## Output
Use this shape:
```markdown
## What LLM-Wiki is
## What it is not
## Main solution families
## Evidence level
## Fit for your case
## Recommended next step
```
## Safety gates
- Do not present generated summaries as verified truth.
- Do not recommend a specific external tool using stale maturity facts.
- Browse for current project status, releases, pricing, compatibility or maintenance claims.
- Do not claim LLM-Wiki is proven universally better than RAG; route evidence questions through `llm-wiki-faq`.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |