From llm-wiki
This skill should be used when the user asks to "create a wiki", "set up a wiki", "initialize a wiki", "start a knowledge base", "scaffold a wiki", or "set up an llm-wiki for this project". Guides Claude through generating a domain-specific wiki schema and directory structure at wiki/ in the project root.
How this skill is triggered — by the user, by Claude, or both
Slash command
/llm-wiki:create-wiki [topic or domain] (optional)[topic or domain] (optional)This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Scaffold a new LLM Wiki in the current project. The wiki is a persistent, compounding knowledge base maintained by Claude — based on Karpathy's LLM Wiki pattern. Claude writes and maintains it; the user curates sources and asks questions.
Scaffold a new LLM Wiki in the current project. The wiki is a persistent, compounding knowledge base maintained by Claude — based on Karpathy's LLM Wiki pattern. Claude writes and maintains it; the user curates sources and asks questions.
The wiki lives at wiki/ in the project root with this layout:
wiki/
├── CLAUDE.md ← schema: conventions the LLM follows (generated per-domain)
├── index.md ← content catalog, updated on every ingest
├── log.md ← append-only event log
├── raw/ ← immutable source documents (user drops files here)
└── pages/ ← LLM-maintained wiki pages
raw/ is never modified by Claude. pages/ and the index/log are entirely Claude's responsibility.
Before generating anything, ask the user the following (all at once):
Wait for answers before proceeding.
Based on the user's answers, generate a domain-specific wiki/CLAUDE.md. This file is the LLM's constitution — it defines conventions for the wiki. Tailor every section to the domain.
The schema must include:
Wiki purpose — 2-3 sentences on what this wiki tracks and why.
Directory structure — confirm the standard layout above, noting any domain-specific subdirs under pages/ if useful (e.g. pages/papers/, pages/people/, pages/concepts/).
Page types — list the recurring page types for this domain. For each, specify:
person-firstname-lastname.md, paper-shortname-year.md)Index conventions — how index.md is organized (by category, by date, alphabetically). Specify the one-line summary format for each entry.
Log conventions — the prefix format for log entries. Standard format:
## [YYYY-MM-DD] <operation> | <title>
Operations: ingest, query, lint, create.
Cross-reference style — how pages link to each other (standard markdown [Page Title](pages/filename.md)).
Contradiction handling — when new sources conflict with existing pages, note the conflict inline with a > ⚠️ Conflict: blockquote and flag it in the lint report.
Source citation style — how to cite sources in page content (e.g. ([Source Title](../raw/filename.md))).
mkdir -p wiki/raw wiki/pages
Generate an empty index appropriate for the domain:
# Wiki Index
> Auto-maintained by Claude. Updated on every ingest.
## [Category 1]
| Page | Summary | Sources |
|------|---------|---------|
| *(empty)* | | |
## [Category 2]
...
---
*Last updated: [date]*
Replace categories with domain-appropriate ones from the schema.
# Wiki Log
> Append-only. Each entry starts with `## [YYYY-MM-DD] operation | title` for easy grepping.
## [today's date] create | Wiki initialized
Wiki created for: [user's stated purpose].
Schema generated at `wiki/CLAUDE.md`.
After creating all files, tell the user:
wiki/wiki/raw/ to ingest themwiki/raw/. They are immutable sources of truth.wiki/log.md — never overwrite existing entries.wiki/CLAUDE.md as the single source of convention truth. If the user wants to change a convention, update it there first.wiki/ already exists in the project, stop and tell the user — do not overwrite an existing wiki.npx claudepluginhub christianasc5/christianasc5-marketplace --plugin llm-wikiFetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
Applies a firm's KYC/AML rules grid to parsed onboarding records: assigns risk rating, checks required documents, outputs rule outcomes with citations, and routes for escalation.
Generates daily or weekly digests of activity from connected sources (chat, email, docs, tasks, CRM), highlighting action items, decisions, mentions, and project updates.