Karpathy LLM Wiki

Trust Score 91/100

Create and maintain a Karpathy-style LLM-backed personal knowledge base: ingest raw sources, compile wiki articles, query and lint the knowledge base.

triggers:llm wikiadd to wikiingest sourcewhat do I know about

GitHub SKILL.md

What it does

Provides a workflow and tooling to build a persistent personal wiki using LLMs: collect immutable raw sources, compile them into curated wiki articles, keep an index and operation log, and run linting and query flows. The skill formalizes ingest, compile, query, archive, and lint steps.

When to use it

Use when you want an LLM-powered knowledge management system for research notes, documentation, or a personal knowledge base. Triggers include 'add to wiki', 'ingest source', 'what do I know about X', and maintenance lint jobs.

What's included

Scripts: none bundled in the skill repo, but templates and file structure conventions are provided for raw/ and wiki/ folders.
References: article/index/raw templates and detailed workflow guidance in SKILL.md and references.
Instructions: step-by-step rules for initializing the wiki, ingesting sources, compiling articles, cascading updates, and deterministic/heuristic lint checks.

Compatible agents

Designed for agents that can read/write project files and run LLM operations (Claude Code, Codex, Cursor-style automation agents).

Audit Summary

A purely declarative skill for building and maintaining a Karpathy-style LLM wiki knowledge base. No scripts included — all instructions guide the agent on file/directory management patterns (raw/ sources, wiki/ compiled articles). Well-structured with clear Ingest, Query, and Lint workflows, cascade update logic, and deterministic vs heuristic lint categories. No security concerns whatsoever.

Watch Out

Requires agent to have file read/write capabilities
Relies on agent discipline to follow strict rules like immutability of raw/ files

Notes

No executable code to audit. Pure instruction-based skill. High security score reflects zero attack surface. Code quality reflects well-written but instruction-only content — no scripts to evaluate for runtime behavior. Architecture is clean with good separation of concerns across three modes. Usefulness is moderate — niche but genuine audience of knowledge workers who want LLM-managed wikis.

Information

Repository: karpathy-llm-wiki
Stars: 998

Trust Score

Overall91

Security100

Code Quality82

Architecture80

Usefulness65

Related Skills

Yielding Bear — Unified LLM Routing

A unified LLM routing API that automatically routes requests to the cheapest capable model across providers to reduce inference cost by ~60-80%.

Markdrop

Convert PDFs to structured Markdown or interactive HTML and generate AI-powered descriptions for images and tables using multiple LLM providers.

Chinese Software Copyright Application

Generate complete Chinese software copyright application materials (application form, code documentation, user manual, design docs) and export them to PDF from

mog — Microsoft Ops Gadget

Command-line toolkit for Microsoft 365 (Mail, Calendar, OneDrive, Contacts, Tasks, Word, PowerPoint, Excel, OneNote) with slug and multi-account support.

CTF Write-up Generator

Generate a concise, reproducible submission-style CTF writeup with a one-path solution script, metadata, and a short checklist for fast verification.

Verifying Markdown Formatting

Checks Markdown files for structural and stylistic formatting issues (headings, lists, code blocks, links) and applies deterministic fixes.

Code Spec Backfill

Identify missing function-level contracts and backfill docstrings and type annotations; report ambiguous gaps with misuse scenarios. Incremental, state-driven w

Azure External Attack Surface Management

Provides expert guidance for Azure External Attack Surface Management (EASM): quotas, configuration, integrations, and exporting inventory to analytics platform

Back to Skills