karpathy-wiki
Skill109 starsChanged 7 months agoMITView on GitHub
Use when building or maintaining a personal knowledge base with LLM assistance. Triggers on ingesting sources, compiling wiki articles, querying knowledge, running health checks, deep research, or initializing a new wiki vault. Based on Karpathy's LLM Wiki pattern with Obsidian integration.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- baklib-tools/skillstheir own version, first seen May 19, 2026
- skytiger6724/qwen-skillstheir own version, first seen Apr 9, 2026
- oozoofrog/codex-skills-projecttheir own version, first seen Mar 28, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- baklib-tools/skills★ 2same name
- skytiger6724/qwen-skills★ 2same name
- oozoofrog/codex-skills-project★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
Files
Karpathy Wiki
LLM-driven personal knowledge base that compounds over time. Raw sources in, structured wiki out.
You never write the wiki. The LLM writes everything. You just steer, and every answer compounds.
When to Use
- User wants to ingest documents, URLs, or text into a knowledge base
- User wants to query, compile, or maintain wiki articles
- User needs deep research to fill knowledge gaps
- User wants to initialize a new wiki vault
When NOT to Use
- One-off Q&A without persistent knowledge storage
- Note-taking without cross-referencing or structure
- Non-text content management (images, videos alone)
Core Principles
- raw/ is immutable — sources are never edited; re-ingest as new version
- purpose.md anchors direction — read on every operation, align output with wiki goals
- Classify before extract — determine entity type during Ingest, choose chapter structure during Compile
- Analyze before generate — two-step Compile: structured analysis → article generation
- Backlink audit is mandatory — bidirectional links are the knowledge graph
- Query answers must be archived — good answers are compounding fuel
- Token budget first — always read index before full articles (L0→L1→L2→L3)
- Human verifies, LLM maintains — LLM writes, user steers
Pipeline
| Stage | Input | Output |
| Ingest | URLs, files, text | raw/ (immutable) + hash cache |
| Compile | raw/ files | wiki/<domain>/ articles by entity type |
| Query | Questions | Cited answers → outputs/queries/ |
| Lint | Full wiki | Health report → outputs/reports/ |
Each stage appends log.md. Lint and Query can trigger Deep Research to fill gaps.
Token Budget
| Level | ~Tokens | When | What |
| L0 | ~200 | Every session | SKILL.md frontmatter |
| L1 | ~1-2K | Session start | purpose.md + Concept Index + Dashboard |
| L2 | ~2-5K | Search | Result summaries |
| L3 | 5-20K | Deep read | Full articles or raw sources |
Entity Types
7 types, each with different required sections (see references/entity-types.md):
concept (default) · person · tool · event · comparison · pattern · overview
Commands
| Command | Action |
init [path] [--template <t>] | Initialize vault (templates: general/research/reading/project) |
ingest <source> | Collect + classify + archive (no wiki generation) |
compile [topic] | Two-step: analyze → generate wiki article |
query <question> | Answer from wiki, archive result |
promote | Elevate quality query answer to wiki article |
research <topic> | Gap → web search → ingest → compile loop |
delete <source> | Safe delete with cascade cleanup |
lint | Health check + graph analysis |
Conventions
- Obsidian CLI first for page operations (see
references/obsidian-cli-cheatsheet.md) - Structured diff for updates: Current/Proposed/Reason/Source, user confirms each
- Sources traceability: frontmatter
sources[]links every wiki page to raw/ files - Incremental cache: SHA256 hash in
.cache/skips unchanged sources
References
references/operations.md— Full procedures for all 9 operationsreferences/frontmatter-schemas.md·compilation-guide.md·entity-types.md·obsidian-cli-cheatsheet.mdassets/— All templates (purpose, wiki article, raw, query, indexes, log)scripts/—init-wiki.sh(with--template),parse-media.py
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.