karpathy-graph-extract skillA
karpathy-graph-extract is agent-read markdown (skill) from andersonlimahw/lemon-ai-hub: Stage 1 of the karpathy-graph pipeline. Turn unstructured context — documents, source files, incident timelines, chat logs, research notes — into typed entities and subject-predicate-object relations that fit the graph ontology. Use when the user asks to extract entities, pull relations out of documents, map a codebase into a graph, or when building a knowledge graph and no nodes exist yet..
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# Extract
Replace the classical NLP pipeline with one schema-constrained pass per
document. There is no trained NER model here — **the schema is the training
data**, and the ontology is the contract.
## Input contract
One unit of context at a time. A "unit" is whatever fits comfortably in one
pass with room for reasoning:
| Context | Unit |
|---|---|
| Documents | one document, or one section of a long one |
| Codebase | one module or package, not one file |
| Incident | one timeline, one postmortem |
| Research | one paper, one thread |
Never batch unrelated documents into one extraction call. Provenance gets
blurred and you cannot tell which source supported which claim.
## Output contract
Emit exactly this shape. Nothing else — no prose, no markdown fence commentary.
```json
{
"nodes": [
{
"type": "Source",
"name": "incident-2026-03-11.md",
"uri": "./docs/incidents/incident-2026-03-11.md"
},
{
"type": "Entity",
"name": "Vendor X",
"description": "component supplier named in the incident report",
"aliases": ["VendorX Ltd", "Vendor-X"]
},
{
"type": "Claim",
"name": "Vendor X supplied Component Z",
…Read the whole file at its exact version.
How to install
mdr add andersonlimahw/lemon-ai-hub/karpathy-graph-extract@v1.0.0mdr add andersonlimahw/lemon-ai-hub/karpathy-graph-extract@sha256:d908f50623d4d0c3Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_c6p7sjfk5ks66ilt)
1 badge views in 30 days
Versions
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (4978 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
andersonlimahw/lemon-ai-hub · 14 stars · license none · pushed 2026-09-20 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_c6p7sjfk5ks66ilt GET https://markdownregistry.com/api/v1/resolve?ref=andersonlimahw/lemon-ai-hub/karpathy-graph-extract GET https://markdownregistry.com/api/v1/blob/d908f50623d4d0c3fb224ae84b35857adfe344461b25ce576bca636ad8b9c2ec
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.