karpathy-graph-resolve skillA
karpathy-graph-resolve is agent-read markdown (skill) from andersonlimahw/lemon-ai-hub: Stage 2 of the karpathy-graph pipeline. Collapse duplicate surface forms into canonical entities using context as evidence, reversibly and with a recorded rationale. Use after extraction when the graph contains aliases, abbreviations, or spelling variants of the same thing, when isolated-node count is high, or when the user asks to deduplicate or merge entities..
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
# Resolve
Extraction creates **surface forms**, not a clean graph. The same thing appears
as an abbreviation, an alias, a former name, a spelling variant, or a partial
name. Resolution decides which of those are the same entity.
This is a reasoning task, not a string-matching task.
## Why string similarity is not enough
It fails in both directions:
- **Misses:** "Edwin Aldrin" and "Buzz Aldrin" share zero character overlap and
are the same person.
- **False merges:** two different people named "J. Silva" have perfect overlap
and are not the same person.
A false merge is the more expensive error. Collapse two entities into one node
and every downstream traversal silently combines their relations, dates, and
sources. Nothing errors. The graph just becomes confidently wrong.
**When uncertain, do not merge.** An unmerged duplicate is visible and fixable.
A false merge is invisible.
## Procedure
### 1. Block by type
Only ever compare entities of the same `type`. An `Entity` never resolves to a
`Source`, a `Metric` never resolves to a `Claim`.
```bash
KG="${CLAUDE_PLUGIN_ROOT:-plugins/karpathy-graph}/scripts/kgraph.py"
…Read the whole file at its exact version.
How to install
mdr add andersonlimahw/lemon-ai-hub/karpathy-graph-resolve@v1.0.0mdr add andersonlimahw/lemon-ai-hub/karpathy-graph-resolve@sha256:a9c49b8eb821bfc3Pin 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_7ksxckfhdwq6gtyz)
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 (4207 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_7ksxckfhdwq6gtyz GET https://markdownregistry.com/api/v1/resolve?ref=andersonlimahw/lemon-ai-hub/karpathy-graph-resolve GET https://markdownregistry.com/api/v1/blob/a9c49b8eb821bfc37b5f56c0753c045111cc60405abb3c076cd8bd6b32fbe47b
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.