multica-ai/andrej-karpathy-skills vs steipete/agent-rules
multica-ai/andrej-karpathy-skills scores higher overall: 61/100 (Experimental-tier) against 54/100 (Deprecated-tier). multica-ai/andrej-karpathy-skills leads on reliability, skill leverage, maintenance; steipete/agent-rules leads on documentation, composability. Scored 2026-07-07, methodology v1.0.
Dimension by dimension
| Dimension | karpathy-skills | agent-rules |
|---|---|---|
| Reliability | 60 | 55 |
| Skill Leverage | 70 | 62 |
| Documentation | 68 | 74 |
| Maintenance | 50 | 15 |
| Safety / Governance | 62 | 58 |
| Evaluation Readiness | 35 | 35 |
| Composability | 55 | 66 |
| Adoption (capped) | 90 | 62 |
| Overall | 61 · Experimental | 54 · Deprecated |
The entries
A single CLAUDE.md distilling Andrej Karpathy's public observations on LLM coding pitfalls into instruction-file rules. An instruction set,…
A well-regarded collection of rules and knowledge files for Claude Code and Cursor. Archived by its maintainer in 2026; indexed with a…
Frequently asked questions
multica-ai/andrej-karpathy-skills or steipete/agent-rules?
multica-ai/andrej-karpathy-skills scores higher overall (61 vs 54, methodology v1.0). But they are the same category, so the dimension table below is the real answer.
How is this comparison generated?
Both scorecards come from the same public rubric with evidence notes, scored by the same editorial process. This page presents them side by side; it adds no new judgments beyond the scores themselves.