LIVE
Mistral Makes Large 4 Available as Open Weights29/10/26 · Mistral AI|Gemini 4 Carbon tops the Artificial Analysis Intelligence Index v529/10/26 · Google|Mistral Large 4 now available on Hugging Face29/10/26 · Mistral AI|Google Introduces "Gemini 4 Carbon" as Next Step in Its Model Lineup28/10/26 · Google|DeepMind unveils Genie 4, a new generation of world model27/10/26 · Google DeepMind|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|New benchmark checks whether LLMs actually test their temporal medical claims29/10/26|An ODE-based sampling method speeds up discrete diffusion models29/10/26|OpenAI revises its Model Spec with five guiding values for AI behavior29/10/26 · OpenAI|Inside Proton's internal clash over its AI pivot29/10/26 · Proton|Mistral Makes Large 4 Available as Open Weights29/10/26 · Mistral AI|Gemini 4 Carbon tops the Artificial Analysis Intelligence Index v529/10/26 · Google|Mistral Large 4 now available on Hugging Face29/10/26 · Mistral AI|Google Introduces "Gemini 4 Carbon" as Next Step in Its Model Lineup28/10/26 · Google|DeepMind unveils Genie 4, a new generation of world model27/10/26 · Google DeepMind|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|New benchmark checks whether LLMs actually test their temporal medical claims29/10/26|An ODE-based sampling method speeds up discrete diffusion models29/10/26|OpenAI revises its Model Spec with five guiding values for AI behavior29/10/26 · OpenAI|Inside Proton's internal clash over its AI pivot29/10/26 · Proton|
SkillRisingofficial publisher

andrej-karpathy-skills

Skill : A reusable method: instructions, procedures, sometimes scripts or templates · requires any-agent

Portable skills inspired by Andrej Karpathy's practices for coding with an agent.

Short, dense, and readable in ten minutes: a good example of a well-written skill.

Official site GitHub · 220,693 ★
Overall score
7.6 /10
Security: one caveat, see the card
Utility
8.0
Reliability
9.0
Security
One caveat
Ease
5 min · Comfortable

Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.

What it does, concretely

Coding agents like Claude Code tend to make silent assumptions, overcomplicate code, and modify parts of a project unrelated to the task at hand. This skill gathers four principles into a single file, drawn directly from Andrej Karpathy's public observations about these pitfalls.

In practice, the skill is installed either as a Claude Code plugin or as a CLAUDE.md file in your project. It injects instructions that push the agent to state its assumptions explicitly, ask questions before coding, write the minimum necessary code, and touch only the lines directly related to the request.

It is aimed at developers who use Claude Code, Cursor, or Codex regularly and have already experienced overly broad diffs, unsolicited refactoring, or oversized implementations. Installation takes under two minutes and the file can be customised to fit the rules of an existing project.

This skill is not useful for trivial tasks such as fixing a typo: the authors themselves note it is designed to reduce costly mistakes on non-trivial work, not to slow down simple changes.

What it offers

  • ·"Think Before Coding" principle: forces the agent to state its assumptions and ask for clarification before starting.
  • ·"Simplicity First" principle: forbids speculative abstractions and unrequested features, targeting the minimum lines that solve the problem.
  • ·"Surgical Changes" principle: limits modifications to lines directly related to the request and prohibits unsolicited adjacent refactoring.
  • ·"Goal-Driven Execution" principle: transforms imperative instructions into verifiable success criteria, with tests written before the code.
  • ·Installation as a Claude Code plugin for availability across all projects, or as a CLAUDE.md file for per-project use.
  • ·Cursor compatibility via a dedicated project rule included in the repository.
  • ·Customisable: the file is designed to be merged with instructions specific to an existing project.

In practice

Price
Free
Commercial use
Commercial use: yes
Available on
claude-code, codex, cursor
Host
any-agent
Licence
MIT
FR interface
No

Security and data

Rating
Security: one caveat, see the card
Your data
Data stays local
Maintenance
Active

Links and repository checked on 18 October 2026 · added on 1 May 2026

How to install andrej-karpathy-skills

1. Open Claude Code in your development environment. 2. Add the marketplace with the command: /plugin marketplace add forrestchang/andrej-karpathy-skills 3. Install the plugin with: /plugin install andrej-karpathy-skills@karpathy-skills 4. For per-project use only, download the file instead with: curl -o CLAUDE.md https://raw.githubusercontent.com/forrestchang/andrej-karpathy-skills/main/CLAUDE.md 5. If you already have a CLAUDE.md, use the append command to avoid overwriting your existing rules. 6. Caution: this file is executed as a system instruction by the agent; review its contents before adding it to a shared or sensitive project.
/plugin marketplace add forrestchang/andrej-karpathy-skills && /plugin install andrej-karpathy-skills@karpathy-skills

In the same category

Whole category Frame (methods, guardrails) →