Individual skill
karpathy-skills-extended
Always-on check for any coding agent, including Codex, Claude, Cursor, Antigravity, Gemini, and Grok. Before code, writing, a plan, or an answer, find knowledge gaps that would change the outcome and that you cannot resolve from the repo, tools, or the conversation. Ask only those, in one short round, then wait. If there are none, do the work and stay quiet about the check. Includes Karpathy coding behavior: think before coding, simplicity first, surgical changes, and a check you can run. Use on every prompt, when the user runs /karpathy-skills-extended, and whenever you are about to assume scope, audience, behavior, names, schema, or success criteria.
skills/karpathy-skills-extended/SKILL.md
Source commit 8fe0d8dfbcafebcfd29ea1b94cd120964883843d
What was checked
- Matches the observed repository HEAD
- Yes
- Scanner rules are current
- No
- Repository scan complete
- Yes
- Skill scan complete
- Yes
- Observed behavior
- Not checked yet
- Last checked
- Nov 11, 2026
These are observations at the recorded commit, not a guarantee of safety. Review the source before running it.
Install this skill
Choose an agent in your Stack or copy a command below. Commands fetch the pinned commit and copy only this skill directory. Existing destinations are refused. Copying a command does not install anything.
An install command is unavailable: current rules, complete repository and skill scans, an eligible behavior level and a permitted license are required.
Repository context
- Repository stars (not skill ratings)
- 25
- Repository last updated
- Oct 24, 2026
- License
- MIT
Observed agent integration: No specific integration evidence. Standard SKILL.md installation does not imply verified compatibility.
- agent-skills Repository topic
- coding-agents Repository topic
- codex Repository topic
- ai-agents Repository topic
- claude-code Repository topic
- cursor-rules Repository topic
- karpathy Repository topic
- prompt-engineering Repository topic
- engineering Derived category
Files and excerpt
File metadata and permitted, redacted excerpts come from the recorded scan. Source links are pinned to that commit.
Browse files
SKILL.mdcore · 5,435 bytesagents/openai.yamlasset · 281 bytes
Read excerpt
# Karpathy skills, extended
Apply this on every user message, including when nobody named this skill. The host may be Codex, Claude, Cursor, Antigravity, Gemini, or Grok. The check is the same in each. Do not announce the check.
The four coding rules below follow Andrej Karpathy's notes on agents that assume, overbuild, and edit code they were not asked to touch (https://x.com/karpathy/status/2015883857489522876), in the form published by https://github.com/multica-ai/andrej-karpathy-skills. The gap check is the part this skill adds. Matt Pocock's grill-me (https://github.com/mattpocock/skills) asks the user to settle every branch of a plan before work starts. That session fits a soft idea. On an ordinary prompt it runs long. This skill keeps one piece of it: if a decision belongs to the user, ask them. Ask only the gaps that pass the tests below, then do the task.
## Gap check
Run this before you edit, write, delete, send, or commit to a plan.
A gap meets all four tests:
1. The user did not state it, and no earlier message in this conversation settled it.
2. You cannot get it by reading the repo, docs, tools, or config.