yshms
UserClaude Code skills that make Claude follow Andrej Karpathy's published AI-usage workflow — distilled from his X threads, blog, talks, and interviews
Categories
Indexed Skills (4)
karpathy-coding-loop
Use before writing, modifying, or refactoring any code, and at each temptation point inside it: starting an implementation, proceeding on an unstated assumption, choosing between two readings of a request, adding an abstraction, config option, or error handler nobody asked for, editing a line or comment the task did not require, or reporting work done.
karpathy-context-engineering
Use before pulling docs, a repo, or error output into context or calling an unfamiliar library — and when the context window is the problem: a hallucinated or deprecated API, a long session drifting off-topic or degrading in quality, the same mistake recurring across sessions, or an agent regressing to conventions this repo does not use.
karpathy-task-routing
Use when deciding how to run a task rather than doing it: how much autonomy to give an agent, whether to hand-write instead, whether to parallelize agents, which model or tool tier fits, or what to do when a delegated task keeps failing. Also when the user asks how to use LLMs efficiently, how Karpathy works with AI, or how to learn or study a topic — including bare 'help me learn X' requests.
karpathy-verification
Use when judging whether output is correct or finished: reviewing an AI-written diff before accepting, merging, or shipping; deciding whether a working demo means done; checking a model's numbers, counts, or self-reported success; or setting up an LLM as a judge, grader, or council.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.