andrej-karpathy

Agente que simula Andrej Karpathy — ex-Director of AI da Tesla, co-fundador da OpenAI, fundador da Eureka Labs, e o maior educador de deep learning do mundo.

Quality

18%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

—

The risk profile of this skill

Fix and improve this skill with Tessl

tessl review fix ./plugins/antigravity-awesome-skills-claude/skills/andrej-karpathy/SKILL.md

Quality

Content

14%Scale 1-3

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

This skill is an extensive encyclopedia about Andrej Karpathy rather than an efficient, actionable skill for Claude. At 1000+ lines, it massively exceeds what's needed — Claude already knows who Karpathy is and can simulate his style with far less context. The content lacks a clear workflow for how to use the information and would benefit enormously from being condensed to ~100-150 lines focusing on distinctive style markers, key frameworks, and response structure.

Suggestions

Reduce content by 80-90%: Remove biographical timeline, detailed Tesla/OpenAI history, full quote collections, and vocabulary lists. Focus on the 10-15 most distinctive style markers and frameworks Claude needs to simulate Karpathy.

Add a clear response workflow: Define a step-by-step process for how Claude should construct Karpathy-style responses (e.g., 1. Lead with intuition, 2. Give precise technical definition, 3. Show real code, 4. Acknowledge limitations, 5. Suggest next steps).

Split into multiple files: Keep SKILL.md as a concise overview (~100 lines) with style guide and response structure, then reference separate files like QUOTES.md, TESLA.md, PROJECTS.md for detailed content.

Remove content Claude already knows: Claude is already familiar with backpropagation, transformers, BPE tokenization, and Karpathy's public work. Focus only on non-obvious style nuances and specific behavioral patterns that differentiate this persona from generic responses.

Dimension	Reasoning	Score
Conciseness	Extremely verbose at ~1000+ lines. Contains extensive biographical details, timelines, debate arguments (cameras vs LiDAR), full quote collections, vocabulary lists, humor descriptions, and personality traits that Claude doesn't need to simulate the persona. Most of this is encyclopedic content that could be drastically condensed.	1 / 3
Actionability	Contains some concrete code examples (micrograd Value class, tokenization with tiktoken, backprop example) but most content is descriptive/biographical rather than instructional. The skill tells Claude about Karpathy rather than giving clear, executable instructions on how to simulate him. Code snippets are often incomplete (truncated micrograd, pseudocode nanoGPT).	2 / 3
Workflow Clarity	There is no clear workflow or sequenced process for how Claude should respond when this skill is activated. The content is organized as an encyclopedia about Karpathy rather than a structured guide for simulation. No validation steps, no decision tree for different query types, no clear sequence of how to construct a Karpathy-style response.	1 / 3
Progressive Disclosure	Monolithic wall of text with everything inline in a single massive file. Content that could easily be split into separate reference files (biography, Tesla details, quotes collection, project descriptions) is all crammed into one document. No references to external files for detailed content.	1 / 3
	Total	5 / 12 Passed

Description

22%Scale 1-3

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

This description focuses almost entirely on establishing the persona's credentials rather than describing what the skill actually does or when it should be used. It lacks concrete actions (e.g., explains deep learning concepts, teaches neural network architectures) and has no 'Use when...' clause, making it difficult for Claude to know when to select this skill over others.

Suggestions

Add concrete actions the skill performs, e.g., 'Explains deep learning concepts, walks through neural network architectures, provides intuitive explanations of AI research papers in Karpathy's teaching style.'

Add an explicit 'Use when...' clause, e.g., 'Use when the user asks for deep learning explanations, neural network tutorials, AI education in a Karpathy-like style, or mentions Karpathy by name.'

Include natural trigger terms users would say, such as 'teach me deep learning', 'explain transformers', 'neural networks tutorial', 'AI concepts', 'machine learning education'.

Dimension	Reasoning	Score
Specificity	The description names a persona (Andrej Karpathy) and lists his credentials, but does not describe any concrete actions the skill performs. There are no verbs indicating what the skill actually does beyond 'simula' (simulates).	1 / 3
Completeness	The description partially answers 'what' (simulates Karpathy) but very weakly, and completely lacks a 'when' clause or any explicit trigger guidance for when Claude should select this skill.	1 / 3
Trigger Term Quality	Contains some relevant keywords like 'Andrej Karpathy', 'deep learning', 'Tesla', 'OpenAI', 'Eureka Labs' that users might mention, but lacks natural trigger terms for when someone would want this skill (e.g., 'explain neural networks', 'teach machine learning', 'AI education').	2 / 3
Distinctiveness Conflict Risk	The specific persona (Andrej Karpathy) provides some distinctiveness, but without clear actions or triggers, it could overlap with any AI education, deep learning, or persona simulation skill.	2 / 3
	Total	6 / 12 Passed

Validation

81%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 9 / 11 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (1170 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	9 / 11 Passed

Repository: popey/claude-code-skills
Path: plugins/antigravity-awesome-skills-claude/skills/andrej-karpathy/SKILL.md
Commit: a5a6601

Reviewed: 5 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.