karpathy-guidelines

Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.

Install with Tessl CLI

npx tessl i github:forrestchang/andrej-karpathy-skills --skill karpathy-guidelines

What are skills?

Review — 83%

Does it follow best practices?

If you maintain this skill, you can automatically optimize it using the tessl CLI to improve its score:

npx tessl skill review --optimize ./path/to/skill

Learn more

Evaluation — 92%

Agent success when using this skill

Validation — 11 / 11 Passed

Validation for skill structure

SKILL.md

Review

Evals

Evaluation results

80%

Simplify the Domain Extraction Utility

Surgical code edits with orphan cleanup

Criteria

Without context

With context

Orphaned import removed

Non-orphaned import kept

100%

Dead function preserved

100%

Unchanged functions untouched

100%

Style matched

100%

Dead code mentioned

100%

No added annotations

100%

Without context: $0.1212 · 24s · 7 turns · 8 in / 1,267 out tokens

With context: $0.2514 · 51s · 14 turns · 14 in / 2,633 out tokens

100%

Log Error Extractor

Simplicity first, no over-engineering

Criteria

Without context

With context

No class abstraction

100%

No extra CLI flags

100%

No config file

100%

No speculative error handling

100%

No extra output files

100%

No unrequested features

100%

Concise implementation

100%

No logging framework

100%

No type annotations on simple code

100%

Without context: $0.1929 · 35s · 14 turns · 11 in / 2,178 out tokens

With context: $0.2887 · 58s · 17 turns · 17 in / 2,545 out tokens

98%

37%

Extend the User Export Script with Filtering

Goal-driven execution with assumption surfacing

Criteria

Without context

With context

Multiple interpretations listed

33%

100%

Chosen interpretation stated

91%

100%

Reasoning given for choice

83%

100%

Numbered implementation plan

33%

83%

Per-step verification

100%

Simpler path noted

90%

100%

No gold-plating

83%

100%

Existing functions unchanged

100%

Without context: $0.2625 · 1m · 14 turns · 14 in / 3,215 out tokens

With context: $0.3122 · 1m 12s · 15 turns · 16 in / 3,682 out tokens

Evaluated: about 7 hours ago
Agent: Claude Code

Table of Contents

Simplify the Domain Extraction Utility Log Error Extractor Extend the User Export Script with Filtering

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.