Skip to content

Karpathy Guidelines - Hermes Adaptation.md

Latest commit

 

History

History
69 lines (48 loc) · 2.46 KB

Karpathy Guidelines - Hermes Adaptation.md

File metadata and controls

69 lines (48 loc) · 2.46 KB

Andrej Karpathy Skills — Hermes Adaptation

Source: https://github.com/forrestchang/andrej-karpathy-skills Stars: 12,227 Date: 2026-04-11 Status: Installed as Hermes skill

What It Is

A set of 4 behavioral guidelines for LLM coding agents, derived from Andrej Karpathy's observations on common AI coding mistakes:

  1. Think Before Coding — Surface assumptions, ask for clarification
  2. Simplicity First — Minimum code, no speculative features
  3. Surgical Changes — Only touch what you must, match existing style
  4. Goal-Driven Execution — Define success criteria, verify with tests

The Problems It Solves

From Karpathy's original post:

"The models make wrong assumptions on your behalf and just run along with them without checking. They don't manage their confusion, don't seek clarifications, don't surface inconsistencies, don't present tradeoffs, don't push back when they should."

"They really like to overcomplicate code and APIs, bloat abstractions, don't clean up dead code... implement a bloated construction over 1000 lines when 100 would do."

"They still sometimes change/remove comments and code they don't sufficiently understand as side effects, even if orthogonal to the task."

What Was Adapted

Original (Claude Code)

  • Single CLAUDE.md file
  • Claude Code plugin format
  • .claude-plugin/ directory
  • No Hermes integration

Adapted (Hermes)

  • Full Hermes skill format (SKILL.md + YAML frontmatter)
  • Added Hermes-specific integration section
  • Added wikilinks to related skills (TDD, debugging, planning, review)
  • Added trigger conditions ("When to Load This Skill")
  • Added anti-patterns quick reference table
  • Preserved full EXAMPLES.md as reference file
  • Categorized under software-development

Installation Location

~/.hermes/skills/software-development/karpathy-guidelines/
├── SKILL.md
└── references/
    └── EXAMPLES.md

Related Skills in Hermes

  • [[test-driven-development]] — Verification loop for Principle 4
  • [[systematic-debugging]] — Debug first for Principle 1
  • [[writing-plans]] — Multi-step planning for Principle 4
  • [[requesting-code-review]] — All principles during review

Key Insight

"LLMs are exceptionally good at looping until they meet specific goals... Don't tell it what to do, give it success criteria and watch it go."

Good code solves today's problem simply, not tomorrow's problem prematurely.

Adapted by Hermes Agent — 2026-04-11