isabela-valonni/prompt-evaluator

README preview

# Prompt Evaluator

> Paste one block into any AI's custom instructions. Every chat starts with quiet, specific feedback on how you prompt — and it gets sharper as it learns your patterns.

[![License: MIT](https://img.shields.io/badge/license-MIT-green.svg)](./LICENSE)
[![Works with](https://img.shields.io/badge/works%20with-Claude%20·%20ChatGPT%20·%20Gemini%20·%20any%20AI-58a6ff.svg)](#quick-start--any-ai-60-seconds)
[![Claude plugin](https://img.shields.io/badge/Claude%20Code%20%2F%20Cowork-plugin-8250df.svg)](#the-personalized-setup)

A small, persistent prompting coach. It scores how you ask, names one thing to sharpen, and remembers your recurring patterns so the feedback gets *more targeted* over time. No app, no service, no telemetry — it's ~20 lines of instruction text.

Inspired by [Andrej Karpathy's observations](https://x.com/karpathy/status/2015883857489522876) on simple, durable AI behaviors — applied to the prompting layer itself: the part most people never get feedback on.

<p align="center">
  <img src="assets/demo.svg" alt="Pixel-art terminal: the evaluator scores a first prompt 6/10, one thing that worked, one thing to sharpen, then answers normally" width="720">
</p>

---

## Quick start — any AI, 60 seconds

View full repository on GitHub →