Install PromptScore and run your first prompt check

Install or run once

PromptScore ships as a public CLI package. You can install it globally or invoke it with `npx`.

npm install -g @promptscore/cli
npx @promptscore/cli analyze prompt.txt

Analyze a file, directory, glob, inline prompt, or stdin

The `analyze` command accepts prompt files, directories of prompt-like text files, glob patterns, inline strings, or piped stdin.

promptscore analyze prompt.txt
promptscore analyze prompts/
promptscore analyze "prompts/**/*.{txt,md}"
promptscore analyze --inline "You are a helpful assistant."
cat prompt.txt | promptscore analyze

Choose a profile

The current public release ships with `_base`, `claude`, and `gpt`. Profiles adjust weight, severity, suggestions, and references while using the same deterministic core engine.

promptscore analyze prompt.txt --model claude
promptscore profiles

Save project defaults

If your team wants shared defaults, add a `promptscore.config.yaml` file to the repo and let the CLI discover it automatically.

model: claude
format: text
fail_on_severity: warning

Use it in code

In Node or server-side code, import `analyze` from `@promptscore/core`. The API returns a Promise<ScoreReport>.

import { analyze, format } from '@promptscore/core';

const report = await analyze('You are a helpful assistant. Summarize the article.', {
  model: 'claude',
});

console.log(report.overall);
console.log(format(report, 'text'));

Where to go next

• Read CLI Guide for command details, formats, and exit codes.
• Read Config to set project-wide defaults and CI policy.
• Read Browser Analyzer for the browser-safe entry point and client usage.
• Read Rules Reference to understand what the score is actually measuring.