v0.4.0 · Open Source · MIT

Lint your prompts
before you send them

Static analysis for LLM prompts. Analyze a prompt in the browser, use the CLI in CI, or embed the core library directly in your tooling with the same deterministic engine.

View on GitHubQuick Start
$npx @promptscore/cli analyze prompt.txt
Live analyzer

The browser analyzer below uses the current deterministic engine directly in the client, so the score and findings stay aligned with the CLI and the core library.

v0.4.0
Runs locally in the browser

This analyzer uses the current deterministic engine from @promptscore/core. No API calls are made.

PS

Run the analyzer to see a real prompt report here.

How it works

Three steps. Zero API calls.

Choose your surface

Analyze prompts directly in the browser on this page, run the CLI on a prompt file, or import @promptscore/core in your code.

Run 12 deterministic rules

PromptScore checks length, structure, output format, examples, constraints, vague language, and more. All offline.

Get actionable feedback

Each finding includes a concrete fix suggestion, and supported profiles can attach Claude- or GPT-specific references.

Features

Built for prompt engineers
who ship in production

12 Deterministic Rules

Check length, structure, examples, output format, constraints, vague language, and more. No LLM calls, no API keys.

Model Profiles

YAML profiles for Claude, GPT, and a universal baseline. Rules adjust severity and suggestions per model.

Browser + CLI + Library

Use the browser analyzer for quick checks, import @promptscore/core in your code, or run promptscore from the terminal.

CI/CD Ready

Use the CLI in CI today. It can scan prompt directories and exits non-zero when findings meet your configured failure threshold.

Actionable Warnings

Each failing rule includes a concrete fix suggestion, and supported profiles can link to official model docs.

Privacy First

Your prompts never leave your machine. No telemetry. No network calls. Fully offline.

Rules

What gets checked

min-lengthMinimum lengthwarning
max-lengthMaximum lengthinfo
no-output-formatOutput format specifiedwarning
no-examplesExamples providedwarning
no-roleRole or persona assignedinfo
no-contextBackground context providedinfo
ambiguous-negationAmbiguous negative instructionsinfo
no-constraintsConstraints definedinfo
all-caps-abuseAll-caps abuseinfo
vague-instructionVague instructionswarning
missing-taskClear task or instructionerror
no-structured-formatStructured formattingwarning
Get Started

Up and running in seconds

# Analyze a prompt file — no install needed
npx @promptscore/cli analyze prompt.txt

# Or pipe from stdin
echo "You are a helpful assistant" | npx @promptscore/cli analyze

# Target a specific model profile
npx @promptscore/cli analyze prompt.txt --model claude