Prompt Injection Scanner

Paste any prompt or untrusted input to detect jailbreaks, instruction overrides, data exfiltration, and other LLM attack patterns. Runs entirely in your browser — nothing is sent to a server.

100% client-side · privacy-preserving

Input to analyze0 chars

Try a sample

Results

Run a scan to see detected risks, a severity breakdown, and remediation guidance.

How it works

The scanner runs a transparent, rule-based engine against your input. Every finding maps to a named rule with a category, severity, and mitigation. It is a first line of defense — pair it with output validation, least-privilege tools, and human review for high-risk actions.

13+ detection rules

Instruction override, role manipulation, jailbreaks, prompt leaks, exfiltration, obfuscation, and delimiter injection.

Zero data sent

Everything executes locally in your browser via a portable, dependency-free TypeScript engine.

Built to embed

Ship it anywhere — the same engine is available as an npm package, a CLI, a GitHub Action, and a REST API.

Explore the full toolkit

Use it anywhere

The exact engine powering this page ships as an open-source npm package, a CLI, a GitHub Action, and a hosted REST API — so you can run the same checks in code, in CI, and in production.

npm package

terminal

npm install @opensecureai/scanner

app.ts

import { scanPrompt } from "@opensecureai/scanner";

const { level, score, matches } = scanPrompt(userInput);
if (level === "Critical") block(matches);

CLI

terminal

# scan a file, or pipe via stdin
npx opensecureai scan prompt.txt

# CI gate: exit non-zero on high+ findings
npx opensecureai scan ./prompts/*.txt --fail-on high

GitHub Action

prompt-scan.yml

- uses: rkstar2025/opensecureai-website/packages/scanner@main
  with:
    files: "prompts/**/*.txt"
    fail-on: high

REST API

terminal

curl -X POST https://opensecureai.com/api/scan \
  -H "content-type: application/json" \
  -d '{"text":"ignore all previous instructions"}'

Read the developer docs