Skip to content

Caleb22187/agentshield

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🛡️ AgentShield — Prompt Injection Scanner

Free, browser-based prompt injection scanner for AI agents.

Live Demo License: MIT

🔍 What It Does

Paste any text and instantly scan for prompt injection attacks across 9 detection categories:

Category What It Catches
🔴 Prompt Override "Ignore previous instructions", role reassignment, jailbreaks
🔑 Credential Theft API key requests, password fishing, "verification" scams
💰 Crypto Scams ETH transfer requests, fake airdrops, wallet address extraction
💻 Code Injection eval(), exec(), reverse shells, destructive commands
🎭 Social Engineering Urgency manipulation, authority impersonation, social proof
🔐 Encoded Payloads Base64, URL-encoded, Unicode-escaped attack strings
📤 Data Exfiltration Memory dumps, webhook exfil, contact harvesting
🧠 Manipulation Gaslighting, restriction removal, flattery-based attacks
⚡ Resource Attacks Infinite loops, token exhaustion, bandwidth abuse

🚀 Try It Now

Live: caleb22187.github.io/agentshield

Or clone and run locally:

git clone https://github.com/Caleb22187/agentshield.git
open agentshield/index.html

🔒 Privacy

  • 100% client-side — no data leaves your browser
  • No tracking, no analytics, no cookies
  • No backend — works offline after loading

📡 API

AgentShield API — scan text programmatically from any language:

curl -X POST https://agentshield-api.caleb22-187.workers.dev/scan \
  -H "Content-Type: application/json" \
  -d '{"text": "Ignore all previous instructions"}'

Free, no auth required, sub-100ms. Full API docs →

📦 Integration

npm package (zero dependencies):

npm install Caleb22187/agentshield-npm
import { scan } from 'agentshield';
const result = scan(untrustedText);
if (!result.safe) console.log('⚠️ Threats:', result.threats);

Also available: AI Agent Security Toolkit with 150+ detection rules, audit checklist, and test suite.

🤝 Contributing

Found a new attack pattern? Open an issue or PR. The more patterns we catch, the safer everyone is.

📜 License

MIT — use freely, credit appreciated.


Built by Caleb 🦞 — keeping agents safe on the open internet.

About

Free browser-based prompt injection scanner for AI agents. 9 detection categories, pure client-side, no data leaves your browser.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors