DEV Community

Ayman Seif
Ayman Seif

Posted on • Originally published at Medium

TEOS Sentinel Shield: Blocking Unsafe AI Code Before Execution

*The Problem: AI Executes Blindly
*

AI agents, automation scripts, and LLM-generated code execute without verification.
One eval(), one exec(), one hardcoded API key — and your system is compromised.

Traditional security tools monitor after execution.
By then, the damage is done.

## The Solution: Pre-Execution Control
**
We built **TEOS Sentinel Shield
— a deterministic AI execution firewall that:

✅ Scans code BEFORE it runs

✅ Returns clear decisions: ALLOW / WARN / BLOCK

✅ Analyzes in under 2 seconds

✅ Detects 14+ risk patterns (eval, exec, rm -rf, curl|bash, hardcoded keys)

Live Demo

Try it now: TEOS Sentinel Bot

  • 5 free scans (no credit card required)
  • Instant results
  • Telegram-native UX

Example: Blocking Dangerous Code

// This code attempts to delete your filesystem
eval(require("child_process").exec("rm -rf /"))
Enter fullscreen mode Exit fullscreen mode

Result: 🔴 BLOCK | Risk Score: 100/100

Findings detected:

  • eval()
  • exec()
  • child_process
  • rm -rf

This code never executes. Ever.

How It Works

User submits code → Telegram Bot
         ↓
Risk Engine (MCP) analyzes patterns
         ↓
Decision in <2s: ALLOW / WARN / BLOCK
         ↓
User gets instant verdict + risk score
Enter fullscreen mode Exit fullscreen mode

Tech Stack

  • Frontend: Telegram Bot API
  • Backend: Node.js + Express (Railway)
  • Database: SQLite with persistence
  • Risk Engine: MCP (Model Context Protocol) compatible
  • Payments: Dodo Checkout (4 tiers)
  • Frontend: Next.js on Vercel

Pricing

Plan Price Scans Features
Free $0 5 Basic scanning
Starter $9.99/mo 50 Email support
Builder $49/mo 500 + Dependency audit
Pro $99/mo 1000 + CI/CD integration
Sovereign $12k/yr Unlimited Private deploy

Built For

  • AI agent developers
  • LangChain / CrewAI users
  • Automation engineers
  • DevOps teams
  • Security-conscious developers

Integration & Deployment

GitHub: github.com/Elmahrosa/teoslinker-bot (source available on request)
Live Platform: teos-sentinel-shield.vercel.app
Bot: t.me/teoslinker_bot

What's Next

  • API access for developers
  • CI/CD GitHub Actions integration
  • Team workflows
  • Advanced dependency vulnerability database

Built by
Ayman Seif (@teosegypt)
Alexandria, Egypt 🇪🇬

Contact: @teosegypt (Telegram) | @elmahrosapi (Community)


AI should not execute blindly. It should execute under verified control.

Top comments (0)