Skip to main content

​
Sentinel

Sentinel is in development. This page describes what’s coming.
Sentinel is the inbound protection layer. It validates AI-generated code before you use it.

​
The problem

AI can generate code that:
  • Has security vulnerabilities (SQL injection, XSS, etc.)
  • Uses copyleft-licensed code in your proprietary project
  • Violates your team’s coding standards
  • Includes deprecated or dangerous APIs
You catch these in code review… if you notice them.

​
What Sentinel will do

Security scanning
  • SQL/NoSQL injection patterns
  • Command injection
  • XSS vulnerabilities
  • Insecure crypto usage
  • Hardcoded secrets (yes, AI sometimes adds new ones)
License compliance
  • Detect GPL/AGPL code in proprietary projects
  • Flag license-incompatible dependencies
  • Identify code that looks like known open source
Custom rules
  • Your team’s patterns to enforce
  • Your team’s patterns to forbid
  • Framework-specific best practices

​
Planned API

import { createSentinel } from '@takumo/sentinel';

const sentinel = createSentinel({
  rules: ['security', 'license'],
  licensePolicy: 'permissive-only',  // No copyleft
  customRules: './takumo-rules.yaml'
});

// Validate before using AI-generated code
const result = await sentinel.validate(aiGeneratedCode);

if (!result.passed) {
  console.log('Issues found:');
  for (const issue of result.issues) {
    console.log(`  ${issue.severity}: ${issue.message}`);
    console.log(`    Line ${issue.line}: ${issue.snippet}`);
  }
}

​
Help shape it

What validations matter most to you? Request access and tell us.