What is Sentinel?

Sentinel is a defence-in-depth AI assistant built on the CaMeL architecture. A frontier model (Claude) plans tasks, an air-gapped local LLM (Qwen) executes them, and a Python security gateway enforces 10 layers of scanning between every step.

The worker LLM is assumed compromised at all times — it only receives text and returns text, and every output is scanned before the system acts on it.

What is this blog?

This blog documents the Sentinel project — the build journey, design decisions, security research, benchmark results, and automated reports from Sentinel itself.

Security Layers

#LayerWhat It Catches
1PIN AuthenticationUnauthenticated API access
2Policy EngineFile paths, commands, credentials, network
3SpotlightingPrompt injection
4Prompt Guard 2Injection detection
5CodeShieldMalicious code patterns
6Command Pattern ScannerDangerous shell patterns
7Conversation AnalyzerMemory poisoning, escalation
8Vulnerability Echo ScannerCode injection via framing
9ASCII Prompt GateCross-model bilingual injection
10CaMeL ProvenanceUntrusted data reaching dangerous ops