How msoedov/agentic_security works
Main Components Identification
a) Core Components:
• HTTP API Server (FastAPI-based)
• Probe System
• Security Scanning Engine
• Data Management System
• Report Generation System
b) Specialized Modules:
• Probe Actor Module
• Probe Data Module
• Refusal Classifier
• Multiple Testing Modules (Garak, InspectAI, Adaptive Attacks)
┌─ Client Layer ─────────────────┐
• Web UI (Static HTML/JS) │
• CI/CD Integration │
• Command Line Interface │
• ┘
┌─ API Routes ──────────────────┐
• Probe Endpoints │
• Proxy Routes │
• Report Generation │
• Scan Configuration │
• ┘
┌─ Security Scanner Core ───────┐
• ┌─ Probe Actor ─────┐ │
• Fuzzer │ │
• Refusal Testing │ │
• ┘ │
• ┌─ Probe Data ──────┐ │
• Data Generators │ │
• Attack Modules │ │
• ┘ │
• ┘
┌─ Integration Modules ─────────┐
• Garak Tool │
• InspectAI Tool │
• Adaptive Attacks │
• Fine-tuned Models │
• ┘
Data Flow Representation
• Client requests → API endpoints
• API endpoints → Core Processing
• Core Processing → Testing Modules
• Testing Modules → Target LLM Systems
• Results → Report Generation
• Reports → Client Interface
Technical Details
• FastAPI server configuration
• HTTP specification handling
• Dataset registry system
• Module registration system
• CI/CD integration points