🔨 A.N.V.I.L. (Automated Network & Vulnerability Identification Logic)

The Multi-Agent Safety Engineering Framework.

---

⚠️ The Problem

Safety is not one-dimensional. A system can be:

• Mechanically Safe (The brakes work).
• Psychologically Unsafe (The operator is disgruntled and has an override key).
• Physically Impossible (The plan relies on perpetual motion).

Standard LLMs blur these lines, giving you a "blended" average that misses critical specific risks.

🛡️ The Solution

A.N.V.I.L. is a multi-agent orchestration framework that subjects a system design to three distinct, adversarial audits:

1. The Engineer (STPA): Looks for missing control loops and unsafe actions.
2. The Psychologist (CPIR): Looks for human stressors and insider threats.
3. The Physicist (T1): Validates the fundamental physical constraints.

The Forge (Orchestrator) synthesizes these reports. If any agent flags a Critical Risk, the system is rejected.

🚀 Quick Start

pip install -r requirements.txt