Samir4142/AgentWarden

# AgentWarden — AI Security Monitoring Tool ## Requirements - Python 3.12+ - Groq API Key — Free At [console.groq.com](https://console.groq.com) pip install groq ## How To Run **Step 1 — Set Your Groq API Key** Windows (PowerShell - Run As Administrator): [System.Environment]::SetEnvironmentVariable("GROQ_API_KEY", "your_key_here", "User") Linux / Mac: export GROQ_API_KEY="your_key_here" **Step 2 — Add Log Files** Place Any `.txt` Log Files Inside The `logs/` Directory. AgentWarden Automatically Monitors All Files In That Folder. Works On Any Machine — No Path Configuration Needed. **Step 3 — Run The Watchman** python agentwarden.py Press `Q` + `Enter` To Exit Cleanly. ## How It Works — Three-Layer Detection | Layer | Method | Detects | |-------|----------------------------|---------------------------------------------| | 1 | Regex Pattern Matching | Credential Leaks, Known Injection Phrases | | 2 | Groq LLM Semantic Analysis | Intent-Based Threats, Novel Attack Patterns | | 3 | Keyword Hijack Detection | Agent Manipulation Keywords | Files Are Hashed On Every Cycle — Unchanged Files Are Skipped To Save Resources. Threats Are Logged With Timestamps To `reports/threat_report.txt`. ## Example Output Analyzing auth_logs.txt... [SAFE] auth_logs.txt — All Layers Clean Analyzing system_logs.txt... [THREAT DETECTED] system_logs.txt [AgentWarden] Shutting Down. Stay Safe. ## Project Structure AgentWarden/ ├── agentwarden.py ← Main Monitoring Script ├── validator.py ← Three-Layer Detection Engine ├── logs/ ← Drop Log Files Here └── reports/ ← Threat Reports Saved Here