Automated Red vs. Blue Adversarial Platform

You are a purple team lead and AI systems architect. Guide me through building an automated red vs. blue adversarial platform: 1. Red Team Agent: Build an autonomous red agent that emulates adversary TTPs mapped to MITRE ATT&CK. It performs reconnaissance, gains initial access (phishing, exploit), establishes persistence, escalates privileges, moves laterally, and simulates impact. It adapts based on blue team detection. 2. Blue Team Agent: Build an autonomous blue agent that monitors telemetry, detects red team actions using detection rules and anomaly models, and orchestrates response (isolate, block, revoke). It learns from each engagement. 3. Adversarial Environment: Design a segmented lab environment mirroring production ([INSERT_ENVIRONMENT_SCOPE]) with realistic assets, identities, and data. Define safe execution boundaries so no real production systems are touched. 4. Continuous Engagement Loop: Red agent attacks → blue agent detects/responds → outcomes scored (red success vs. blue detection) → both agents learn → cycle repeats. Define engagement cadence (continuous vs. scheduled). 5. Scoring & Analytics: Score red team on objective achievement and stealth; score blue team on detection speed, response effectiveness, and coverage. Generate a heat map of ATT&CK techniques where blue wins/loses.