Building Safe Agents with Long-Term Memory: SoulScan, Persona Engine & Swarm Memory

SoulClaw v2026.3.21 introduces three safety features for autonomous AI agents: SoulScan (inline security scanning for soul files with prompt injection and data exfiltration detection), Persona Engine (drift detection that monitors agent personality deviation from SOUL.md definitions), and Swarm Memory (git-based memory synchronization across multiple agents and devices). These tools address the gap in agent safety that emerges when AI agents run continuously with tool access and sensitive data.