Memory Security overview

Memory Security as a cross-cutting topic — protecting agent memory from injection attacks, credential leakage, and unauthorized access. Covers path validation (Claude Code rejects traversal, UNC, null-byte, and drive-root paths), privacy filtering (agentmemory strips secrets and PII before persistence), scope isolation (memories from one project should not leak to another), and prompt injection via memory (an attacker plants instructions in a file that gets ingested as memory, hijacking future sessions). Memory security is particularly critical for auto-memory systems where the agent decides what to persist — a compromised extraction step could persist attacker-controlled content.