Agent Security Hardening

Security hardening patterns for production AI agents. Covers prompt injection defense (7 rules), data boundary enforcement, read-only defaults for external i...