Guardrail node and more system tools

A new GUARDRAIL node lets you check inputs and outputs against configurable policies — PII, profanity, jailbreak attempts — and route to a fallback path when something's caught, instead of hand-rolling the check in a CONDITIONAL node.

Alongside it, the built-in system tools available to agents and TOOL nodes are expanding beyond the current set, so fewer workflows need a full MCP server for common operations.