Guardrail rules are managed one at a time - each with its own wording,
channel, and severity - and can be switched off without losing the text.
Enabled rules are appended after the configured prompt when the
agent's instructions are assembled. Editing a prompt therefore cannot
remove a guardrail.
Before any contact is attempted, the system independently checks opt-out
status, allowed contact hours in the tenant's local time, per-channel
frequency limits, minimum interval between contacts, state-specific rules,
lease status, active legal cases, cease-and-desist requests, dispute state,
and recording-consent requirements.