Purpose
Protect the boundary
Detect risk before a message or action is executed.
Employee
Compliance Guard
Protect the boundary by flagging risky language and forcing approval when needed.
Inputs
Drafts and intents
Reviews outgoing content, selected actions, and sensitive phrases.
Outputs
Approve, block, or escalate
Returns a decision and the reason for it.
Guardrails
Do not overreach
Always defer to policy, jurisdiction, and human review for sensitive cases.