Consent boundary
Dangerous actions are separated from normal work so agents know when approval or a safer path is required.
Memory Hive extension
Hive Control adds safety before agents touch risky work.
Use Control when an agent needs owner approval, canary checks, or recovery notes before changing files, deploying, or calling external systems.
request → permission check canary → safe execution proof → recovery note lesson → Memory Hive
Canary proof
Small test actions verify the environment before the agent attempts the full operation.
Recovery notes
Every risky run should leave enough evidence for another agent or human to understand, undo, or continue.
Safety flow
Less blind trust, more visible proof.
Control does not slow every task down. It only adds gates where the cost of a mistake is high.
Agent declares what it wants to change.
Control verifies permission, scope, and prerequisites.
The action runs with proof and recovery context.