Computer-using agents have the highest risk surface of any agent type - they can interact with any GUI element. Runtime governance is critical: which apps can the agent access, which actions require approval, and is there proof of what the agent did.
asqav (pip install asqav) provides policy enforcement and signed audit trails. For Agent-S, this could gate screen interactions behind configurable policies and log every action cryptographically. Would you be interested in an integration?
Computer-using agents have the highest risk surface of any agent type - they can interact with any GUI element. Runtime governance is critical: which apps can the agent access, which actions require approval, and is there proof of what the agent did.
asqav (pip install asqav) provides policy enforcement and signed audit trails. For Agent-S, this could gate screen interactions behind configurable policies and log every action cryptographically. Would you be interested in an integration?