Agent Firewall blocks now say who can lift them and where

When the Agent Firewall blocked a request, the agent used to be able to say only "ask the workspace owner," which left people guessing. Every firewall refusal now explains itself: which firewall stopped the action, where it can be allowed, and who is allowed to change it.

How it works

• One shared answer. For each firewall that denied an operation (the workspace's, the connection's own, or both), the refusal names the page that allows it: Settings, Agent Firewall for workspace rules, or the connection's own page.
• Who can lift it. The refusal says whether you can change the rule yourself, whether it takes a workspace Owner or Admin, or names the connection's owner. Only a connection's owner can change that connection's firewall; a workspace admin can't.
• Everywhere agents run. The same answer comes from the sandbox's outbound proxy, the geni exec command, and the integration relay, and the agent relays it with a link.
• Warned before trying. When an agent looks up an API operation that's blocked for a connection, the docs say so before it makes the call.
• Never fails open. Older sandboxes ignore the new details, and an unreadable policy falls back safely, so a version mismatch can't switch enforcement off.