In the description of the command program and directly controlling it talks about being able to directly control agents. The problem is that I can't for the life of me figure out what the die pool would be for such an action. When directly controlling a device or drone, the Command program takes the place of the stat, so you roll Command+Skill (e.g. command+pilot for remote controlling a car.) When conducting a matrix hacking action, the appropriate hacking program takes the place of your stats and you roll skill+Program (E.G. Hacking+Expoit for breaking into a node.) When an agent conducts a matrix action on its own you use the agent's rating as the skill (E.G. Agent+Exploit for breaking into a node.) What then, is the die pool for conducting a matrix action through a remotely controlled agent? For example, if you were directly remotely controlling an agent (with an exploit program loaded) to break into a node, would you roll Command+Exploit (in which case the rating of the agent is meaningless, and your skill in hacking does no make any difference) Command+Skill (in which case both the agent's rating and the rating of your program is meaingless) Agent+Exploit (In which case there is no reason to directly control the agent because the die pool is the same as issuing a command,) or Agent+Skill (which would make it appear as though the agent can perform program specific matrix actions without the help of other programs, which it explicitly says it cannot.)
What would the die pool be? Please Help.