You are correct, arguments do not take place in a void, the arguers have to share common principles or presuppositions to make an argument possible. If both are allowed to reject each other's premises and background assumptions without limitations they will just be talking past each other.
Logical pragmatics, tacit and context dependent aspects of arguments, is a big part of modern argumentation theory. The language of "inside and outside the argument" was used on the Less Wrong blog recently. The more traditional term is "begging the question against the opponent", a move where one of the parties argues from, or presupposes, a premise she knows (or should know) the opponent would reject (or rejects a premise he accepts). In your example Bob begs the question against Jim, he is "outside" the argument.
Unfortunately, Wikipedia's article on begging the question overlooks its pragmatic dimensions, and simply identifies question-begging with circular reasoning. But circular reasoning may well be pragmatically justified in a context if the supposedly circular premise happens to be shared by the participants. It is only begging against the participants' premises that violates the tacit "pact" made at the start of the argument. You can look at Walton's Begging the Question as a Pragmatic Fallacy for a discussion:
"Robinson argued that there are only two proper ways of condemning an argument - because the conclusion does not follow from the premises, or because the premises are not acceptable to the person to whom the argument was directed. Arguing that begging the question does not fit into either category, Robinson concluded that it is not a proper criticism of an argument.
Robinson, continued to build up his skeptical case by arguing that begging the question has traditionally been thought to be a fallacy because it is a breaking of the rules of the old-fashioned game of elenchus (two-person contestive question-reply argumentation as found in Aristotle). Each participant has a conclusion (question) to be proved, and one of the rules, according to Robinson, was that a question must not directly ask for conclusion."
Robinson gave an example similar to the OP's:"God has all the virtues. Therefore, God is benevolent". On the traditional account the argument begs the question by assuming its conclusion. But as long as the premise is "acceptable to the person to whom the argument was directed" it is contextually valid nonetheless. The objection is "outside" the argument. On the other hand, if the purpose of a debate is a "search for truth" then the strictures of the elenchus may not apply, as Robinson himself argued. This may reflect a pragmatic pact among the participants to keep their assumptions open to external evidence. For instance, Rips in Argumentative Thinking describes an approach of Hahn, Harris, and Corner:
"This approach toward argumentation takes the strength of an inference to be the conditional probability of its conclusion given available evidence, where the evidence in some cases may be drawn from outside the argument itself. In the case of arguments from ignorance, for example, the strength of an argument such as “Drug X is safe because JAMA reported 10 studies with no side effects,” depends on the strength of the evidence (the number of studies) and on the reliability of the source (JAMA)."
Chomsky argued, controversially, that adherence to "established" premises in public debates is aided by "concision" (saving time) in the media, which leads to conformism and "thought control", see Dimock's Critiquing Debate:
"Debaters who keep their positions within the very narrow range of the dominant paradigm have a considerable tactical advantage over those who attempt to argue from outside that paradigm. Concision and the overreliance on authority are practices which make it very difficult to challenge the dominant paradigm. Because they reinforce the dominant ideology, which has tended to favor some groups... while marginalizing others."