Metagaming is the use of out-of-character knowledge in an in-character situation...
Though originally defined in relation to role-playing games, similar phenomena may arise in philosophical debates. That is: when engaging in hypothetical reasoning, we need to be careful not to illegitimately import external information (e.g. the external fact that we are engaging in hypothetical reasoning) into the space of hypothesized claims. Let me illustrate with a few examples.
(1) Conflating the supposition that p with the supposition that the supposer supposes that p (as discussed here). Note that conditional proof begins by supposing that p, noticing that q follows, and hence concluding that if p then q. This would go badly wrong if one were to reason as follows: Suppose that p. So, you're supposing that p. Hence, if p then you're supposing that p.
This is bad reasoning because we began by merely supposing that p, and not that anyone is supposing it. It happens to be true, "externally" speaking, that in the process of making this argument we were indeed supposing that p. But that wasn't included among the hypothesized facts, and hence isn't eligible for inclusion in the conditional proof (which draws only on hypothesized, not external, facts).
(2) Conflating actual with hypothesized judgments. Sometimes we face the opposite problem. Rather than drawing on external information when we shouldn't, in this case we get so wrapped up in a scenario that we fail to draw on our independent/external judgments when we should. In moral philosophy, especially, we often need to evaluate counterfactual scenarios in which our counterparts are massively deluded or have manipulated desires/values. In such cases, it is vital to appreciate that our considered judgment about how we prefer the scenario to unfold, need not conform to the counterfactual desires that we are stipulated to have in the scenario. As explained here, there's an important difference between 'I do desire that, were I in situation S, P' and 'In situation S, I would desire that P'. (I suspect that this conflation often underlies knee-jerk subjectivism.)
(3) Fleshing out schematic arguments. As recently noted, when discussing arguments indirectly (e.g. by way of a schema with mere placeholders 'P' and 'Q'), it's crucial that one understands what features are built into the fleshed-out argument, and what features are merely used 'externally', e.g. in selecting and setting up how the argument is to be fleshed out. The linked post explains in more detail how this applies in case of the zombie argument in particular.
The key insight to note is that there's a world of difference between (i) conceiving and reasoning from an opaque statement like 'the complete specification of the [non-]physical base facts', and (ii) conceiving and reasoning from the actual specification P [or NP] -- which is selected to be, as it happens, a complete list of the [non-]physical base facts, though there might be nothing internal to P [NP] itself which specifies its completeness. Failure to appreciate this distinction can lead to some very sloppy reasoning, as further explained in my linked discussion.
(4) Other metaphysics/epistemology mix-ups. You can fill in your own favourite examples here. They all seem to share the common feature of conflating what is supposed to be true with what is supposed to be supposed (believed, known, etc.). I wonder whether the 'metagaming' metaphor would help students to better understand (and hence learn to avoid) this common mistake? (It seems helpful to me, but is the idea still as clear to those less-geeky souls who've never had the pleasure of playing RPGs?)