Games are an unusually good fit for SpecDevShip precisely because they need both of the workflow's collaboration models at once: tight, gated discipline for systems that must actually work, and real creative autonomy for the parts that are fundamentally aesthetic judgment calls.
Where full tier actually belongs in a game codebase
A new save-system architecture, a multiplayer networking layer, a core progression-system overhaul — these are full-tier by any reasonable read, carrying the kind of migration and architectural risk the tier decision is built to catch. A new enemy type or a UI screen for an existing menu system is usually light tier, even though it might involve plenty of creative work — the tier decision is about technical/architectural risk, not creative scope.
Art direction and design direction, doing what they're built for
This is the clearest case for art-direction and design-direction's "pitch a few directions, get a pick, then run with full autonomy" model. A game's visual and UX identity is precisely the kind of decision where forcing a back-and-forth over every specific choice produces worse results than trusting a clear, chosen direction. Grounding that direction in any prior art-analysis (if you're matching or departing from a reference game's look) keeps the autonomy purposeful rather than arbitrary.
Investigation for the bugs that only reproduce sometimes
Game bugs are disproportionately state-dependent, timing-dependent, or dependent on a specific sequence of player actions nobody thought to test. The Investigate skill's insistence on verifying real behavior — running the actual scenario, not reasoning about it from code — matters enormously here, where "probably a physics timing issue" is an easy, wrong, and expensive guess to act on.
Product-analysis for cloning and homage work
A huge amount of game development is explicitly building on or responding to existing games — a clone-with-a-twist, a spiritual successor, a genre entry that needs to hit familiar beats. product-analysis and design-analysis extracting what makes a reference title's core loop and identity actually work — rather than a vague creative impression of it — is a genuinely different and more rigorous starting point than "make something like X."
Games don't need less discipline than enterprise software. They need discipline applied unevenly, on purpose — tight where systems have to work, loose where the whole point is a creative call.