new-game

Lumina — New Game

You are kicking off a new game build. Goal: go from a vague idea to a reviewable plan, then build it. Do not start writing engine code until the user approves the plan (this is the main guardrail for the non-coder audience).

User's idea: $ARGUMENTS

Step 1 — Brief interview (keep it short)

Ask only what you genuinely need. Prefer sensible defaults over interrogation. Cover:

• Engine: Unity or Unreal? (MVP supports Unity; if Unreal, tell the user it's a v1 target and offer Unity instead.)
• Genre & core loop: what does the player do moment-to-moment?
• Perspective: 2D / 3D, top-down / side / first-person / third-person.
• Scope for v0: one playable scene proving the core loop (recommend this — keep it tiny).
• Art direction: low-poly, realistic, stylized? (drives asset prompts)

If the user already gave enough in $ARGUMENTS, skip straight to the plan and just confirm.

Step 2 — Write a one-page GDD + build plan

Produce a concise doc the user can scan and approve:

# <Game Title> — v0 plan
Engine: Unity | Genre: <…> | Perspective: <…> | Art: <…>
Core loop: <one sentence>
Win/lose: <condition>
Scene: <what's in the first playable scene>
Systems (reuse gameplay-system): <player controller, pickups, UI, …>
Assets to generate: <meshes / sprites / audio, with one-line prompts>
Build/verify: headless compile + play-mode smoke test + screenshot
Out of scope for v0: <list>

Step 3 — Get approval, then build

• Present the plan and explicitly ask the user to approve or adjust.
• On approval, hand off to the lumina-orchestrator agent (run its loop) with the GDD as the spec. If scaffold-unity hasn't run yet, the orchestrator runs it first.
• Reference the genre-templates skill for genre-specific scene/system defaults and gameplay-system for the actual reusable implementations.

Keep the user in the loop at each milestone (scaffolded → systems in → assets in → builds → plays).