Skip to content

Use Case: Game Designer Using AI to Create a Roadmap & PRD

A Game Designer can use an AI coding agent to brainstorm features, write a detailed PRD, and create a design spec -- without knowing how to code.


Context

You're a PM/Game Designer developing a mobile MOBA game called "The Ended World." You need to design a Daily Fortune Spin mini game (spin-to-win wheel) with the goal of keeping players coming back every day.

Tools:

  • AI coding agent (Claude Code, Cursor, or similar)
  • Superpowers framework installed

What You'll Learn

In this tutorial, you'll see how to apply:

  1. Brainstorming skill -- AI asks questions one at a time to explore requirements, doesn't dump everything at once
  2. Visual companion -- AI displays wireframes/mockups directly in the browser
  3. Structured spec -- Turns ideas into a clearly structured design spec

Step 1: Brainstorming -- Exploring Requirements

You tell the AI:

"I need to design a spin-to-win mini game for the game The Ended World."

AI activates the brainstorming skill and asks one question at a time:

Question 1 -- Primary objective:

AI: "What's the primary objective of this spin-to-win mini game?

  • A. Monetization -- generate revenue from players (buy spins with real money/gems)
  • B. Retention -- keep players coming back daily (free spin every day)
  • C. Engagement booster -- increase play time (spin after completing quests)
  • D. Combination of multiple objectives"

You: "B"

Question 2 -- Spin frequency:

AI: "What spin frequency and mechanics do you have in mind?

  • A. 1 free spin per day (simple, classic)
  • B. Spin based on login streak
  • C. Spin after completing daily quests
  • D. Combination: free + extra spins from quests"

You: "A"

Question 3 -- Reward types:

AI: "What rewards go on the wheel?

  • A. In-game resources (gold, gems, exp...)
  • B. Combat items (runes, buffs...)
  • C. Cosmetics (skins, avatar frames, emotes...)
  • D. Mix of all -- tiered by rarity"

You: "D"

Question 4 -- Probability mechanics:

AI: "Rarity mechanics?

  • A. Fixed probabilities (Common 60%, Rare 25%, Epic 12%, Legendary 3%)
  • B. Pity system -- guaranteed after X spins (like Genshin Impact)
  • C. Progressive wheel -- changes based on player level"

You: "A"

Question 5 -- Number of slots and Mystery:

AI: "How many slots on the wheel?

  • A. 6-8 slots, rewards clearly shown
  • B. 8-10 slots, rewards clearly shown
  • C. 6-8 slots, some hidden as "Mystery Reward" (adds a surprise element)
  • D. You have a specific number or different idea?"

You: "C"

Question 6 -- Interface theme:

AI: "Theme for the wheel?

  • A. Post-apocalyptic / Sci-fi -- hologram, digital, ruined futuristic style
  • B. Fantasy dark -- ancient wheel, glowing runes
  • C. Military / Tactical -- military radar
  • D. You already have art direction for the game, I'll follow that style"

You: "A"

Question 7 -- Animation:

AI: "Animation level?

  • A. Simple -- spin and stop
  • B. Medium -- deceleration, glow, light shake, popup
  • C. Grand -- slow-motion, particle effects, Legendary special effects"

You: "B"

Question 8 -- Desired output:

AI: "Platform and tech stack?

  • A. Unity (C#)
  • B. Unreal Engine (C++)
  • C. Cocos Creator (TypeScript)
  • D. No codebase yet, just need the PRD/Roadmap documentation
  • E. Other engine/framework"

You: "D"

Why this step matters:

Without Superpowers, AI would immediately dump a 5-page PRD without understanding the context. With the brainstorming skill, AI is forced to ask one question at a time -- helping the PM make intentional decisions at each step, rather than being overwhelmed by a massive document.


Step 2: AI Summarizes & Proposes Approaches

After collecting enough information, AI summarizes and proposes 3 directions:

FactorDecision
ObjectiveRetention -- keep players coming back
Frequency1 free spin per day
RewardsMix of all (resources, items, cosmetics) by rarity
ProbabilitiesFixed (Common -> Legendary)
Slots6-8 slots, with Mystery Rewards
ThemePost-apocalyptic / Hologram sci-fi
AnimationMedium (deceleration, glow, shake, popup)
OutputPRD/Roadmap documentation

3 proposed approaches:

Approach A: Simple "Daily Spin" A standalone spin wheel, 1 free spin per day, fixed rewards. Simple, suitable for MVP.

Approach B: "Daily Spin + Seasonal Rotation" (AI recommended) Same as A but the reward table changes by season/event. AI recommends this because it balances simplicity with long-term retention -- seasonal rotation keeps the wheel fresh without over-complicating things.

Approach C: "Daily Spin + Collection System" Tied to a collection system -- collect pieces to assemble bigger rewards. Extremely strong retention but complex.

You choose: "A" -- focus on MVP, keep it simple first. Although AI recommended Approach B, you decide to start with the simplest option.

Nice touch: AI gives a reasoned recommendation but respects your final decision. Seasonal rotation can be added later once the MVP is running smoothly.


Step 3: AI Writes the Design Spec Section by Section

AI doesn't dump everything at once. Instead, it presents each section and waits for your confirmation before moving on.

Section 1: Player Flow

Feature name: Daily Fortune Spin

1. Open game -> see spin wheel icon (red badge if spins available)
2. Tap -> opens hologram wheel screen with 8 slots (6 visible + 2 Mystery)
3. Press "SPIN" -> wheel spins, gradually decelerates, light shake near stop
4. Stops -> winning slot highlights, popup shows reward
5. Press "Claim" -> reward goes to inventory
6. SPIN button -> "Come back tomorrow", resets at 00:00 UTC

Conditions:
- Spins don't accumulate (miss a day = lose the spin)
- Can't buy extra spins (staying true to retention goal)

You: "OK" -> AI moves to section 2.

Section 2: Reward Table & Probabilities

SlotTypeRarityExample RewardProbability
1ResourceCommon500 Gold25%
2ResourceCommon200 EXP25%
3ItemRareRandom Rune15%
4ResourceRare20 Gems13%
5ItemEpic24h Combat Buff8%
6CosmeticEpicAvatar Frame7%
7MysteryRare-EpicRandom from hidden pool5%
8MysteryEpic-LegendaryRandom from hidden pool2%

Total: 100%

Mystery Pool details:

  • Mystery Rare-Epic: could drop hero fragments, skin fragments, or 2x gems
  • Mystery Epic-Legendary: could drop rare skins, special recall effects, or new heroes

Rules:

  • Probabilities calculated server-side, client only receives the result and plays the animation
  • Result determined the moment SPIN is pressed, animation is purely visual

You: "OK" -> AI moves to section 3.

Section 3: Visual Wireframe

AI pushes the wireframe to the visual companion -- a local server that displays mockups directly in the browser.

Practical note: During this work session, the visual companion server needed to restart 3 times before running smoothly (port changed from 49983 -> 59628 -> 57802). This is still a fairly new feature -- if you don't see the web page loading, ask AI to restart the server.

After the server is running, you open the link and see:

  • Dark background (#0a0e17), cyan hologram (#00f0ff) as the primary color
  • Circular wheel with 8 slots, glowing cyan border
  • 2 Mystery slots showing orange/red question mark icons
  • SPIN button with cyan border, glow effect
  • Rarity legend at the bottom

Visual companion feature allows AI to display mockups/wireframes directly in the browser, helping PMs visualize the interface during brainstorming.

You: "OK" -> AI moves to section 4.

Section 4: States & Edge Cases

StateDisplayAction
Spin availableSPIN button glows cyan, red badgeAllow spin
Already spun today"COME BACK TOMORROW" + countdownDisabled
Currently spinningButton disabled, spin animationWait for result
Claiming rewardPopup + "CLAIM" buttonTap claim -> inventory
Connection lost"Connection lost, retrying..."Auto retry 3 times

Edge cases:

  1. App closed before Claim -> Server already recorded, reopening shows claim popup
  2. Time manipulation -> Reset based on server time (UTC)
  3. Spinning at exactly 00:00 -> Server locks request, completes before resetting
  4. Inventory full -> Reward sent to in-game mailbox
  5. Mystery pool empty -> Falls back to a fixed Epic reward

Step 4: Roadmap & Milestones

AI creates a 3-phase roadmap:

Phase 1 -- MVP (Weeks 1-2):
  |-- Design UI for 8-slot hologram-style wheel
  |-- Spin logic: fixed probabilities, server-side random
  |-- 1 spin/day, reset at 00:00 UTC
  |-- Medium animation (deceleration, highlight, popup)
  +-- Integrate into main screen (icon + badge)

Phase 2 -- Polish (Week 3):
  |-- Mystery Reward reveal animation
  |-- Sound effects by rarity
  |-- Countdown timer to next spin
  +-- Handle edge cases (connection loss, app close, full inventory)

Phase 3 -- Analytics & Tuning (Week 4):
  |-- Tracking: DAU opening wheel, daily spin rate, D1-D7 retention
  |-- A/B test reward tables
  +-- Internal dashboard for reward distribution

KPIs:

MetricTarget
Daily spin rate> 80% DAU
D7 retention lift+5-10% compared to before the feature
Time to open wheel< 30 seconds (smooth UX)

Step 5: AI Writes Spec Document & Reviews

AI automatically:

  1. Writes the complete spec document to docs/superpowers/specs/2026-03-19-daily-fortune-spin-design.md
  2. Dispatches a reviewer -- a separate agent checks spec quality

Review result: Approved with a few minor recommendations (advisory):

  • Mystery Pool needs more specific config at implementation time
  • Should clarify which edge cases are handled in Phase 1 vs Phase 2
  • Confirm whether "2-3 seconds" is total duration or just the deceleration phase

AI asks about next steps:

After the spec is approved, AI doesn't take initiative. Instead, it asks:

AI: "What would you like me to do next?

  1. Write a detailed implementation plan -- break down into specific code tasks (requires knowing the tech stack)
  2. Complete here -- the PRD/Roadmap is sufficient, you take the documentation to implement with your team
  3. Add more -- more detailed wireframes, API spec, database schema..."

You: "2" -- complete here.

Superpowers principle: AI doesn't expand scope on its own. After completing the original goal, AI asks the user instead of doing more unprompted.


Final Results

After ~15 minutes working with AI, the PM has:

DeliverableStatus
Complete design specCompleted & Approved
Reward table & probabilitiesCompleted
Visual wireframe (hologram style)Completed (viewable in browser)
States & edge casesCompleted
3-phase roadmap (4 weeks)Completed
KPIs for measurementCompleted

Superpowers Principles Applied

PrincipleHow Applied
Brainstorming firstAI asks one question at a time (8 questions) to understand context before writing
YAGNIChose "Simple Daily Spin" instead of the complex Collection System
Present section by sectionSpec divided into 5 sections, each awaiting confirmation before continuing
Visual companionWireframe displayed directly in browser, not just described in text
Auto reviewAI automatically dispatches a reviewer to check spec quality

Takeaway: You don't need to know how to code to use Superpowers effectively. The framework helps PMs and AI collaborate with discipline -- ask before doing, break into sections for confirmation, and automatically review output quality.