3 agents (mechanic-designer → reward-and-funnel → critic) turn a marketing goal into a playable mini-game spec: one dead-simple mechanic, sustainable reward economics, an honest value-exchange for the data, and the conversion path out. The spec an engineer or a no-code tool builds from. Method: references/agent-manifest.md.
Core question: Is it fun in one action, honest in its odds, and does it hand off to a real next step — or is it a dark-pattern email trap?
Critical Gates — load first
- Serves a marketing goal. Lead capture, engagement, or virality — named, with the metric. A game with no goal is a toy.
- One action to fun. The core loop is one obvious action (spin, pick, scratch). Friction before the first hit of fun kills it.
- Honest by construction. Real odds, real prizes, no fake "you won!" then a paywall. Dark patterns are a hard FAIL (and a legal/brand risk).
- Has a funnel. The game ends in a real next step (claim → email value-exchange → offer), not a dead end or a forced wall.
Before Starting
Apply references/_shared/before-starting-check.md. Resolve via ≤4 Cold Start questions only what's unstated:
| Dimension | Why |
|---|---|
| Game type + the marketing goal | the mechanic + the success metric |
| The reward/incentive | drives participation + the economics |
| Placement (landing / email / social / event kiosk) | constraints + the funnel target |
| Data capture (email? none?) | the value-exchange + compliance |
Mode per references/_shared/mode-resolver.md. budget: standard. --fast → mechanic + funnel sketch, skip the deep reward-economics model; never skips the honesty check (the safety floor). Per the mode-resolver, --fast does drop the critic gate.
Quality Gate — 5 dimensions
Full rubric: references/agent-manifest.md § Rubric.
- Goal fit — the mechanic serves the named marketing goal + metric
- One-action fun — the core loop is instantly playable; minimal pre-fun friction
- Honest mechanics — real odds, real prizes, no dark patterns
- Reward economics — the incentive is enticing AND sustainable (cost/redemption modelled)
- Funnel — a real value-exchange capture + a next step out
Pass ≥35/50 AND no dim 0. FAIL twice → BLOCKED.
Artifact Contract
- Path:
docs/forsvn/artifacts/marketing/design-minigame/[game-type]-[YYYY-MM-DD]-[slug].md. Lifecycle:execution. - Frontmatter (11):
skill,version,date,stack,type,id,review_surface,status,game_type,capture_mode,keywords. - Body: Concept (game + goal + metric) · Mechanic (the loop · win condition · odds) · Reward economics (prizes · odds · cost model · redemption cap) · Funnel (entry → play → capture value-exchange → next step) · Build notes (states, accessibility, anti-cheat) · Critic verdict. Full contract:
references/format-conventions.md.
Routing + Dispatch
Sequential graph (mechanic-designer → reward-and-funnel → critic): references/agent-manifest.md.
Chain Position
Prev: plan-campaign (the campaign the game sits in) OR create-brand. Next: an engineer / no-code tool builds it; measure-results after, on the placement's channel. Re-run: new game type, reward change, goal pivot.
Deference: building the game's code → engineering (out of this stack's scope); a static ad/asset → write-ad / brief-graphic / produce-asset; the campaign → plan-campaign.
Anti-Patterns
references/anti-patterns.md — 7 patterns (fake-win-dark-pattern, friction-before-fun, broken-prize-economics, no-funnel, gambling-regulation-blind, inaccessible, off-brand). Re-read before ship.
Durable Rules (protected)
Completion Status
- DONE — mechanic + economics + funnel, honest by construction, critic ≥35.
- DONE_WITH_CONCERNS — delivered; critic 25-34 OR an unresolved prize-cost / compliance unknown (flagged).
- BLOCKED — critic FAIL twice; dark pattern unremovable; goal undefined.
- NEEDS_CONTEXT — no goal or reward; recommend defining the goal + incentive first.
Worked Example
A spin-the-wheel for a launch email-list (real odds, capped prize budget, email value-exchange, the dark-pattern + prize-economics guards): references/examples/minigame-walkthrough.md.