AI Prompts for a Marble Auto-Battler: From Rules to a Build Plan

Write AI prompts for a marble auto-battler that connect physical collisions and hero abilities to working gameplay. Move from a first brief to a playable prototype.

Seele Editorial TeamUpdated November 17, 2026

A marble auto-battler lets players plan a collision and then watch the lineup resolve. The setup phase can carry strategy: which hero launches first, which lane catches an enemy, and whether a limited ability changes the impact. The simulation becomes easier to understand when every contact produces a visible result that follows a small set of rules.

The heroes, arena, values, and combat below are fictional game-design examples. The prompt sequence is unrun; no playable result or real physics test is claimed.

Two colorful hero marbles line up opposite an armored enemy marble on a compact fictional auto-battler arena.
Show the lineup and collision lane before the automatic round begins.

Define the arena and team

Separate setup decisions from automatic resolution. In this fictional example, the arena has two launch lanes. During setup, the player places one hero in each lane and chooses a launch direction. A marble travels until it contacts the first enemy in its lane or reaches a boundary. After each collision, the game reports impact, damage, and any displacement before continuing.

Give each hero a small profile. The Amber Scout has 4 HP and impact 2. The Amber Bulwark has 6 HP and impact 1; its contact pushes a target one marked space. A one-use hero ability, Overcharge, adds 1 impact to the first launch. The opposing Bronze Sentinel has 6 HP and an intent to deal 2 damage to the last ally that collided.

UnitHPImpact / effectRole
Amber Scout42 impactFast first strike
Amber Bulwark61 impact; push 1 spaceControl and follow-up
Bronze Sentinel6Intent: 2 damageTarget to defeat before response

Keep an automatic round deterministic while developing the rules. If two units reach the same contact point, specify which collision resolves first. Define whether the arena edge stops or reflects a marble, how a pushed target interacts with a wall, and when HP reaches zero. These choices prevent a dramatic collision from creating ambiguous state.

Write collision and ability rules

Describe a collision in resolvable terms: identify the moving marble, find the first valid target, apply impact damage, resolve a push, check for defeat, then continue any remaining launches. The player should know whether a push creates another collision or simply moves the target to a new tile. Start with one clear rule for each effect.

Overcharge is a setup ability, not an extra collision. Mark it unavailable after the first launch, add 1 to that launch's impact, and keep the team's base impact unchanged. If the Scout is placed in the first lane, its impact is 3 for that contact. Do not accidentally apply the bonus again when the marble rebounds or when the next hero launches.

A fictional auto-battler diagram shows an Overcharge glow on a launched hero marble, collision arrows, enemy health, and a one-space push marker.
Keep the ability bonus and collision feedback visible as separate rule steps.

A collision event can be expressed as a small record:

{
  "attacker": "amber_scout",
  "target": "bronze_sentinel",
  "baseImpact": 2,
  "abilityBonus": 1,
  "damage": 3,
  "abilityConsumed": true
}

Record base impact separately from temporary bonuses. That makes a later upgrade easier to audit: the same Scout remains impact 2 after Overcharge is spent. Add feedback for the target's HP change, a brief collision animation, and a readable indication of the ability's used state.

Trace a two-round fight

Round 1 begins with the Scout in lane one and the Bulwark in lane two. Use Overcharge on the Scout's first launch. The Scout deals 3 damage; the Bulwark then deals 1. The Bronze Sentinel falls from 6 HP to 2. Its intent triggers after both collisions and hits the last ally that collided, the Bulwark. The 2 damage reduces Bulwark HP from 6 to 4.

In Round 2, the Scout launches first for its normal impact of 2. That contact reduces the Sentinel from 2 HP to zero. The enemy is defeated, so its intent does not resolve. The Bulwark remains at 4 HP. This outcome makes launch order meaningful: the player spends Overcharge once, then keeps the more durable ally ready for a later node.

StepScout HPBulwark HPEnemy HPResolution
Start466Overcharge ready
Round 1 — Scout463Impact 2 + 1 ability bonus
Round 1 — Bulwark462Impact 1
Round 1 — intent442Bulwark takes 2 damage
Round 2 — Scout440Impact 2 defeats enemy; skip intent

This is a hand-calculated rules example, not a collision simulation or playtest. Use it as a test case for state updates: Overcharge is spent once, each collision subtracts impact from target HP, the Bulwark push uses its own rule, and defeated enemies do not trigger an intent.

Stage prompts and check the feedback

Build the prompt in layers. First request a two-lane arena and two heroes. Then specify collision order, the one-use ability, and enemy intent. Finally ask for impact feedback and checks that compare displayed HP with underlying state. Avoid adding upgrades, obstacles, and multiple enemies before a single lane resolves clearly.

Create a compact fictional marble auto-battler with two launch lanes. The Amber Scout has 4 HP and impact 2. The Amber Bulwark has 6 HP and impact 1; its collision pushes the target one marked space. Overcharge is a once-per-encounter ability that adds 1 impact to the first launch only. The Bronze Sentinel has 6 HP and deals 2 damage to the last ally that collided after both heroes launch, if it survives. Resolve collisions in lane order, show each HP update, and skip enemy intent after defeat. Do not add random physics, extra abilities, or hidden bonuses yet.

These templates are unrun drafts. When using them, record the exact prompt, tool, date, output version, and correction. A storyboard or static mockup can illustrate intended feedback but cannot prove that a collision resolves correctly in a playable build.

SEELE's official AI Game Maker page describes text-based game creation, browser testing, and follow-up iteration. This is general product context only; this marble auto-battler prompt has not been run and no generated result is claimed.

  • Lane check: each launch hits the first valid target in its lane.
  • Ability check: Overcharge adds 1 impact once, then becomes unavailable.
  • Collision check: damage and push resolve in the stated order.
  • Intent check: the enemy hits the last colliding ally only if it survives.
  • Defeat check: an enemy at zero HP does not act.
Two stylized hero marbles remain in the fictional arena while the defeated enemy rests in the background and a dimmed ability icon appears on the lower status strip.
Make the round's result and spent ability visible before offering the next encounter.

Once the two-lane baseline is clear, add one obstacle or one new hero ability. Keep collision order unchanged and compare the new result with the baseline. That helps the player understand whether a change came from lineup, impact, or an arena rule.

Marble auto-battler prompt FAQs

What should players control in an auto-battler?

Let them make a few meaningful setup choices, such as lane, launch order, or a limited ability, then make automatic resolution visible and consistent.

How do I make collisions readable?

Show the moving unit, target, impact, HP change, and any displacement as separate feedback steps. Keep the order stable before adding chained contacts.

Should hero abilities change base stats?

Track temporary bonuses separately from base impact. That makes single-use abilities and later upgrades easier to inspect.

When should enemy intent resolve?

Define one stable phase, such as after all allied launches. Skip it if the enemy has already been defeated.

Are these prompts or marble physics tested?

No. They are unrun prompt drafts and the encounter is a fictional hand calculation. No playable game or physics test is claimed.

Make your own marble auto-battler

Start with two lanes, two heroes, and one enemy intent. Give collision order and ability limits a clear rule, then inspect the same round state after each prompt revision. For general game-creation context, explore the SEELE AI Game Maker and adapt this unrun prompt sequence to your own prototype.

SEELE AI

Turn this guide into a playable build.

Start with a focused brief, then iterate in the Seele AI Workspace.

Start creating