Start with one coupled habitat loop
An ecosystem simulation prompt needs more than a list of animals and plants. It needs a habitat value the player can change, a short chain of responses, a fixed order for updating those responses, and a condition that tells the player whether the habitat is holding together. Write that loop before asking for a larger world.
Mireglass Fen is a fictional game example created for this guide. Moonreed, Glimmerwings, Glassbacks, their relationships, and every number below are invented design rules. They are not real species or ecological advice. The prompts are templates for readers to try; no prompt run, playable game, or test is being shown. The worked v0-to-v1 example is a hand calculation, clearly labeled as a simulation.
The player manages one sluice in a small marsh. Raising, lowering, or holding the water changes Moonreed first; the plant count then affects Glimmerwings, whose count affects Glassbacks. The player is not asked to balance a real ecosystem. They are asked to make one fictional game loop predictable enough to inspect.
Define the fictional habitat state
Keep the first version small: one water level, three population counters, and a four-turn objective. Give each value a bound so a request cannot silently create runaway population growth.
| State | Start | What changes it |
|---|---|---|
Water level T | 1 of 5 | The player raises, lowers, or holds the sluice by one step each turn. |
Moonreed R | 1 of 5 | Grows by one when water is 2–4; otherwise falls by one. |
Glimmerwings G | 1 of 5 | Grows when the updated reed count is at least 2; otherwise falls. Starting Glassbacks may reduce it by one. |
Glassbacks F | 1 of 5 | Grows when the updated Glimmerwing count is at least 2; falls only when that count reaches 0. |
| Stability streak | 0 | Increases when all three populations are at least 2 at turn end; otherwise resets to 0. |
The goal is to keep all three populations at 2 or higher for two turns in a row within four turns. This is a made-up win condition, not a claim about what makes a marsh healthy. The count cap, turn limit, and streak give the prompt an inspectable ending instead of an open-ended “make the ecosystem thrive” request.
Prompt for the rule contract before the scene
First ask the builder to restate the transitions without adding new species or systems. This separates a rule that is incomplete from a scene that is merely unfinished.
Treat Mireglass Fen as an invented game ecosystem, not a real wetland model. Convert the supplied rules into a compact state table and deterministic turn order. Preserve only T, R, G, F, and the stability streak; do not add species, seasons, food chains, events, or ecological claims. Define each clamp and threshold. Hand-calculate the provided state T=2, R=2, G=2, F=1 with action hold, using start-of-turn F0 for Glimmerwing pressure. Return the expected state and list any ambiguity before proposing a build. Label this as rule analysis, not a test.Check that the answer uses the starting Glassback count for Glimmerwing pressure and updates the Glassback count afterward. If the response adds weather, migration, food, or extra habitats, remove those additions before moving on; each would introduce another transition to verify.
Request a minimal habitat slice
Once the rule contract is fixed, ask for one screen and the shortest interaction that can expose it. Name the controls, counters, turn limit, and reset. Tell the builder which systems to leave out so the first request stays focused.
Create a small single-screen ecosystem-management game concept for the fictional Mireglass Fen. Let the player choose one action each turn: raise, lower, or hold a sluice. Show a water-level control and readable counters for Moonreed, Glimmerwings, Glassbacks, and the two-turn stability streak. Use exactly the supplied four-turn update rules and fixed order; clamp all values to 0–5. Include a restart action and a clear win state when the stability streak reaches 2. Do not add real species, scientific explanations, settlement growth, trade, unlocks, extra resources, weather, or random events. Keep each turn's start state and end state inspectable so a reader can compare them with the rules.These instructions describe a proposed first slice, not a feature guarantee. When a reader tries them, useful checks include whether raising the sluice changes only water first, whether each population then updates in the stated order, and whether the streak resets whenever any count ends below 2.
Hand-check one habitat turn
Before asking for a build, run the arithmetic on paper. Start at T=2, R=2, G=2, F=1, choose hold, and resolve the rules in order. The starting Glassback count is below the pressure threshold.
| Step | Calculation | State after step |
|---|---|---|
| Hold sluice | T stays at 2. | T=2, R=2, G=2, F=1 |
| Update Moonreed | Water is within 2–4, so R gains 1. | T=2, R=3, G=2, F=1 |
| Update Glimmerwings | Reeds are at least 2, so Gtemp gains 1; starting F0=1 causes no pressure. | T=2, R=3, G=3, F=1 |
| Update Glassbacks | Glimmerwings are at least 2, so F gains 1. | T=2, R=3, G=3, F=2 |
| Check stability | All three populations are at least 2, so the streak becomes 1. | T=2, R=3, G=3, F=2; streak 1 |
The state is a prediction from the authored rules. It does not prove an implementation follows them. The second consecutive hold turn would produce T=2, R=4, G=3, F=3 and streak 2, meeting the example's fictional goal.
Revise one hypothetical rule mismatch
The first draft says that Glassbacks respond to Glimmerwings and reduce Glimmerwings, but it does not say whether that pressure uses the starting Glassback count or the count after the current update. That omission produces two possible hand-calculated states:
| Draft interpretation | End state from the same start and hold action |
|---|---|
Use starting F0=1 for pressure, then update Glassbacks. | T=2, R=3, G=3, F=2; streak 1. |
Update Glassbacks first, then use new F=2 for pressure. | T=2, R=3, G=2, F=2; streak 1. |
The simulated revision adds one line: Glimmerwing pressure uses the start-of-turn F0 value. On that rule, the expected v1 state is T=2, R=3, G=3, F=2, streak 1.
This v0-to-v1 walkthrough is an author-created simulation, not a prompt run, game build, playtest, or measured outcome. If a reader later has a real build, they can replace the hypothetical state with an exact action and observed state before requesting a one-rule correction.
This is a hypothetical rules ambiguity, not an observed game result: from T=2, R=2, G=2, F=1, action hold, one reading ends at T=2, R=3, G=3, F=2 and another at T=2, R=3, G=2, F=2 because the draft never defines which Glassback count drives Glimmerwing pressure. Make only one change: specify that pressure uses F0, the start-of-turn count. Preserve all other rules and return the revised transition and expected state. Do not say this change was tested or measured.Take the rule sheet to a creation workspace
When the brief is ready, describe the audience and game format, name the first player action, define the scene and its state, and add constraints that keep the first version small. Then inspect the actual result: do the displayed counts follow the transition order, and does the stability streak reset under the stated condition?
SEELE's current Free AI game maker page describes a general workflow for describing a game idea in plain English, reviewing a playable 2D or 3D result, iterating in Workspace, and reviewing the selected model, available Koin balance, and displayed generation cost before submitting. It does not document Mireglass Fen, ecosystem-specific behavior, or this example's rules. Use the prompts as starting instructions and verify the result yourself.
Ecosystem simulation prompt FAQ
How many species should a first ecosystem prompt include?
Start with the fewest counters needed to show one dependency chain. Mireglass Fen uses three invented populations to model a habitat change, a growth response, and a pressure rule. Add another population only when it creates a decision the existing loop cannot express.
Should the species in an ecosystem game be real?
They can be, but claims about real species and their relationships need reliable sources. This example deliberately uses invented names and authored rules, so it should not be presented as ecological guidance or a realistic wetland model. If you choose real species, cite a source for each relationship and keep sourced facts separate from invented rules.
What should an ecosystem simulation prompt specify first?
Name the habitat variables, population bounds, player action, transition order, and win condition. Then include one exact starting state and ask for the expected state after a single turn. Those details make a result easier to inspect than a prompt built from atmosphere alone.
How do I resolve conflicting population results?
Find the first rule whose timing is ambiguous. State whether an interaction uses the count from the start of the turn or after another population updates, change that rule alone, and recalculate the same state. A real build should be checked separately before calling the correction tested.
Were the prompts and simulated example tested?
No. The prompts are templates to try, and the v0-to-v1 walkthrough is hand-calculated. It does not show a build, playtest, or measured result. Compare your own output with the rule contract before describing its behavior as verified. Only call behavior observed after recording an exact action and result from a real build.
