CapybaraHow to prompt Capybara for a stronger first game plan
A practical prompting guide for turning a rough game idea into a focused Capybara plan that the builder can turn into a playable first slice.
The best Capybara prompt does not try to describe every asset, animation, and screen. It gives the planning agent a clear game fantasy, then lets the planning conversation shape the first build into something playable.
Start with the promise, not the parts list
A weak prompt says: “make a farming RPG with crops, NPCs, shops, quests, combat, fishing, seasons, and pets.” It sounds complete, but it gives the planner too many equal priorities. The first build becomes a pile of features instead of a game loop.
A stronger prompt says what the player should want, what they do minute to minute, and what makes the game memorable. Capybara’s planning agent is designed to ask one or two high-leverage questions before it creates the plan, so your job is to give it a useful starting point rather than a final spec.
- Name the genre: cozy mystery, farming sim, dungeon crawler, shop RPG, social deduction game.
- Name the player goal: solve one case, restore one shrine, survive one night, make the first sale.
- Name the hook: AI villagers remember rumors, crops mutate at night, suspects destroy evidence, customers haggle in free text.
Let the planner ask the important fork questions
Capybara’s planning flow is intentionally conversational. The planning agent does not jump straight from idea to build. It asks clickable questions when a choice changes the whole first asset slate or gameplay loop: personal toy vs shareable demo, cozy vs tense, simple villagers vs AI NPCs with memory, one room vs separated rooms for suspicion or stealth.
That matters because every fork has a cost. A mystery with suspects who patrol between rooms needs spatial separation. A farming sim with relationship-focused villagers needs different NPC behavior than a crop-only prototype. A commercial demo needs a clear store-page promise, not just a list of cozy features.
- Answer planning questions based on the experience you want players to retell.
- Pick the smallest version that still proves the hook.
- Avoid asking for a giant world before the first loop feels satisfying.
Upload art style references instead of describing the vibe
If visual style matters, upload it as an art style reference before you prompt. Describing “cozy painterly pixel anime with warm Ghibli vibes” can be interpreted many ways. A style upload gives Capybara a stronger visual anchor for maps, characters, props, HUD screens, and the title menu.
Use the prompt for gameplay and the upload for style. That separation keeps the plan cleaner: the planner can focus on what the player does, while the builder has a concrete style reference when generating the asset batch.
- Good style upload: one coherent screenshot, mood board, or concept image.
- Good prompt detail: “rainy harbor detective RPG with AI suspects and a three-clue accusation loop.”
- Bad prompt detail: a long paragraph of competing visual adjectives with no gameplay priority.
Prompt for the first five minutes
Capybara plans a first playable slice, not a wiki for a future franchise. A useful prompt says what should happen after the player presses Start: the short narrated setup, the first objective, the first interaction, the first small win, and the reason to keep going.
This is also where engagement is born. Players stay oriented when they see a clear objective, get immediate feedback, find one surprising reward, and glimpse something they cannot reach yet. Ask for that shape directly.
- “Start with one clue already in the case ledger.”
- “Let the player see a locked garden gate they can open after finding evidence.”
- “Give every failed interaction a useful reason, like ‘Find the brass key first.’”
Before you build
- Upload an art style reference if visuals matter.
- Describe the player fantasy in one or two sentences.
- State the first playable objective.
- Answer the planner’s fork questions honestly.
- Approve a focused plan before building the first asset batch.
Keep reading
More practical guides for building better first slices.