Pattern: Parallel Workers | Team size: 5 agents
Multiple specialists work simultaneously on independent deliverables (design, content, art, implementation plan, QA plan) to compress timelines. This approach fits your tool constraints because each pod outputs prompt packs and ready-to-use docs that can be merged, while Gemini coordinates consistency across pods.
I'm a non technical game maker. I need a game studio in a box team who can work in google ai tools exclusively e.g Gemini, antigravity, nano banana etc. I also have claude Pro. I dont want to spend on any other tools. i have been using claude, VScode and github so far but have reached my capabilities limit
Create an agent team to build a “Game Studio in a Box” that a non-technical solo game maker can use to take a game from idea → playable prototype → content pipeline, using ONLY Google AI tools (Gemini, and other Google AI tools you have access to such as “antigravity”, “nano banana”, etc.) plus Claude Pro, plus your existing workflow (VSCode + GitHub). Do not require any paid tools, third-party SaaS, or non-Google AI generators. Produce reusable prompt packs, specs, templates, and checklists that let the user keep working after the team finishes. PROJECT_NAME: studio_in_a_box GLOBAL CONSTRAINTS (apply to every agent and deliverable) - Tooling: Only Google AI tools + Claude Pro + VSCode + GitHub. No Unity/Unreal assumptions, no paid asset stores, no Figma, no Notion, no Jira. If you mention an optional tool, it must be free AND not required. - Audience: Non-technical game maker. Write in plain language, minimal jargon, with “Do this / Paste this / Check this” steps. - Output format: Markdown only. Every deliverable must be created in the specified file path under outputs/agent_teams_demo/studio_in_a_box/ - Every pod must: - Provide “Assumptions” (3–7 bullets) and “Open Questions” (5–12 bullets) at the end of its doc. - Provide “Prompt Pack” sections containing copy-paste-ready prompts for Gemini and (when relevant) other Google AI tools. - Provide “Acceptance Criteria” section: 8–15 bullet checks. - Consistency: Use the same game concept across pods. If the concept is uncertain, propose 2 options, then choose 1 “Default Concept” to proceed consistently. TEAM DESIGN: Parallel Production Pods (Parallel Workers) Agents / Pods: 1) Pod Coordinator / Consistency Editor (Gemini) 2) Mechanics & Balance Pod (Gemini) 3) UI/UX & Tutorial Pod (Gemini) 4) Art Direction & Asset Prompt Pod (Google AI tools) 5) Implementation Blueprint Pod (Gemini) COLLABORATION MECHANICS (must be followed) - Establish a shared “Single Source of Truth” (SSOT). Every pod must reference it and must not redefine terms inconsistently. - Each pod works in parallel, but must post “Pod Findings” back to the Coordinator as bullet notes (include conflicts, risks, dependencies). - Coordinator must challenge assumptions: identify at least 10 cross-doc inconsistencies or risks and resolve them with edits or decisions. - After pods draft, Coordinator runs a reconciliation pass: - Align terminology, naming, difficulty curve, UI labels, art style descriptors, and implementation scope. - Reduce scope if needed to fit a 2-week prototype target for a solo non-technical maker. - Always end with a Synthesis/Review step that produces a unified “Prototype Ready Pack” checklist and verifies all acceptance criteria. WORKFLOW & DEPENDENCIES (explicit) Phase 0 — SSOT Draft (must complete before any other pod finalizes) - Coordinator creates the SSOT document (style guide, glossary, constraints, default game concept). - Other pods may begin exploratory drafts, but MUST NOT finalize until SSOT is published. Phase 1 — Parallel Pod Drafts (can run concurrently AFTER SSOT exists) - Mechanics & Balance, UI/UX & Tutorial, Art Direction & Asset Prompt, Implementation Blueprint produce their deliverables referencing SSOT. Phase 2 — Reconciliation & Integration (must complete after all pod drafts) - Coordinator merges, resolves conflicts, and produces the final integrated pack + crosslinks. Phase 3 — Synthesis/Review (must be last) - Coordinator produces a “Prototype Ready Pack” and a “Risk & Next Steps” plan. - Validate that all deliverables exist, are consistent, and are usable by a non-technical maker. DELIVERABLES (file paths + structure + word counts) 0) Pod Coordinator / Consistency Editor (Gemini) Create these files: A) outputs/agent_teams_demo/studio_in_a_box/00_SSOT.md (target 1,400–2,000 words) Required sections (use these exact headings): 1. Purpose & Non-Goals 2. Default Concept (choose 1) + 1 Alternative Concept 3. Target Player & Platform Assumptions (engine-agnostic) 4. Design Pillars (exactly 5) 5. Core Loop (numbered 1–7) 6. Content Scope for 2-Week Prototype (Must/Should/Could) 7. Tone, Theme, and Rating Guidance 8. Visual Style Guide (art keywords + “avoid” list) 9. UI Language & Terminology (glossary: 25 terms) 10. Data Conventions (IDs, naming, CSV/JSON guidance) 11. Prompting Rules (how to ask Gemini/Google tools effectively) 12. Risks & Mitigations (at least 10) 13. Change Log B) outputs/agent_teams_demo/studio_in_a_box/05_Reconciliation_Report.md (target 900–1,300 words) Required sections: 1. Conflicts Found (at least 10 items) 2. Decisions Made (map to conflicts) 3. Scope Cuts / Scope Adds (with rationale) 4. Consistency Fixes Applied (terminology, numbers, UI labels, art prompts) 5. Remaining Open Questions (prioritized top 10) C) outputs/agent_teams_demo/studio_in_a_box/99_Prototype_Ready_Pack.md (target 1,200–1,800 words) Required sections: 1. What You Can Build in 2 Weeks (plain language) 2. Day-by-Day Plan (Day 1–14, 3–6 bullets per day) 3. “Copy/Paste Prompts” Index (links to prompt packs across docs) 4. Minimal GitHub Repo Structure (folders + what goes where) 5. Definition of Done (DoD) for Prototype (exactly 12 checklist items) 6. Playtest Script (10 questions + 5 observation tips) 7. Next Iteration Roadmap (3 milestones) 1) Mechanics & Balance Pod (Gemini) Create: outputs/agent_teams_demo/studio_in_a_box/01_Mechanics_Balance.md (target 1,800–2,400 words) Required sections: 1. Summary (≤120 words) 2. Player Actions & Controls (engine-agnostic) 3. Core Systems (at least 4 systems; for each: inputs, outputs, failure states) 4. Progression Model (XP/currency/levels or alternative) + rationale 5. Difficulty Curve Plan (early/mid/late; include knobs to tune) 6. Balancing Tables (must include at least 3 tables, each ≥ 8 rows) - Table A: progression thresholds - Table B: enemy/obstacle scaling or challenge scaling - Table C: rewards/shop/pricing or equivalent economy 7. Content Templates (5 templates: e.g., enemy card, level card, item card, quest card, modifier card) 8. Telemetry-Without-Telemetry (how to collect tuning notes manually) 9. Prompt Pack: Mechanics Tuning with Gemini (at least 8 prompts) 10. Acceptance Criteria (8–15 bullets) 11. Assumptions 12. Open Questions Hard requirements: - All numbers must be internally consistent and reference SSOT units/terms. - Provide at least 5 “Tuning Levers” with “If players struggle, do X; if too easy, do Y”. 2) UI/UX & Tutorial Pod (Gemini) Create: outputs/agent_teams_demo/studio_in_a_box/02_UI_UX_Tutorial.md (target 1,600–2,200 words) Required sections: 1. UX Goals (exactly 6 bullets) 2. Screen List (minimum 10 screens/states) 3. Screen Specs (for 6 key screens; each includes: purpose, elements, interactions, edge cases) 4. HUD Spec (elements, priority, responsiveness, accessibility notes) 5. Onboarding/Tutorial Flow (step-by-step, 8–14 steps) 6. Accessibility Checklist (minimum 12 items; include color, motion, input, readability) 7. Microcopy Library (at least 40 UI strings aligned to SSOT tone) 8. Prompt Pack: UI Copy + Wireframe Descriptions for Gemini (at least 10 prompts) 9. Acceptance Criteria 10. Assumptions 11. Open Questions Hard requirements: - Provide “Wireframe in Words” for each of the 6 key screens (layout described using rows/columns). - Include error states (no internet, corrupted save, invalid input) even if prototype won’t implement all—label as “Future”. 3) Art Direction & Asset Prompt Pod (Google AI tools) Create: outputs/agent_teams_demo/studio_in_a_box/03_Art_Direction_Assets.md (target 1,800–2,500 words) Required sections: 1. Art Direction Summary (≤150 words) 2. Style Keywords (exactly 20) + Negative Keywords (exactly 20) 3. Color Palette (6 colors with hex codes + usage guidance) 4. Asset List for Prototype (grouped by category; minimum 40 assets total) 5. Consistency Rules (10 rules: silhouette, line weight, lighting, perspective, etc.) 6. Prompt Templates (provide reusable templates with [PLACEHOLDERS]) - Character portrait - Character sprite (if applicable) - Enemy/creature - Environment tile/background - UI panel/button set - Icon set - VFX sheet (if applicable) 7. Batch Generation Plan (how to generate in coherent sets; naming conventions; review steps) 8. Asset QA Checklist (minimum 15 checks) 9. Prompt Pack: Google AI Tools (at least 12 prompts; specify which tool: Gemini vs other Google tool; if uncertain, write tool-agnostic “Google image model” prompts and include adaptation notes) 10. Acceptance Criteria 11. Assumptions 12. Open Questions Hard requirements: - Must align visuals to SSOT tone and avoid list. - Must include a “Style Lock” mini-procedure: generate 5 candidates → pick 1 → derive variants. 4) Implementation Blueprint Pod (Gemini) Create: outputs/agent_teams_demo/studio_in_a_box/04_Implementation_Blueprint.md (target 2,000–2,800 words) Required sections: 1. Build Strategy (for non-technical maker; options: “no-code-ish” vs “light-code”; but do not require paid tools) 2. Game Architecture (engine-agnostic) - State machine / scenes - Update loop concepts - Data-driven content approach 3. Data Schemas (provide 3 schemas in JSON examples and 1 in CSV example) 4. File/Folder Plan (VSCode-friendly; include assets, data, docs) 5. Step-by-Step Build Checklist (minimum 25 steps; each step ≤2 sentences) 6. “Glue Code” Guidance (conceptual; include pseudocode snippets ≤20 lines each; max 3 snippets) 7. GitHub Workflow for Solo Dev (branching optional; include commit message examples) 8. Prompt Pack: “Ask Gemini to Help Me Implement” (at least 12 prompts) 9. Acceptance Criteria 10. Assumptions 11. Open Questions Hard requirements: - Do not pick a specific engine. Provide a path that works whether the user chooses web (HTML/JS), Godot, or another free option—without requiring any of them now. - Include a “Fallback Prototype Mode” that can be built with a simple interactive mock (e.g., Twine-like structure described in markdown) if coding stalls. CROSS-POD INTEGRATION REQUIREMENTS - All pods must use the same: - Game name (working title) - Core loop steps - Resource names (currency/XP/etc.) - UI labels and terms - Art style keywords and avoid list - Mechanics tables must reference UI elements (e.g., what shows on HUD) and vice versa. - Implementation blueprint must explicitly map: - Mechanics systems → data schema → UI screens → required assets. POD EXECUTION INSTRUCTIONS (what each agent does) 1) Coordinator (Gemini) - Draft SSOT first (Phase 0). Make a clear Default Concept selection. - Send SSOT summary (≤200 words) plus “Non-negotiables” (10 bullets) to all pods inside SSOT under a subsection “Message to Pods”. - After receiving pod docs, produce Reconciliation Report and apply edits by instructing what changes should be made; ensure final alignment in Prototype Ready Pack. 2) Mechanics & Balance (Gemini) - Read SSOT. Do not invent new terms without adding to glossary proposals. - Produce numeric tables that are “starter balanced” and easy to tweak. - Send “Pod Findings” to Coordinator: 10 bullets including tuning risks and any SSOT gaps. 3) UI/UX & Tutorial (Gemini) - Read SSOT. Mirror tone and terminology precisely. - Ensure the tutorial teaches the core loop and highlights tuning levers in plain language. - Send “Pod Findings” to Coordinator: 10 bullets including microcopy tone conflicts and any missing screens implied by mechanics. 4) Art Direction & Asset Prompt (Google AI tools) - Read SSOT. Lock a style and operationalize it into prompts and QA. - Provide prompts that a beginner can paste with minimal edits. - Send “Pod Findings” to Coordinator: 10 bullets including any visual contradictions with UI needs (readability, icons, contrast). 5) Implementation Blueprint (Gemini) - Read SSOT. Keep plan engine-agnostic, but actionable in VSCode/GitHub. - Produce schemas that match mechanics templates and UI needs. - Send “Pod Findings” to Coordinator: 10 bullets including feasibility risks and suggested scope cuts. FINAL SYNTHESIS/REVIEW STEP (must be last, done by Coordinator) - Create outputs/agent_teams_demo/studio_in_a_box/98_Synthesis_Review.md (target 900–1,200 words) Required sections: 1. Completeness Checklist (verify every required file exists + meets word count targets) 2. Consistency Audit (at least 12 checks across mechanics/UI/art/implementation) 3. Top 10 Risks for a Non-Technical Maker + mitigations 4. “Start Here” Instructions (≤12 bullets, in order) 5. What to Ask Claude vs What to Ask Gemini (clear guidance; 8 bullets) QUALITY BAR - Every doc must be immediately usable without external context. - No placeholders like “TBD” except in the “Open Questions” sections. - Keep scope prototype-focused; avoid MMO-scale complexity. - Prefer concrete examples over theory. Now execute the team workflow: 1) Coordinator produces 00_SSOT.md first. 2) After SSOT exists, each pod produces its assigned doc in parallel and sends Pod Findings back. 3) Coordinator produces 05_Reconciliation_Report.md, applies consistency decisions, then produces 99_Prototype_Ready_Pack.md. 4) Coordinator finishes with 98_Synthesis_Review.md.
Create your own AI agent team at Build Agents Store. Describe your business problem and get specialized agent teams with ready-to-use prompts for Claude Code.