Build Autonomous Game Development Teams with AutoGen and Gemini 3.1 Pro
Develop an autonomous multi-agent swarm capable of designing and prototyping simple mobile games to ride the current vibe coding wave. Using AutoGen, you will orchestrate specialized agents that manage game logic, asset generation, and codebase refinement. These agents will leverage Gemini 3.1 Pro for high-speed generation tasks and utilize E2B for sandboxed code execution and testing. You will also integrate Upstage for structured data preparation and use StackAI to automate the orchestration of the agent lifecycle within a web-based dashboard. The challenge requires setting up a team of autonomous agents that interact through conversational interfaces, providing feedback loops for iterative game development. By automating the game creation process, you will demonstrate how modern agents can replace manual effort in rapid prototyping environments. All interactions must be governed by AutoGen agent proxy patterns to ensure stability during the build process.
What you are building
The core problem, expected build, and operating context for this challenge.
Orchestrate an autonomous multi-agent swarm using AutoGen and Gemini 3.1 Pro to rapidly prototype and test mobile games.
How work is evaluated
Validation of agent autonomy and successful code execution in sandboxed environments.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Execution Check
Verify code runs without errors
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Generation Speed
Tokens per second for game code • target: 150 • range: 50-200
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Configure AutoGen multi-agent orchestrations with specialized coding agents
Implement E2B SDK integration to allow agents to run Python games in secure containers
Apply Upstage data preparation techniques to refine game design parameters
Construct automated workflows in StackAI to trigger agent swarms from external inputs
Design Ellipsis-based feedback channels for human-in-the-loop review of AI-generated assets
Master inter-agent messaging protocols using AutoGen group chat manager
Reference links and supporting material
Sample data for 1 tasks
Challenge setup
Validation of agent autonomy and successful code execution in sandboxed environments.
Challenge input
Game theme description
AutoGen
Agent Systems · Multi-Agent Systems
Google Gemini
Google's multimodal AI model
StackAI
No-code AI workflow builder
Evaluated output
Python code snippet
- Verify code runs without errors
- Tokens per second for game code • target: 150 • range: 50-200
- Generation Speed target: 150
- 1 public reference case
- Docker execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- AutoGen · Required
- Google Gemini · Optional
- StackAI · Optional
Evaluation contract
- Execution Check · Weight 1
- Generation Speed · Weight 1
Metrics
- Generation Speed · Target 150
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: Game theme description
Output: Python code snippet
Execution harness: Docker
Recipe freeze marker
No recipe freeze time is recorded.