Challenge

Multi-Agent Code Review & Refactoring

This challenge focuses on building an advanced multi-agent system using the OpenAI Agents SDK. The system will be designed to automate code review processes, identify potential bugs or inefficiencies in a given codebase, and suggest intelligent refactoring strategies. It will leverage the o4-mini model for its strong code understanding and generation capabilities, enabling nuanced analysis and creative solutions. The solution will incorporate Kiln AI for robust agent management and lifecycle, ensuring the agents operate reliably and can be scaled. Composio will be used for integrating various external developer tools, such as code analysis suites and version control systems, allowing agents to interact with real-world development environments. Metaflow will orchestrate the complex CI/CD workflow, from code ingestion to analysis, refactoring suggestions, and simulated integration. Optionally, Synthflow can be used to add a voice-based interaction layer for developers to query code status or request refactorings verbally. This project demonstrates cutting-edge multi-agent orchestration for significantly enhancing software development productivity and quality.

Special Purpose AgentsHosted by Vera
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

This challenge focuses on building an advanced multi-agent system using the OpenAI Agents SDK. The system will be designed to automate code review processes, identify potential bugs or inefficiencies in a given codebase, and suggest intelligent refactoring strategies. It will leverage the o4-mini model for its strong code understanding and generation capabilities, enabling nuanced analysis and creative solutions. The solution will incorporate Kiln AI for robust agent management and lifecycle, ensuring the agents operate reliably and can be scaled. Composio will be used for integrating various external developer tools, such as code analysis suites and version control systems, allowing agents to interact with real-world development environments. Metaflow will orchestrate the complex CI/CD workflow, from code ingestion to analysis, refactoring suggestions, and simulated integration. Optionally, Synthflow can be used to add a voice-based interaction layer for developers to query code status or request refactorings verbally. This project demonstrates cutting-edge multi-agent orchestration for significantly enhancing software development productivity and quality.

Datasets

Shared data for this challenge

Review public datasets and any private uploads tied to your build.

Loading datasets...
Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
5 scoring checks
Binary
5 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

CorrectIssueIdentification

Checks if the agent correctly identified all expected issues from the input.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

ValidRefactoringSuggestions

Checks if refactoring suggestions are well-formed, relevant, and provide actionable advice.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 3

MockPRDescriptionPresent

Verifies that a simulated Pull Request description is generated.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 4

CodeQualityImprovementScore

A score indicating the comprehensiveness, accuracy, and impact of the suggested refactorings (0-100). • target: 85 • range: 0-100

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 5

AgentProcessingLatencyMS

Average time taken by the agent system to process a code review request, in milliseconds. • target: 2000 • range: 100-6000

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Master the OpenAI Agents SDK for defining agent roles, capabilities, and tool-calling functions for structured interactions.

  • Implement advanced prompting techniques with o4-mini for sophisticated code understanding, vulnerability detection, and transformation tasks.

  • Design and manage complex agent workflows using Kiln AI for scalable, observable, and resilient multi-agent deployments.

  • Integrate Composio to provide agents with programmatic access to Git repositories, linters, testing frameworks, and other developer tools.

  • Orchestrate complex AI-driven CI/CD pipelines using Metaflow for automated code quality gates, compliance checks, and deployment simulations.

  • Build robust error handling and feedback mechanisms within the agent system for continuous improvement and developer collaboration.

How this agent runs

The solution will be evaluated on its ability to autonomously review a provided Python codebase, accurately identify issues, suggest correct and actionable refactorings, and demonstrate conceptual integration with a m...

Preview configuration

Challenge input

{"file_path": "str", "code_content": "str", "expected_issues": ["str"]}

OpenAI

OpenAI AI model provider

Composio

Tool integrations for AI agents

o4-mini

Policy Serving in the agent workflow.

Evaluated output

{"identified_issues": ["str"], "refactoring_suggestions": "str", "mock_pr_description": "str"}

Checks for
  • Checks if the agent correctly identified all expected issues from the input.
  • Checks if refactoring suggestions are well-formed, relevant, and provide actionable advice.
  • Verifies that a simulated Pull Request description is generated.
Proof of success
  • CodeQualityImprovementScore target: 85
  • 1 public reference case
Runtime evidence
  • Python execution harness
View technical recipe

Configured tools

Action Space
  • OpenAI · Optional
  • Composio · Optional
Policy Serving
  • o4-mini · Required

Evaluation contract

  • CorrectIssueIdentification · Weight 1
  • ValidRefactoringSuggestions · Weight 1
  • MockPRDescriptionPresent · Weight 1
  • CodeQualityImprovementScore · Weight 1
  • AgentProcessingLatencyMS · Weight 1

Recipe state

This is a preview. The configuration can change before the evaluation recipe is locked.

Run this agent on your dataset and AI stack

Bring your dataset, model providers, and success criteria. We will scope the right managed run for your team.

Scope a managed run
Start from your terminal
$npx -y @versalist/cli start multi-agent-code-review-refactoring

[ok] Wrote CHALLENGE.md

[ok] Wrote .versalist.json

[ok] Wrote eval/examples.json

Requires VERSALIST_API_KEY. Works with any MCP-aware editor.

Docs
Manage API keys
Explore

Find another challenge

Jump to a random challenge when you want a fresh benchmark or a different problem space.

Useful when you want to pressure-test your workflow on a new dataset, new constraints, or a new evaluation rubric.

Frequently Asked Questions about Multi-Agent Code Review & Refactoring