Create Issue: Let AI Detect Codebase Problems and Turn Them into Issues

Introduction

"This code looks problematic, but I don't have time to write an issue"—as development moves forward, improvements like these often get overlooked.

DevLoop Runner's Create Issue automatically analyzes your repository with AI and creates GitHub issues for the improvements it finds. It discovers problems and opportunities that humans tend to miss, supporting continuous quality improvement.

The Three Issue Run Modes

Create Issue is one of three modes available in Issue Run.

Mode	Description
Create Issue	Analyzes a repository and creates new issues automatically (this article)
Rewrite Issue	Rewrites an existing issue to improve clarity and structure
Close Issue	Automatically closes stale or outdated issues

The normal workflow with Dev Run is "Issue to PR", but Issue Run works in the reverse direction.

Dev Run Flow: Issue → AI Analysis → Implementation → PR

Create Issue Flow: Repository Analysis → Problem Detection → Automatic Issue Creation

This article focuses on Create Issue. For the other modes, see:

• Rewrite Issue → Rewrite Issue Guide
• Close Issue → Close Issue Guide

Three Categories Create Issue Detects

1. Bug Detection (Bug)

Detects potential bugs and problematic code patterns.

Detection examples:

• Missing null checks
• Lack of error handling
• Type inconsistencies
• Security concerns

2. Refactoring Candidates (Refactor)

Identifies refactoring opportunities that lead to improved code quality.

Detection examples:

• Consolidating duplicate code
• Splitting overly complex functions
• Improving naming conventions
• Removing dead code

3. Feature Suggestions (Enhancement)

Suggests ideas for improving existing functionality.

Detection examples:

• Performance optimization
• Usability improvements
• New feature proposals
• Accessibility improvements

How to Use

Basic Steps

1. Navigate to the Run page and select the "Issue Run" tab
2. Choose "Create Issue" from the mode selection
3. Enter the repository
   • Format: owner/repo (e.g., my-org/my-project)
4. Select a category
   • Bug Detection / Refactoring / Feature Suggestions
5. Set the generation limit
   • Specify 1–5 issues
6. Click "Start Issue Run"

Checking Results

After execution completes, you can check results on the job detail page:

• List of created issues
• Links to each issue
• Success/failure status

Effective Usage

Regular Code Health Checks

Run Create Issue weekly or monthly for continuous code quality monitoring.

Every Monday: Detect 5 issues with Bug category
1st of each month: Detect 3 issues with Refactor category

Pre-Release Quality Check

Run the Bug category before releases to discover potential problems early.

1. Run Create Issue (Bug) on the release branch
2. Address detected issues by priority
3. Fix important problems before release

Visualizing Technical Debt

Run the Refactor category to visualize technical debt as issues.

• Manage as backlog items
• Include in sprints for planned resolution

Combine with Rewrite Issue

Issues automatically generated by Create Issue can be further refined with Rewrite Issue, tailoring them to your project's conventions and format.

Things to Keep in Mind

AI Judgment Is Not Perfect

• False positives are possible
• Review created issues before taking action
• It's perfectly fine to close unnecessary issues

Set Appropriate Limits

• Creating too many issues at once makes management difficult
• Start with 1–3 and gradually increase

Repository Access Rights

• The selected GitHub account needs write permission
• repo scope is required for issue creation

Summary

• Create Issue uses AI to analyze your repository and automatically create GitHub issues for improvements
• Detects across three categories: Bug Detection, Refactoring, and Feature Suggestions
• Use regular runs and pre-release checks for continuous quality improvement
• Combine with Rewrite Issue to further enhance issue quality

Leverage Create Issue to efficiently discover improvements that might otherwise be overlooked.