debug-agent  by millionco

AI agents debug code with runtime evidence and logs

Created 1 month ago
267 stars

Top 95.9% on SourcePulse

GitHubView on GitHub
Project Summary

Debugging skill for AI agents, debug-agent addresses the common issue of AI-driven debugging relying on guesswork rather than concrete runtime data. It enhances AI coding assistants like Claude Code, Codex, and Cursor by enabling them to debug with verifiable evidence. This provides users with more reliable bug fixes and a clearer understanding of the debugging process.

How It Works

The agent's core approach involves generating hypotheses about potential bugs. It then instruments the target code with lightweight NDJSON logs to capture runtime behavior. After prompting the user to reproduce the bug, the agent analyzes the collected logs to confirm or reject its hypotheses. Fixes are only applied when the agent achieves 100% confidence, supported by log evidence, and these fixes are subsequently verified.

Quick Start & Requirements

Installation is initiated with npx debug-agent@latest init, which integrates the skill into detected AI agents. Usage varies by agent: in Claude Code and Codex, run /debug-agent [describe your issue]; in Cursor, instruct the agent to "use the debug-agent skill" and describe the issue.

Highlighted Details

  • Leverages runtime evidence and NDJSON logging for precise debugging.
  • Systematically generates and validates hypotheses against captured log data.
  • Ensures fixes are applied with 100% confidence, backed by log evidence and verification.

Maintenance & Community

No specific details regarding contributors, sponsorships, community channels, or roadmaps were provided in the README.

Licensing & Compatibility

The project is licensed under the MIT license, which is permissive and generally allows for commercial use and modification, provided attribution is maintained.

Limitations & Caveats

The primary caveat lies in the "100% confidence" claim for fixes, which may present challenges in highly complex or intermittent bug scenarios. The agent's effectiveness is contingent on its ability to accurately instrument code and the user's capacity to reliably reproduce the bug for log generation.

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
1
Issues (30d)
1
Star History
38 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.