debug-mode

Run parallel AI tracks to debug flaky tests and CI failures.

4|Updated Dec 12, 2025
One-click install
npx skills add https://github.com/nicobailon/debug-mode --skill debug-mode-nicobailon
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: debug-mode
Source: https://github.com/nicobailon/debug-mode/tree/main
Command: npx skills add https://github.com/nicobailon/debug-mode --skill debug-mode-nicobailon

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Debug Mode enables hypothesis-driven debugging by running two autonomous AI tracks in parallel, each reviewing and refining its own fixes, then synthesizing their findings for high-confidence changes.

Core Features & Use Cases

  • Dual-track debugging workflow that leverages Opus 4.5 and GPT 5.2 to explore complementary fixes.
  • Context-building, reproduction strategy, and patch coordination across tracks.
  • Use Case: When standard debugging fails or flaky CI tests occur, spin up parallel AI tracks to converge on a fix quickly.

Quick Start

Install prerequisites (bun, tmux, Codex CLI), initialize the debug-session with debug-mode init /path/to/project, run track iterations with debug-mode codex run track-a 1 /tmp/debug-track-a-prompt.md and debug-mode codex run track-b 1 /tmp/debug-track-b-prompt.md, monitor status with debug-mode status track-a and debug-mode status track-b, compare changes with debug-mode diff track-a / debug-mode diff track-b, and finally apply with debug-mode apply track-a /path/to/project.

Frequently Asked Questions about debug-mode

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I debug flaky CI tests that are hard to reproduce?

Flaky CI tests can be debugged by running two parallel AI tracks that independently review and refine their own fixes, then synthesize findings for high-confidence changes. This dual-track approach converges on a fix quickly when standard debugging fails.

What is dual-track AI debugging and how does it work?

Dual-track AI debugging runs two autonomous AI tracks in parallel using Opus 4.5 and GPT 5.2 to explore complementary fixes. Each track independently reviews and improves its own work, then synthesizes findings to converge on a high-confidence fix.

Do I need git worktrees to run parallel debugging sessions?

Yes, isolated git worktrees are required. The dual-track AI debugging workflow uses isolated worktrees to allow each track to independently build context, reproduce the issue, and coordinate patches without interfering with the other track.

How do I set up and run a dual-track debugging session for E2E failures?

Initialize the session with debug-mode init, run track iterations using debug-mode codex run for both track-a and track-b, monitor status with debug-mode status, compare changes with debug-mode diff, and apply the chosen fix with debug-mode apply.

What prerequisites do I need to use dual-track AI debugging?

You need bun, tmux, and Codex CLI installed, along with Opus 4.5 and GPT 5.2 compatible tooling. These tools enable the CLI to initialize, run, poll, diff, apply, and clean up isolated parallel debugging tracks.

When should I use parallel AI tracks instead of standard debugging?

Parallel AI tracks are best for complex debugging scenarios across large codebases, specifically when standard debugging fails or when dealing with flaky tests and CI/E2E failures that are hard to reproduce consistently.