diagnose

Guide a 6-phase workflow to reproduce, diagnose, and fix software bugs.

Updated May 28, 2026
One-click install
npx skills add https://github.com/kyo0221/claude-config --skill diagnose-kyo0221
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: diagnose
Source: https://github.com/kyo0221/claude-config/tree/main/skills/diagnose
Command: npx skills add https://github.com/kyo0221/claude-config --skill diagnose-kyo0221

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates unproductive, guesswork-driven debugging of complex, hard-to-reproduce software bugs and performance regressions that waste developer time and often lead to incomplete or incorrect fixes.

Core Features & Use Cases

  • Structured 6-phase workflow: Guides users through building a feedback loop, reproducing the bug, generating falsifiable hypotheses, instrumenting code, implementing fixes with regression tests, and performing post-mortem cleanup for consistent, repeatable debugging.
  • Reproduction loop guidance: Provides 10 ranked strategies to build fast, deterministic, agent-runnable pass/fail signals for even the most elusive non-deterministic bugs.
  • Use case: When a critical production API timeout regression appears with no clear root cause, this Skill guides you to build a repro harness, generate testable hypotheses, and implement a fix with a regression test to prevent recurrence.

Quick Start

Use the diagnose skill to apply the disciplined 6-phase workflow to debug the reported intermittent checkout failure that only occurs during peak traffic.

Frequently Asked Questions about diagnose

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

FAQPage Schema
What is the best way to debug hard-to-reproduce software bugs and performance regressions?

To debug hard-to-reproduce software bugs and performance regressions effectively, use a structured 6-phase workflow that builds deterministic reproduction loops, generates falsifiable hypotheses, instruments code, and implements regression tests to replace guesswork.

How do I build a deterministic reproduction loop for non-deterministic bugs?

Building a deterministic reproduction loop for non-deterministic bugs requires applying ranked strategies to create fast, agent-runnable pass/fail signals, ensuring consistent replication of runtime errors and unexpected application behavior before generating testable hypotheses.

How do I fix a performance regression in a production API with no clear root cause?

Fixing a performance regression in a production API without a clear root cause involves building a repro harness, generating falsifiable hypotheses, instrumenting code for targeted debugging, and implementing a fix alongside regression tests to prevent recurrence.

Does systematic hypothesis testing work for resolving intermittent application failures?

Systematic hypothesis testing works for resolving intermittent application failures by providing a disciplined workflow to generate falsifiable hypotheses and instrument code, eliminating inefficient guesswork when dealing with unexpected application behavior.

When should I not use guesswork-driven debugging for runtime errors?

You should not use guesswork-driven debugging for runtime errors when encountering complex failing tests or performance slowdowns that cannot be resolved via simple code inspection, as this often leads to incomplete fixes and wasted developer time.

Why do I need regression testing after fixing complex software bugs?

You need regression testing after fixing complex software bugs to validate the implemented fix, ensure the failing tests pass deterministically, and perform post-mortem analysis that prevents future occurrences of the same performance regressions.