bug-diagnosis_skill

Overview

This skill applies a scientific debugging methodology through conversational investigation. It guides a user from symptom observation to verified root cause and a minimal fix while keeping the user in control. Responses use progressive disclosure and report only actions that were actually performed.

How this skill works

I start by confirming the precise symptom and collecting context, then form testable hypotheses. I perform targeted, observable checks (logs, code, git history, runtime state), report findings as verifiable actions, and narrow the cause. When the root cause is identified I propose a minimal fix, run or suggest tests, and verify results only after you approve each step.

When to use it

• Investigating a reproducible or intermittent bug
• Tracing an error back through code or call paths
• Performing root cause analysis after a regression
• Designing tests or experiments to validate hypotheses
• Coaching developers through stepwise debugging decisions

Best practices

• Start with a concise symptom summary before deep dives
• State only what you actually inspected or ran (logs, files, commands)
• Form clear, prioritized hypotheses and test them one at a time
• Propose changes before applying them and request user approval
• Use smallest possible code changes and run focused tests to verify fixes

Example use cases

• I read a failing stack trace, located the function and proposed the minimal patch
• I compared dependency versions between environments and found a mismatch
• I used git log to find a suspect commit and suggested a bisect strategy
• I added a targeted log statement to expose a race condition and validated the timing
• I drafted a one-line test to reproduce the failure and verified it fails before fixing