Systematic Debugging

Find and fix bugs by following a strict process: no fixes without root cause first.

Why this matters: Random fixes feel faster but often introduce new bugs and leave the original cause in place. A few minutes of real investigation—reproduce, trace, hypothesize—usually leads to one right fix instead of a long chain of patches. It works the same whether you’re in a Node app, a Python script, or a distributed system: understand, then change.

Core principle

No fix until the root cause is found: a fix aimed at a symptom leaves the cause in place and usually adds a new bug. Propose a fix only after Phase 1.

Redact before you show anything. This skill quotes commands, outputs and captured artifacts back to the user, and the transcript leaves the machine. Write <REDACTED> in place of every token, password, connection string, cookie and customer identifier; build loops against environment variables so the credential stays in the environment rather than in the command you paste; and from a captured artifact (a HAR file, a log dump, a request trace) quote only the lines carrying the signal, because auth headers ride in the rest. If the redacted output is genuinely not enough to diagnose the bug, say so and ask the user rather than pasting the raw capture.

When to Activate

• User reports a bug, error, or "it doesn't work"
• Test failures, build failures, or integration issues
• Unexpected behavior or performance problems
• User asks to "debug", "find the bug", or "why does X happen"