debug

Start by capturing the exact symptom: what happens, what should happen, and how to reproduce it. Then form 2-3 competing hypotheses about the cause, gather evidence for each, and eliminate disproved theories until one root cause remains.

Reproduce the failure consistently first, then check recent changes with git log and git diff since regressions often come from recent commits. Form hypotheses about the cause and test each one against evidence rather than guessing at fixes.

State clearly that the issue is not reproducible and investigate environmental differences instead: configuration, data, dependencies, or runtime conditions. Without reproduction, you cannot verify that any fix actually works.

A wrong guess wastes more time than methodical investigation and can mask the real problem or introduce new bugs. Narrowing the space of possible causes with evidence ensures the fix addresses the actual root cause.

Trivial fixes like one-line changes or config updates do not need full SDLC ceremony with branching, planning, and PR review. Assess the blast radius of the change and let the user decide how much process the fix warrants.