Systematic debugging procedure - build a feedback loop, reproduce and minimize, hypothesize, instrument, fix, and regression-proof. Use whenever anything is broken - failing tests, exceptions, wrong output, flaky behavior, performance regressions, "works on my machine", or production incidents. Also the operating manual for a dedicated debugger agent in a pipeline. Trigger on any error message, stack trace, or "why doesn't this work" - before proposing any fix.
