Debugging

Overview

Turn an unknown failure into a falsifiable cause before changing production code. Diagnosis-only requests stop at evidence; fix requests continue through the smallest proven repair.

When to Use

• Bug reports, incidents, failing tests, and unexpected behavior with an unknown cause
• Flaky, timing-sensitive, environment-specific, or multi-component failures
• Repeated speculative fixes or an unclear failing boundary

When Not to Use

• Use engineering when the cause and required change are understood
• Use test-design when the main deliverable is a test strategy, suite, or explicit TDD workflow
• Do not widen one failure investigation into an architecture audit without evidence