Frame the problem. Record the observed and expected behavior, scope, impact, environment, and when it began. Keep observations separate from interpretations.

Locate the divergence. Gather the smallest relevant evidence and trace the behavior toward the earliest point where reality differs from expectation. Preserve useful inputs, versions, timestamps, and environment details.

Reproduce when safe. Find the smallest faithful check. For intermittent problems, repeat under controlled conditions and vary one factor at a time. If reproduction is not possible, state what the available evidence can and cannot establish.

Test competing explanations. Keep only plausible hypotheses. For each, identify an observation it predicts; choose the next check for its ability to distinguish explanations, and look for evidence that could disprove the leading one.

Establish the causal mechanism. Trace the evidence from symptom to trigger to responsible boundary and mechanism. Challenge the strongest alternative. Do not treat correlation or proximity as proof of cause.

Report the result. State whether the cause is confirmed, likely, or unresolved; give the supporting evidence, material alternatives considered, checks performed, and remaining uncertainty. Include the smallest supported fix and how to verify it. If a fix is requested, use these findings as input to a separate implementation step following the project's conventions; this skill does not require another skill to be installed.