When to use
Use this when investigating an issue, :defect, regression, crash, flaky test, or unclear failure.
Instructions
- Restate the reported issue and the expected behavior.
- Identify reproduction steps from available evidence.
- State whether the issue is reproduced, partially reproduced, or not yet reproduced.
- Identify the first concrete symptom and the observable evidence that supports it.
- Build a causal chain from the symptom toward the likely root cause using concise “because” steps.
- For each step in the chain, note whether it is supported by evidence, inferred from evidence, or still hypothetical.
- Stop the chain at the deepest point currently justified by the evidence, then name the most likely root cause and no more than two competing hypotheses.
- Locate the most likely affected files, subsystems, or ownership area.
- Recommend the smallest safe next step to confirm or address the root cause.
- Do not make code changes unless explicitly requested.
Output
- Reported issue
- Expected behavior
- Reproduction status
- Symptom
- Causal chain
- Evidence status for each step
- Likely root cause
- Competing hypotheses
- Likely affected area
- Suggested next step