HiAi-gg Investigate a Regression
Use this skill when an error spikes, reappears, or starts at a clear point in time — the classic regression signal.
Workflow
1. Establish the timeline
- First seen and the volume curve (when did it start / spike?).
- Align that timestamp with the release cadence: which release/deploy landed closest to the spike?
2. Identify the release
- Fetch the releases around the spike (before/at/after).
- For the candidate release, list the commits included (or the PRs merged).
3. Correlate the change with the error
- Compare the issue's culprit (file/function) with the files touched in the candidate release's commits.
- A matching file/function is strong evidence; also check for dependency, config, or SDK version changes in the same window.
4. Narrow further
- If multiple candidates: compare event attributes before vs after the spike (exception message, stack frames, affected users) — a changed stack frame or message points at the new code path.
- Check if the same issue exists on an older release line (did it never happen before, or was it always broken?).
5. Report
State: the spike timeline, the candidate release(s), the commit/change most likely responsible (with evidence), and the recommended verification (reproduce, roll back, or fix forward). Mutating issue/release state requires explicit user intent.