Start with the engineering issue and its linked support ticket. Review the symptom, affected version, reproduction, and evidence before deciding on a fix.
- Open Investigations and start with the engineering issue URL and original support ticket URL, or paste support’s reviewed handoff. Include any new reproduction or customer facts recorded on the engineering issue after the automated report was created.
- Select the affected codebase branch. Ask: “Trace this reproduction through the cited code. Which condition explains the failure, and what evidence would distinguish it from the alternative in the handoff?”
- Give Driver the reproduction result from your development environment. Ask which callers, configuration branches, and tests a fix must cover. Compare the answer with the actual reproduction before choosing a change.
- If you use a coding assistant connected through Driver MCP, ask the assistant to retrieve the cited source on that branch. With prepared Advanced context, ask it to use
get_relatedandchange_impacton the relevant file or symbol to identify dependents and candidate tests. Supply your local diff for changes that have not reached the connected branch. - Implement and test through your normal development workflow. Reference the engineering issue in the pull request, then verify the fix and report its rollout state to support.
You can also open Continue investigation in Driver on the original automated report and send a message to start a private continuation. Add the engineering handoff and later Jira facts in that message: the shared starting snapshot does not include support’s private follow-ups. Your own private thread also reaches support through the results you copy to Jira.