When a support ticket turns out to be a bug, what reaches engineering is often a title and a link. The details an engineer needs are already in the ticket, but they are spread across the conversation, so the first thing anyone does is read the whole thread again.
Ask your agent to do that reading and write the result into the escalation task that already exists.
What the agent starts with
The agent has two cards on the same Timestripe board. In Queue there is ticket #4471, and its description is the entire conversation with the customer — six messages over two days, pasted in as they came. In Triage there is a goal someone opened while escalating the ticket, holding a title and a link and nothing else.
Everything useful is in the thread and nowhere else. The customer names her browser and operating system in the third message, gives her plan and version in the fifth, and support finds a temporary workaround in the last one. The cause turns out to be a filter that was active when she pressed export.
The prompt
Two of those instructions matter more than they appear to. Rewriting the existing goal instead of creating a new one keeps the escalation in one place, which is worth insisting on because someone has already linked to it. The last line is what stops a task from reading as finished when the thread never contained enough to reproduce anything.
What comes back
The goal in Triage on the Support board now holds a bug report. It is named after the behaviour that is broken rather than the ticket number, the environment is quoted from the customer rather than assumed, and the reproduction steps end at the filter, which is the detail that makes the bug reproducible at all. It also carries the number of other tickets on the board describing the same thing.
The workaround support had already found is in there too. That line decides how urgent this is, and until now it was visible only to the person who wrote it.
