A customer asking how something works gets a draft from Cursor. You still send it.
A customer with a stack trace gets the bug reproduced and filed. They hear back from you.
{
"agent": "Claude Code",
"job": "Reproduce the bug, file or attach a Linear issue, and leave a private note. Do not reply.",
"contact": "${contact_name}",
"subject": "${subject}",
"conversation": "${url}"
}Fernand fills in ${contact_name}, ${subject}, and ${url} when this runs.
AI runs one is-about rule per email. If you already have a rule whose phrase is "bug report", this one and that one compete. Keep this phrase for bugs a coding agent should reproduce, and turn the other off if Linear is getting two issues.
The key needs read, notes, and Linear. It does not need reply. Claude Code reads the Fernand API docs, searches for an existing issue, then attaches or creates one.