Zendesk to GitLab bug triage agent

Build me an agent that turns customer bug reports in Zendesk into well-written GitLab issues so engineering stops missing real bugs buried in our support queue.

Trigger: cron, every hour during business hours (let me configure the exact hours and timezone).

Each run, the agent should:

1) Use Zendesk Search Tickets to pull tickets created or updated since the last successful run that match a configurable set of tags or views (defaults to tags like bug, engineering, or escalation). Let me set the exact tags, views, or search query, plus the target GitLab project and labels, as inputs to the workflow.

2) For each candidate ticket, call Zendesk Show Ticket and List Ticket Comments to read the full thread, including the customer messages and any internal notes.

3) Decide whether it is a reproducible defect or a how-to / account / billing question. Skip the non-bugs and record a one-line reason in the run summary. Do not file issues for non-defects.

4) For real bugs, dedupe before filing: call GitLab List Project Issues on the configured project (state=opened, recent first) and compare titles and key symptoms to avoid creating duplicates of issues opened in the last few weeks. If a clear duplicate exists, skip the create and note the existing issue in the run log instead.

5) For new bugs, use GitLab Create an Issue with a clean, specific title and a structured body containing: Summary, Steps to reproduce, Expected vs Actual, Environment (browser/OS/version if known), Customer impact (who and how many), and a link back to the Zendesk ticket. Apply severity and area labels inferred from the ticket content from a label allowlist I provide (for example sev1/sev2/sev3 and area:billing, area:auth, area:api).

6) After the GitLab issue is created, call Zendesk Update Ticket to add an internal note (not a public reply) containing the GitLab issue URL and a one-line summary, so the support agent can see the engineering handoff happened.

7) For high-severity bugs (configurable threshold, default sev1 and sev2), also use Slack Send a Message to post in a configured engineering Slack channel with the customer impact in one line, the GitLab issue link, and the Zendesk ticket link.

Hard rules: never auto-reply to the customer on the Zendesk ticket. Never change ticket status, priority, assignee, or public-facing fields. Only add internal notes. Skip tickets that already have an internal note containing a GitLab issue URL so a re-run does not double-file.

At the end of each run, output a short summary listing tickets filed (with GitLab links), tickets skipped as not-a-bug (with reason), and tickets skipped as duplicates (with the existing GitLab link), so a human owner can spot-check.

Additional information

What does this prompt do?

Reads new and updated Zendesk tickets tagged as bugs and pulls the full conversation so nothing gets lost in translation.
Decides whether each ticket is a real defect or a how-to question, and skips the non-bugs so engineering only sees real work.
Files a clean GitLab issue with a clear title, a structured repro, expected vs actual behavior, severity labels, and a link back to the Zendesk ticket.
Checks recent open issues in GitLab before filing so you do not get five copies of the same bug.
Posts a heads-up in your engineering Slack channel for high-severity bugs and leaves an internal note on the Zendesk ticket with the GitLab link.

What do I need to use this?

A Zendesk account with permission to read tickets and add internal notes.
A GitLab account with access to the project where engineering bugs should land.
A Slack workspace with a channel where engineering wants high-severity pings.
An agreed set of Zendesk tags or views that mark a ticket as worth triaging, for example bug, engineering, or escalation.

How can I customize it?

Change which Zendesk tags or views the agent watches, so it scans only the queues your support team actually routes bugs into.
Pick the GitLab project and label set the agent should use, including your severity and area labels.
Set the Slack channel and the severity threshold that triggers a ping, so the channel stays signal heavy.
Adjust how often it runs, for example every 30 minutes during business hours or every 2 hours overnight.

Frequently asked questions

Will the agent ever reply directly to the customer?

No. It only adds a private internal note on the Zendesk ticket with the GitLab link. All customer-facing replies stay with your support team.

What stops it from filing the same bug twice?

Before creating a new GitLab issue, the agent scans recent open issues in the target project and compares titles and content. If it finds an obvious match, it skips the create step and leaves a note instead.

How does it decide what is a bug versus a how-to question?

It reads the full ticket thread, including the customer messages and any agent replies, and only files when there is a reproducible defect. Account questions, billing, and how-tos are skipped and logged.

Can I point it at a specific GitLab project or group?

Yes. You configure the GitLab project the agent should file into, plus the labels it can apply for severity and area. You can run a separate copy for each engineering team if you want different routing.

What does a high-severity ping in Slack look like?

A short message in your configured engineering channel with the customer impact in one line and a link straight to the new GitLab issue, so on-call can act without opening Zendesk.