Working playbook
Bug Reproduction
Turn a report into a reproduction pack. Stay out of production customer data.
Updated August 27, 2026
A Bot that can operate a browser is unusually good at this job — and unusually dangerous if it logs into production as a customer.
What this Bot owns
Turning reports into reliable reproduction packs.
Connect: Issue tracker, staging environment, browser, and network tools.
If no structured plugin exists, the Bot can still operate the website from the shared cloud computer. Prefer a plugin when one exists; use computer use for everything else.
First task
Run this once. Do not schedule it or enable send access until the result is something you would actually review.
Copy-ready first task
Read this bug report and reproduce it in staging using a fresh test account. Return exact steps, expected and actual behavior, screenshots, browser and OS details, relevant console or network notes, and a minimal test case if possible. Do not use production customer data.
How to inspect the result
Follow the steps yourself once. Confirm screenshots match staging, not production. Credentials should have arrived through a secure handoff, not chat.
Where it must stop
Production logins, customer data, and ticket status changes that close the issue wait for you.
- Do not use production customer data.
- Do not paste passwords or one-time codes into chat.
- Do not mark the ticket fixed.
Failure rule: If it does not reproduce, say so and list what you tried. Do not file a guessed root cause as fact.
When to make it a routine
Useful as an on-demand specialist, not a scheduled closer of tickets. Pair with an engineering Bot only after the pack is accepted.
You can also ask the Bot to follow along once, then save the path as a routine. See teach a Bot the job.
Bot profile
Put this in the Bot profile. Keep this week’s list, dates, and files in the conversation.
Profile template
Name: Bug Reproduction Job: Staging reproduction packs from written reports Sources: issue tracker, staging, browser and network tools Working style: exact steps; expected vs actual; screenshots; no guessed fixes Approval boundary: never use production customer data or close tickets Failure rule: report non-repro instead of inventing a cause
Use the bot roster builder to generate a paste-ready version, or return to the playbook hub.
Sources: xAI use cases · approvals · how Bots work