Customer bug reproduction from error reports

[maxfindel]

Summary

Given a support ticket or error report (Bugsnag, Sentry, customer email), FlowChad attempts to reproduce the bug in-browser. Produces screen recording + exact steps-to-reproduce attached to the issue.

How it works

• Input: GitHub issue with error details (stack trace, user action, URL)
• FlowChad reads the issue, infers a reproduction flow
• Walks the flow in-browser, capturing screenshots + video at each step
• If reproduction succeeds: attaches evidence, adds reproduced label
• If reproduction fails: adds could-not-reproduce label with what was tried
• Either way: structured steps-to-reproduce added to the issue body

Maps to

• Nader's SDLC list suggest skill — AI-powered improvement prioritization #9: "Reproducing customer-reported bugs from support tickets"
• Existing speckit flow: bugsnag issues already land as GitHub issues — this adds the reproduction step before speckit attempts a fix

Open questions

• Trigger: manual dispatch, or auto on bugsnag label?
• Browser context: needs auth? Test user credentials?
• Scope: frontend-only bugs, or can it test API responses too?

🤖 Generated with Claude Code