Act as a senior engineer who runs production apps. Set up error monitoring for my {{tech_stack}} app with Sentry, or the tool already installed. The app handles {{data_to_store}}. Before changing anything, check what error handling and logging exist today. Capture: - Browser and server errors, unhandled promise rejections and failed API calls, tagged by environment and release - Source maps uploaded to the tool during the build, never served publicly - Error boundaries in the UI that show a friendly message with a retry, never a blank screen - Alerts for new issues and spikes, sent to one channel, with noise from bots and browser extensions filtered out Privacy, the hard part: - Default personal data collection off, including IP addresses and cookies - A scrubbing hook before sending that removes emails, phone numbers, tokens, passwords, card numbers and authorization headers from messages, breadcrumbs, URLs and request bodies - Query strings stripped, since they often carry reset tokens or emails - Users identified only by an internal random ID, never email or name - Session replay off, or with all text and inputs masked and media blocked - Data stored in the region our visitors need, kept 30 to 90 days - Server logs follow the same rules Test the scrubbing: trigger an error on a form holding a fake email and token, and confirm neither reaches the dashboard. Sampling: every error, but a low rate for performance traces, so we stay within the plan. Remind me to name the tool as a processor in the privacy policy. Finish with the files changed, a test error I can trigger, and the alert rules.