Section 19 · Error Tracking & Reliability
Error Reporting
Error tracking, configuration, deployment integration, and AI-driven error handling
This guide walks you through auditing a project's error reporting capabilities, covering Sentry (or similar) setup, configuration, deployment integration, and AI-driven error handling.
The Goal: Errors That Fix Themselves
Production errors should flow from detection to diagnosis to fix with minimal friction. A mature error pipeline means every error is tracked, readable, correlated to deploys, and triaged for action.
- Captured — error tracking tool installed, initialized early, and properly configured
- Privacy-aware — PII handling intentionally configured with scrubbing or documented justification
- Readable — stack traces and source maps show original code, not minified output
- Correlated — deployment integration ties errors to specific releases
- Actionable — AI-driven triage, prioritization, and auto-PR creation for fixes
Before You Start
- Identify the error tracking tool (Sentry, Bugsnag, Rollbar, Raygun, etc.)
- Get access to the error tracking dashboard
- Identify frontend vs backend error tracking (may be separate)
Setup
Sentry or similar error tracking SDK installed, initialized early, and DSN configured
“How many errors happened in production in the last 24 hours?”
Personal data handling is intentional - either scrubbed or enabled with documented rationale
“Are user passwords or tokens ending up in your error logs?”
Stack traces enabled and useful; performance tracing configured with appropriate sample rate
“Do your stack traces point to real source lines or minified junk?”
Source maps generated, uploaded to Sentry via CI/CD, stack traces show original source
“What line of code caused last week's frontend error?”
Releases created in Sentry, deploy notifications sent, errors filterable by release
“Did errors spike after your last deploy, and did you catch it?”
AI-Driven Error Handling
Scheduled AI/Claude review of errors with actionable output (Slack, issues, or PRs)
“When did a human last actually review your error inbox?”
AI identifies fixable errors and creates labeled PRs requiring human review
“How many fixable errors have sat in Sentry for months?”
Errors prioritized by impact, routed to owners, known errors filtered, issue tracker integration
“Who owns the error your most important user hit yesterday?”