This is a workflow for actually fixing Sentry errors instead of just closing them. It walks Claude through three phases: triage (using Sentry MCP to find and classify issues), investigation (pulling event details, cross-referencing with Axiom logs by traceId, tracing upstream inputs), and fixing with tests. The philosophy is pretty clear: treat errors as signals pointing to root causes, not noise to suppress. It explicitly bans common shortcuts like batch-ignoring issues, downgrading log levels without investigation, or treating fallbacks as acceptable. You'd use this when production errors pile up and you need systematic debugging rather than whack-a-mole fixes. Requires the Sentry MCP server already configured.
npx skills add https://github.com/brianlovin/claude-config --skill fix-sentry-issues