buildbetter-customer-issue-triage
BuildBetter Customer Issue Triage
Operating Contract
Treat CUS- as the primary signal that the source of truth is a Linear Customer Support issue. Start by reading the Linear issue, comments, attachments, linked Slack thread, customer, org, status, assignee, and any related issues or PRs before touching code or data.
Reconstruct the customer source material before deciding. Verify the underlying Slack thread, recording, transcript, email, attachment, or live product state instead of trusting a title, summary, or adjacent memory. If exact search fails, widen to participant, customer, time-window, and concept searches, then confirm with transcript or source context.
Build an evidence packet before acting. Include issue key/link, customer/org/user/recording identifiers, source thread or attachment, impact, timeline, live state, code paths, logs, related Linear/GitHub/Sentry coverage, diagnosis, proposed action, verification plan, customer-safe response, and remaining uncertainty.
Use explicit live-state surfaces for production work. For BuildBetter production data, prefer production Hasura, direct DB paths, CloudWatch, or kubectl --context production ... as appropriate. Keep environment, time window, query ids, and exact counts visible in the handoff.
Avoid duplicate tracking. Search exact issue keys, customer names, Sentry short IDs, SQL families, PR markers, and existing Linear issues before creating anything new. Update the existing issue or PR when the evidence belongs to an existing lane.
Resolve along the narrowest correct lane: answer/support explanation, customer/vendor comms, guarded data repair, code fix with tests, incident customer mapping, or coordinator/remote-thread handoff. Do not convert notes into transcripts, invent customer impact, or ship speculative fixes when the source evidence is weak.
For code changes, follow the repository AGENTS.md, docs/rules/codebase-standards.md, area-specific AGENTS.md, and repo PR workflow. Use focused tests that prove the touched behavior, and publish evidence-rich Linear/PR comments while keeping PR bodies concise.