maintainable-app

Build, edit, and review applications so common production failures recover automatically, degrade intentionally, and leave enough signal for one maintainer to understand what happened.

Passive Trigger

Load this skill in the background whenever a coding task affects a production-facing application, background worker, scheduled task, webhook, payment flow, import/export, notification, API integration, database state transition, cache, file upload, queue, cron, auth/session flow, observability, logging, tracing, alerting, or reliability behavior.

Keep it lightweight for small changes: run a quick failure-mode pass silently, then mention only the failsafes, edge cases, and observability decisions that change the implementation. Also load {{ skill:maintainable-code }} for source-code changes and {{ skill:maintainable-tests }} when adding or changing tests, if available.

Decision Tree

What are you building or changing?

• User-facing request, workflow, or state transition: Read references/principles.md. Define the failure states, retry safety, user-visible fallback, and recovery path before coding the happy path.

• Queue, background job, cron, webhook, import/export, email, payment capture, or notification: Read references/jobs-and-queues.md. Require idempotency, unique work identity, bounded concurrency, retry limits, dead-letter handling, stale-work recovery, and worker timeouts.