prisma-patterns

Prisma Patterns is a reference skill that captures production-grade patterns and non-obvious traps for using the Prisma ORM in TypeScript backends. It solves the problem of subtle, hard-to-debug behaviors that experienced teams keep rediscovering — for example, that updateMany returns an affected-row count rather than the records themselves, that migrate dev can reset the database, that the @updatedAt field is skipped on bulk writes, that $transaction calls can time out, and that serverless deployments can exhaust the connection pool. It opens by reminding the reader to check their Prisma version, since the API surface has shifted across major releases.

The skill is documented guidance rather than executable tooling. It covers schema design (ID strategy trade-offs between cuid, uuid, and autoincrement; when to add indexes; declaring soft-delete columns upfront), the difference between include and select for controlling over-fetching, mapping entities to response DTOs to avoid leaking sensitive fields, choosing between array-form and interactive transactions, the PrismaClient singleton pattern to prevent connection exhaustion during hot reload, avoiding N+1 queries, and cursor-based pagination for feeds and large datasets. Guidance is illustrated with annotated good and bad TypeScript snippets.

It targets backend and full-stack TypeScript engineers who build on Prisma, particularly those deploying to serverless platforms such as Vercel, Lambda, or Cloudflare Workers, or implementing soft delete and multi-tenant row filtering. It is most useful when designing schemas, writing queries or transactions, planning migrations, or reviewing code for the traps it enumerates.