prisma-pro-dev
Prisma ORM guidance — schema design, migrations, query optimization, relations, and production operations.
- License 许可
- MIT
- 最近更新
- 2026-11-07
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- prisma-pro-dev
- category
- development
- description
- Prisma ORM guidance — schema design, migrations, query optimization, relations, and production operations.
03 SKILL.md 正文
Overview
Prisma is a next-generation ORM for Node.js and TypeScript: you declare models in a human-readable schema file, and Prisma generates a fully typed client. The schema becomes the single source of truth for your data model, and the generated client makes invalid queries a compile error rather than a runtime surprise.
The tradeoffs are real: the schema DSL adds a layer between you and SQL, migrations need care on large tables, and the client hides query behavior you must still understand. This skill covers using Prisma well — schema design, the migration workflow, relation loading without N+1s, and production operations.
When to use
- Starting a new TypeScript project and choosing an ORM.
- Designing a Prisma schema (models, relations, indexes).
- Running migrations safely (dev workflow, production deploys).
- Fixing slow queries or N+1 problems in Prisma code.
- Modeling complex relations (many-to-many, self-relations, polymorphic patterns).
- Configuring Prisma for serverless or edge runtimes.
- Debugging Prisma Client errors and migration drift.
Core concepts
- Schema as source of truth.
schema.prismadeclares models, relations, indexes, and the datasource. The generated client mirrors it exactly — change the schema, regenerate, and TypeScript tells you what broke. - Migrations are versioned SQL.
prisma migrate devgenerates SQL migrations from schema diffs;prisma migrate deployapplies them in production. Never edit a migration that's been applied anywhere; always add a new one. includevsselect.includeloads relations (watch for over-fetching),selectpicks scalar fields. Combine them to fetch exactly what's needed — returning whole object graphs by default is how APIs get slow.- N+1 still applies. Prisma batches some relation loads, but nested
includechains and loops over queries still produce N+1 patterns. Inspect generated SQL with logging or the Prisma Studio/query insights before assuming efficiency. - Relation modes and referential actions.
onDelete: Cascade/SetNull/Restrictencode lifecycle rules in the schema. Choose deliberately — cascading deletes are convenient until they delete the wrong thing. - Indexes belong in the schema.
@@index,@@unique, and@@fulltextmap to real database indexes. Add them for every foreign key you filter/join on and everywhere/orderBypattern in hot queries. - Transactions.
prisma.$transaction([...])for interactive transactions; keep them short, touch rows in consistent order to avoid deadlocks, and never put network calls inside. - Raw SQL escape hatch.
queryRaw/executeRawfor what the fluent API can't express (complex aggregations, CTEs, window functions). Use typed variants so results stay type-safe. - Connection pooling. Prisma manages its own pool; in serverless, connections explode without an external pooler (PgBouncer) or Prisma Accelerate. Set
connection_limitdeliberately. - Multi-schema and multi-database. Prisma supports multiple schemas and databases with separate clients or datasource blocks — know the setup before committing to a topology.
- Middleware and extensions. Client extensions (
prisma.$extends) for soft deletes, computed fields, and query logging — cleaner than scattering logic across call sites. - Seeding.
prisma/seed.tswith idempotent seeds for dev/staging; never seed production with destructive scripts. - Introspection.
prisma db pullreverse-engineers an existing database into a schema — the pragmatic start for brownfield projects. - Prisma Studio. A GUI for browsing data during development. Useful for inspection; never a substitute for proper admin tooling or migrations.
Practical workflow
- Design the schema first. Model entities and relations on paper; encode referential actions, indexes, and uniqueness in the schema from the start.
model Order { id String @id @default(cuid()) status OrderStatus @default(PENDING) total Decimal @db.Decimal(10, 2) items OrderItem[] createdAt DateTime @default(now()) @@index([status, createdAt]) } - Migrate in dev, deploy in prod.
prisma migrate dev --name add_orderslocally; review the generated SQL; in CI/CD runprisma migrate deployas a separate release step — never from the app boot path.- For large tables: expand-contract migrations (add nullable → backfill → add constraint) to avoid long locks.
- Generate and use the typed client. One shared
PrismaClientinstance (singleton in dev to survive hot reload); import types from the generated client for DTOs. - Shape every query. Default to
selectfor scalars +includeonly for needed relations; paginate with cursor-based pagination for feeds.const orders = await prisma.order.findMany({ where: { status: "PAID" }, select: { id: true, total: true, createdAt: true, items: { select: { sku: true, qty: true } } }, orderBy: { createdAt: "desc" }, take: 20, }); - Wrap multi-write operations in transactions. Keep them short and ordered; handle serialization failures with retry at the call site.
await prisma.$transaction(async (tx) => { const order = await tx.order.create({ data: {...} }); await tx.inventory.updateMany({ where: {...}, data: { reserved: { increment: 1 } } }); }); - Add client extensions for cross-cutting logic. Soft-delete filters, audit fields, and query timing in one place instead of repeated inline code.
- Configure for your runtime. Serverless: external pooler or Accelerate,
connection_limit=1per instance; long-lived servers: tune pool size to workload; always set statement timeouts. - Observe. Enable query logging in dev; in production track slow queries via the database (pg_stat_statements), not the ORM — the DB sees the truth.
Common pitfalls
- N+1 via nested includes — deep
includechains generating dozens of queries; inspect SQL and restructure. - Editing applied migrations — corrupts every environment's migration history; always create a new migration.
- Running
migrate devin production — it can reset the database;deployis the only prod-safe command. - Connection exhaustion in serverless — each instance opening its own pool; use PgBouncer or Accelerate.
- Missing indexes on foreign keys — Prisma doesn't add them automatically; declare
@@indexfor join/filter columns. - Cascading deletes by default — convenient in dev, catastrophic with real data; choose referential actions per relation.
- Long interactive transactions — holding locks across awaits and network calls; keep transactions short and ordered.
- Returning full models from APIs — leaking internal fields and over-fetching; project with
selectat the boundary. - Multiple PrismaClient instances — hot-reload creating clients until the pool exhausts; singleton pattern in dev.
- Drift between schema and database — manual DB changes bypassing migrations;
prisma migrate diffto detect and reconcile. - No statement timeouts — one bad query holding connections; set timeouts at the database level.
- Seeding production destructively — seed scripts that truncate; make seeds idempotent and environment-gated.
04 同仓库更多 Skills
web-scrape
Turn web pages into LLM-ready markdown via the Firecrawl API (renders JS, anti-bot handling). Use when browser_open/fetch fails on JS-heavy pages, or when you need clean markdown/structured extraction from a site. Requires FIRECRAWL_API_KEY.
design-taste-frontend
tasteskill v2, the anti-slop frontend skill. Art direction for websites, distilled from a study of 247 curated gallery sites (One Page Love, Awwwards Sites of the Day). Deals researched directions (idea, composition, type, palette, photographs, section blueprints) from a library, enforces a strict no-slop house style with a real-browser inspector, and runs a look-and-fix loop on screenshots. Use this whenever you build, design, redesign, restyle, polish or review a landing page, marketing site, portfolio, product page, one-pager, hero or any public-facing web page, in any stack (HTML/CSS, React, Next.js, Vue, Svelte, Astro, Tailwind) — also when the user says a page looks generic, templated, "AI-made", sloppy, or asks to make it look designed, premium, award-level or less like every other site. Not for dashboards or app screens.
test-driven-development
Use when implementing any feature or bugfix, before writing implementation code
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes
requesting-code-review
Use when completing tasks, implementing major features, or before merging to verify work meets requirements
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).