database-pimp

Active router for ALL database and ORM requests — classifies by database type, use case, or ORM preference and routes to the correct skill. Use when choosing a database, setting up an ORM, or working with any data layer.

Model	Source
sonnet	pack: database

Full Reference

If the request involves databases, data layer, ORM, caching, vector search, or storage in ANY way — choosing a database, schema design, queries, migrations, connections, seeding, RLS, realtime, embeddings, edge data, or anything data-related — you MUST route through this skill FIRST.

This is not optional. This is not negotiable. You cannot skip this.

Database Pimp

The orchestration layer for all database expertise. Not documentation — an active router. Every database request flows through this routing table before any response.

Mandatory Announcement — FIRST OUTPUT before anything else:

┏━ 🗄 database-pimp ━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ [one-line description of request/routing]       ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛

No exceptions. Box frame first, then route.

Quick Context

The database pack covers 6 databases + 2 ORMs — serverless Postgres, BaaS, document, cache, vector, and edge storage. Pair with orm pack (Drizzle/Prisma) for type-safe access.

Routing Table

Classify the request. Invoke the matching skill. No response before invocation.

Request Pattern	Skill
Serverless Postgres, Neon, branching, connection pooling	neon
Supabase, RLS, realtime subscriptions, storage buckets	supabase
MongoDB, document modeling, aggregation pipelines	mongodb
Redis, Upstash, caching, rate limiting, pub/sub	redis-upstash
Vector search, embeddings, pgvector, similarity queries	postgresql-pgvector
Edge database, SQLite, libSQL, Turso, local-first, vector search in SQLite, AI agent databases, embedded replicas, branching preview DBs, Turso CLI, Turso pricing	turso
Drizzle ORM, type-safe queries, push migrations	drizzle
Prisma ORM, schema modeling, prisma migrate	prisma
”Which database should I use?”	Decision matrix below

Decision Matrix

When the user asks which database to use, route through this before any skill invocation:

Use Case	Recommendation
Serverless Postgres, new project, Vercel/Cloudflare	Neon — branch-per-PR, autoscale to zero
Postgres + auth + storage + realtime in one	Supabase — BaaS, RLS, storage, realtime
Document data, flexible schema, nested objects	MongoDB — document modeling, aggregation
Caching, sessions, rate limiting, queues	Redis/Upstash — serverless Redis, global edge
Semantic search, AI embeddings, similarity	pgvector — postgres extension, vector indexing
Edge runtime, SQLite, offline-first, AI agent DBs, vector search in SQLite	Turso — libSQL, distributed edge, embedded replicas, native vector search, per-agent DBs
New project, type-safe schema, push workflow	Drizzle — lightweight, SQL-like, no shadow DB
Existing Prisma schema, complex relations	Prisma — mature, rich ecosystem, generate types

Neon vs Supabase: Both are serverless Postgres. Choose Neon for pure database needs (branching, pooling). Choose Supabase when you also need auth, storage, or realtime out of the box.

Drizzle vs Prisma: New projects → Drizzle (faster, lighter, push-based). Existing Prisma schema → stay on Prisma.

State Detection

Before routing, check project state:

• stack.json → read database and orm keys, route directly
• package.json deps → detect installed DB clients, ORM packages
• drizzle.config.* → Drizzle is in use
• schema.prisma → Prisma is in use
• .env / .env.example → check for DATABASE_URL, SUPABASE_URL, TURSO_*, UPSTASH_*

State	Recommendation
No database detected	Ask use case → run decision matrix
DATABASE_URL with neondb.net	Route to neon
SUPABASE_URL present	Route to supabase
TURSO_DATABASE_URL present	Route to turso
UPSTASH_REDIS_REST_URL present	Route to redis-upstash
drizzle.config.* present	Route to drizzle
schema.prisma present	Route to prisma

Chaining Patterns

User Says	Chain
”Set up my database”	Decision matrix → DB skill → ORM skill
”Add caching to my app”	redis-upstash
”Add search to my app”	postgresql-pgvector (semantic) or mongodb (text)
“Set up auth + database”	supabase (handles both)
“I need a type-safe ORM”	drizzle (new) or prisma (existing schema)
“Set up database for edge”	turso → drizzle (libSQL adapter)
“Add vector search to SQLite”	turso (native — no pgvector needed)
“Build a RAG app with SQLite”	turso (vector search + AI embeddings)
“Per-agent or per-tenant database”	turso (Platform API, one DB per agent)

Hard Rules

• Never respond about databases before invoking the target skill
• No summarizing, planning to invoke, or explaining what you’re about to do
• If unclear which database, run the decision matrix — ask ONE question if needed
• The skill’s content has the verified facts — always defer to it
• ORM questions that are DB-agnostic → route to drizzle or prisma directly
• Never recommend a database that conflicts with stack.json without flagging it