db-intelligence

One skill, four engines. Contract: schema before SQL, evidence before claims, reads before writes. The output of this skill is not just query results — it is a domain evidence artifact other agents (specifier, coder, QA) can consume.

Step 1 — Detect the engine

Signal	Engine	Reference
`psql`, port 5432, `pg_catalog`, DATABASE_URL `postgres://`	PostgreSQL	`reference/postgres.md`
`mysql`, port 3306, JDBC `mysql:`, MyBatis/JPA against MySQL	MySQL	`reference/mysql.md`
a `.db`/`.sqlite`/`.sqlite3` file, `better-sqlite3`, embedded/local store	SQLite	`reference/sqlite.md`
`mongodb://` URI, collections, aggregation, `mongosh`	MongoDB	`reference/mongodb.md`

Can't tell? Look at the repo: ORM config (application.yml, mybatis, prisma/schema.prisma, knexfile, mongoose), docker-compose services, and .env* keys name the engine. Multiple engines in one task are normal — open each engine's reference and keep one evidence artifact.