A data-model brief, target database, and one create/read/update journey.
One complete data journey that still works after a restart, with validation and ownership enforced on the server.
Create data, restart the app, and retrieve it as the correct user; run database-specific checks. The included SQLite checker does not test PostgreSQL.
An app that forgets its data on restart, or lets one user change another's records, is not ready to share. This skill guides your coding agent through adding persistence for one feature, from prerequisites to a checked result.
The agent models one feature first: entities, required fields, relationships and ownership rules, using your project's existing database stack. It generates and inspects a migration, tests it on a disposable database with representative data, and backs up existing data before a change that can remove or transform it. Server operations validate input and enforce ownership inside the database operation, with transactions for related writes and credentials kept out of browser code.
It finishes by proving persistence: create, read, update and delete through the app, restart and confirm the data remains, and confirm that invalid input and another user's record are rejected. If a migration fails, it inspects the current schema and migration history instead of erasing a populated database to make tests pass.
You need a running app, a database chosen for where it will be deployed, and a coding agent. Provider accounts, model weights, GPU compute and hosting are not included.
No reviews yet.
Sign in to rate this Vibe.
Get one working feature first, in a layout that can grow into shared packages without imports breaking later.