mongoose-to-prisma

Invoke the skill with the path to one Mongoose model file, such as /mongoose-to-prisma apps/app/src/server/models/tag.ts. It analyzes the model, verifies schema.prisma, generates a Prisma extension, updates callers, and cleans up Mongoose code, asking for your approval before each write.

Statics become methods inside a Prisma.defineExtension model block, using context.findMany, context.findUnique, and context.count. Virtuals become result compute aliases whose needs list every scalar field the getter reads.

No, Prisma does not support sparse indexes. The skill flags any sparse index and proposes a migrate-mongo migration using the native MongoDB driver's collection.createIndex with the sparse option, plus a comment in schema.prisma.

Mongoose's autoIndex still creates the model's indexes on connect, so the schema block is kept until all models have migrated. Deleting it early requires an explicit migrate-mongo migration to recreate the indexes, or they silently stop existing.

Decide by checking whether the referenced model has a hard-delete path that leaves orphans, not by copying Mongoose's required flag. If hard deletes can leave dangling references, make the relation optional to avoid runtime 'Inconsistent query result' errors.

It handles only one model file per invocation, never modifies frontend code, and never runs prisma db push or prisma generate. Factory-pattern models that depend on the crowi instance require a case-by-case design decision with the developer.