Agent skills for Prisma Next — a small set of SKILL.md files that teach an LLM agent how to operate Prisma Next end-to-end without re-deriving the API from documentation each time.
Edit your data contract. Prisma handles the rest.
Install the version that matches your Prisma Next version. Skills ship in lockstep with the Prisma Next CLI/runtime. If your project uses Prisma Next
0.8.0, install fromprisma/prisma-next#v0.8.0so the skill surface matches the runtime surface.
One package, eleven skills. Each skill is a SKILL.md with its own description field that an agent runtime matches against the user's prompt:
| Skill | Scope |
|---|---|
prisma-next |
Router — catches vague prompts and routes to a specific skill. |
prisma-next-quickstart |
Adoption: greenfield projects and brownfield databases. |
prisma-next-contract |
Contract authoring — PSL, TS builder, no-emit. |
prisma-next-migrations |
Migration authoring — db update, migration plan, data transforms. |
prisma-next-migration-review |
Deployment + concurrency — "what runs on merge?", diamond convergence. |
prisma-next-queries |
Queries — SQL DSL, Raw SQL, ORM client, TypedSQL. |
prisma-next-supabase |
Supabase — the extension pack, RLS policy authoring, role-bound runtime (asUser / asAnon / asServiceRole), auth.* admin reads. |
prisma-next-runtime |
Wiring db.ts — middleware, connection, environment. |
prisma-next-build |
Build-system / dev-server integration — Vite plugin today, Next.js / Webpack / esbuild / Rollup are gaps named instead of fabricated. |
prisma-next-debug |
Debugging — error envelopes, signal-routing to error-code references. |
prisma-next-feedback |
Hand a question or report off to the team — file a GitHub issue (bug or feature request), or route Q&A / design feedback / direct-team-contact (extension authors included) to the Prisma Discord. The canonical destination of every other skill's What PN doesn't do yet routing. |
Every skill follows the same shape: preamble + canonical mental-model headline, When to Use / When Not to Use, Key Concepts, Workflow, Common Pitfalls, What Prisma Next doesn't do yet, Reference Files, and Checklist.
The skill is normally installed for you by prisma-next init, which operates in the current working directory:
mkdir my-app && cd my-app
pnpm dlx prisma-next initTo install standalone (existing project or a new agent runtime added after init):
# `--all` installs every skill in the cluster (the cluster works as a unit —
# the router skill routes between the others) for every agent runtime the CLI
# detects on this machine, without prompting per-skill or per-agent.
pnpm dlx skills add prisma/prisma-next#v<your-prisma-next-version> --allThe skill is always installed at the project level — there is no host-wide / global install path. The cluster's surface (commands it references, exit codes it expects, capability claims it makes) tracks the project's @prisma-next/* version, and a global install would have to pick a single version for every project on the host. Pinning per-project keeps the skill, CLI, runtime, and extension packs coherent on every project the user works in.
If you have multiple agent runtimes installed and want the skill cluster active in only one of them, swap --all for --skill '*' -a <agent> (e.g. -a claude-code, -a cursor, -a codex). The skills CLI's --help lists the supported agent ids.
Prisma Next is in early access (0.x). Each skill carries a What Prisma Next doesn't do yet section that names features the framework doesn't implement (model validations, lifecycle callbacks, Studio, runtime-apply migrations, EXPLAIN, prepared statements, db.batch(), multi-database routing, Next.js plugin, …) along with the workaround and a route to the prisma-next-feedback skill so the request becomes a tracked issue instead of a one-line URL.
The pattern is deliberate: it gives the agent something concrete to say when a user asks about an unbuilt feature, instead of confabulating a plausible-looking API call against something that doesn't exist.
The skills source is versioned with the rest of Prisma Next. Keep the git ref aligned with your Prisma Next version (see the call-out at the top of this README).
Authoring rules, cluster conventions, and the worked example for concepts-over-procedures live in DEVELOPING.md. Read that before adding or rewriting a SKILL.md. Skill sources live in this skills/ directory in the prisma-next monorepo.
Apache-2.0.