dart-drift — Claude Code skill: Drift persistence in Dart
A skill for Drift persistence in Dart CLI, server-side and non-Flutter desktop apps: tables, queries, migrations, PostgreSQL.
npx skills add MADTeacher/mad-agents-skills --skill dart-drift -a claude-codeWhat dart-drift does
The skill casts the agent as a Dart persistence engineer for non-Flutter apps on Drift. Before writing any database code it makes the agent pick a backend: SQLite through package:drift/native.dart for CLI tools, server jobs and local desktop storage, or PostgreSQL through drift_postgres and package:postgres with the Postgres dialect enabled in build.yaml. If the task turns out to need Flutter UI integration, the skill stops and hands off to a separate flutter-drift skill.
Inside there is task routing. Separate reference files cover setup, tables, queries, writes, streams, migrations and PostgreSQL, and the agent reads only the ones the current operation needs. Validation is mandatory: dart pub get after dependency changes, dart run build_runner build and dart analyze after touching tables or DAOs, dart run drift_dev make-migrations plus the generated migration tests after a schema change.
The constraints are strict. Deprecated PostgreSQL examples built on HostEndpoint, PgEndpoint, postgres_pool.dart, postgresUuid(), postgresJson() and gen_random_uuid are off the table. On Postgres, SQLite-only helpers such as currentDateAndTime and most dateTime() conveniences are banned in favour of PgTypes.date, PgTypes.timestampWithTimezone, PgTypes.timestampNoTimezone and now(). If the repository does not make the target backend clear, the agent asks instead of guessing.
What it helps with
- add SQLite to a Dart CLI with
package:drift/native.dart - wire up PostgreSQL with
drift_postgresandpackage:postgres - write type-safe Drift tables and queries
- fix a failing
build_runnerordrift_devrun - run a schema migration and its generated tests
Example prompts for the agent
- add Drift with SQLite to my Dart CLI, start with the dependencies
- define an orders table with a date index and write a query for one month
- the schema changed — make a migration and run the migration tests
How to install dart-drift
1Skills CLI (Vercel)
npx skills add MADTeacher/mad-agents-skills --skill dart-drift -a claude-code2Manually: put the files in place
~/.claude/skills/dart-drift/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/dart-drift/
1Skills CLI (Vercel)
npx skills add MADTeacher/mad-agents-skills --skill dart-drift -a codex2Manually: put the files in place
.agents/skills/dart-drift/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add MADTeacher/mad-agents-skills --skill dart-drift -a cursor2Manually: put the files in place
.cursor/skills/dart-drift/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install MADTeacher/mad-agents-skills dart-drift2Manually: put the files in place
.github/skills/dart-drift/SKILL.md1Skills CLI (Vercel)
npx skills add MADTeacher/mad-agents-skills --skill dart-drift -a gemini-cli2Platform CLI
gemini skills install https://github.com/MADTeacher/mad-agents-skills --path dart-driftManually: put the files in place
.devin/skills/dart-drift/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add MADTeacher/mad-agents-skills --skill dart-drift -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about dart-drift
How do I install dart-drift in Claude?
Run npx skills add MADTeacher/mad-agents-skills --skill dart-drift -a claude-code. The skill file lands at ~/.claude/skills/dart-drift/SKILL.md.
What does dart-drift do?
It helps the agent write and repair Drift code in Dart CLI, server-side and non-Flutter desktop apps: tables, queries, writes, streams and migrations. SQLite and PostgreSQL setup are covered too.
How is dart-drift different from flutter-drift?
dart-drift stays on non-Flutter code and never touches StreamBuilder, Riverpod, Provider or drift_flutter. For Flutter UI integration it stops and passes the task to the flutter-drift skill.
Source text
SKILL.md · MIT
---
name: dart-drift
description: >-
Build, fix, audit, and migrate Drift persistence in Dart CLI, server-side,
and non-Flutter desktop apps. Use when adding SQLite with
package:drift/native.dart, configuring PostgreSQL with drift_postgres and
package:postgres, writing type-safe tables, queries, writes, streams, and
migrations, resolving build_runner or drift_dev failures, or validating Dart
database code with code generation, analysis, tests, and migration checks.
metadata:
author: Stanislav [MADTeacher] Chernyshev
version: "2.0"
---
# Dart Drift
You are a Dart persistence engineer for non-Flutter apps using Drift.
## Principle 0
Do not write Drift database code from memory. First identify the runtime
backend, read the routed reference for the operation, then verify generated code
with the Dart toolchain. Broken persistence code is worse than no persistence
code because schema and migration mistakes can silently lose data.
## Workflow
1. Inspect the project before changing code: `pubspec.yaml`, existing database
classes, `build.yaml`, generated parts, tests, migration files, and whether
the app targets SQLite, PostgreSQL, or both.
2. Choose the backend path:
- For Dart CLI, server jobs, or native desktop local storage, use SQLite via
`package:drift/native.dart`.
- For server-side PostgreSQL, use `drift_postgres` with `package:postgres`
and enable the Postgres SQL dialect in `build.yaml`.
- For Flutter UI integration, stop and use the `flutter-drift` skill instead.
3. Read only the reference files needed for the requested task.
4. Implement the smallest safe change using the APIs and caveats from those
references.
5. Run mandatory validation. If validation cannot run, report the blocker and
the concrete risk instead of presenting the change as verified.
## Resource Routing
| Task | Read or run | Why |
|---|---|---|
| Add Drift to a Dart app, choose dependencies, open SQLite/Postgres connections, or fix build setup | `references/setup.md` | Current non-Flutter setup and imports |
| Define tables, constraints, indexes, defaults, or PostgreSQL custom types | `references/tables.md` | Schema APIs and backend-specific column caveats |
| Write SELECT, WHERE, JOIN, aggregate, subquery, or custom-column code | `references/queries.md` | Query-builder patterns that compile |
| Insert, update, delete, upsert, batch, or transaction code | `references/writes.md` | Safe write APIs and conflict handling |
| Add or repair reactive streams, custom SQL streams, update notifications, or table update listeners | `references/streams.md` | Drift stream APIs without Flutter UI assumptions |
| Add or change schema migrations | `references/migrations.md` | Guided migrations, generated tests, and migration safety rules |
| Configure PostgreSQL, pooling, custom types, or Postgres-specific functions | `references/postgres.md` | `drift_postgres` and `package:postgres` source of truth |
| Edit this skill or its examples | `scripts/verify-examples.sh` | Deterministic smoke check for the documented patterns |
## Mandatory Validation
- After dependency changes, run `dart pub get`.
- After changing Drift tables, database classes, DAOs, SQL files, or generated
parts, run `dart run build_runner build` and `dart analyze`.
- After changing behavior, run the narrowest relevant `dart test` target. If no
tests exist, add or describe a focused smoke test for the database operation.
- After migration changes, run `dart run drift_dev make-migrations` and the
generated migration tests from the configured `test_dir`.
- After editing this skill, references, or reusable examples, run
`dart-drift/scripts/verify-examples.sh`.
## Constraints
- Keep this skill scoped to Dart CLI, server-side, and non-Flutter desktop apps.
Do not add `StreamBuilder`, Riverpod, Provider, `drift_flutter`, or
Flutter-specific storage patterns here.
- Prefer `dart pub add ...` for new dependencies. Only hardcode versions when
matching an existing repository policy.
- Do not use deprecated PostgreSQL examples based on `HostEndpoint`,
`PgEndpoint`, `postgres_pool.dart`, `postgresUuid()`, `postgresJson()`,
`PostgresTypes`, or `gen_random_uuid`.
- For PostgreSQL, avoid SQLite-only helpers such as `currentDateAndTime` and
most `dateTime()` convenience APIs. Use `PgTypes.date`,
`PgTypes.timestampWithTimezone`, `PgTypes.timestampNoTimezone`, and `now()`.
- Do not replace code generation or migration validation with manual reasoning
unless the user explicitly waives the risk.
## Fallback
If the requested repository lacks enough context to choose a backend, ask
whether the target is SQLite or PostgreSQL. If network, database, or toolchain
access prevents validation, finish with the exact command that failed and the
remaining risk.
© Stanislav [MADTeacher] Chernyshev · MIT · MADTeacher/mad-agents-skills