Firebase Data Connect skill: firebase-data-connect (SQL Connect) explained

Source: firebase/agent-skills · Maintainer: Firebase (Google) · License: Apache-2.0 · Checked: 24 October 2026

What is the Firebase Data Connect skill?

firebase-data-connect is Firebase's official Agent Skill for Firebase SQL Connect, the product formerly called Firebase Data Connect. It teaches a coding agent to design a GraphQL schema that maps to PostgreSQL, write queries and mutations with authorization built in, add real-time updates and generate type-safe SDKs for web, Android, iOS, Flutter and Node.js.

Its SKILL.md description says it "Builds and deploys Firebase SQL Connect (aka Firebase Data Connect) backends with PostgreSQL securely." It applies "when designing schemas with tables and relations, writing authorized queries and mutations, configuring real-time data updates, or generating type-safe SDKs." The frontmatter lists Google LLC as author and Databases as the category.

Firebase's SQL Connect docs confirm the rename: "Firebase Data Connect is now known as Firebase SQL Connect." They describe it as "a relational database service for mobile and web apps," backed by Cloud SQL for PostgreSQL, with schemas and operations in GraphQL and generated client SDKs.

As of 24 October 2026, its skills.sh page listed 168.1K installs (snapshot, 24 October 2026).

What do the different names mean?

This skill goes by several names depending on where you look:

The Agent Skills specification says the name field "Must match the parent directory name," so tools may label this skill differently or flag the mismatch. In Claude Code, the skills docs say the frontmatter name, when set, becomes the command you type. Firebase's CLI reference still uses dataconnect in commands such as deploy --only dataconnect and dataconnect:sdk:generate.

What does it do?

The SKILL.md gives the agent a project layout, a strict workflow and a map of reference files to read for each feature.

Project layout. A dataconnect/ folder holds dataconnect.yaml, schema/schema.gql, a connector/ folder with connector.yaml, queries.gql and mutations.gql, and a local-only seed_data.gql.

GraphQL first. The skill tells the agent to default to native GraphQL and use native SQL only when you ask for it or the task needs PostgreSQL features such as PostGIS, window functions or complex aggregations. It warns that native SQL "lacks type safety."

Workflow:

1. Data model. Types with @table, @col and @default, relationships with @ref, and data types such as UUID, Vector and JSON.
2. Authorized operations. Queries with where, orderBy and pagination; _insert, _update, _delete and _upsert mutations; @transaction for multi-step writes; @auth levels and @check and @redact for row-level security; @refresh for real-time updates.
3. Typed SDKs. Configure javascriptSdk, kotlinSdk or swiftSdk output in connector.yaml, then run dataconnect:sdk:generate.

Validation. The agent reviews the generated schema extensions in .dataconnect/schema/main/ and runs dataconnect:compile to check operations against the schema.

Local and deploy. It starts the emulator with emulators:start --only dataconnect, seeds test data, then deploys to Cloud SQL with deploy --only dataconnect.

What's in the folder: SKILL.md, examples.md, templates.md, and a reference/ folder with files for schema, operations, security, real-time, native SQL, search (vector and full text), config, data seeding, Cloud Functions triggers and SDKs for web, Android, iOS, Flutter and the Node.js Admin SDK.

How do you install the Firebase Data Connect skill?

Firebase's agent skills docs install the full Firebase set per tool.

Claude Code plugin (all Firebase skills plus the MCP server):

claude plugin marketplace add firebase/agent-skills
claude plugin install firebase@firebase

Codex plugin:

codex plugin marketplace add firebase/agent-skills
codex plugin add firebase@firebase

Cursor:

npx skills add firebase/agent-skills --agent=cursor

Just this skill, using the command from skills.sh (note the flag uses firebase-data-connect, not the folder name):

npx skills add https://github.com/firebase/agent-skills --skill firebase-data-connect

Add firebase-basics if you install it alone, for the CLI login and project setup. For where Codex and Cursor read skills, see Codex skills and Cursor skills.

How do you use it?

The skill loads when you ask for a relational database in a Firebase project or mention SQL Connect or Data Connect.

Example prompts (suggestions, not quotes from the repo):

• Set up SQL Connect in this Firebase project for a recipe app: users, recipes, ingredients and a many-to-many between recipes and ingredients.
• Write a mutation that lets a signed-in user upsert their own profile, and make sure nobody can edit someone else's.
• Generate the TypeScript SDK for my connector into web/src/lib/dataconnect and use it to list a user's recipes.
• Add full-text search on recipe titles and descriptions.
• Start the SQL Connect emulator, seed ten sample recipes and check every operation compiles.

Before running init dataconnect, the skill tells the agent to make sure it understands your app idea and to ask questions if it is unclear, so a vague prompt may get questions back first.

Tips and limits

• Socket audit: Warn (9 October 2026). On skills.sh, Gen Agent Trust Hub and Snyk show Pass and Socket shows Warn, with one low-severity "Anomaly" alert in examples.md. Socket says the file "appears to be application documentation rather than malware," but that some examples, such as rating updates and a transfer mutation, lack authorization checks and "should not be deployed unchanged."
• Review generated rules. Following Socket's note, read any @auth and @check directives the agent copies from examples before you deploy.
• Billing. As of October 2026, Firebase's SQL Connect pricing page says Spark projects are limited to 8,300 operations per day, Blaze projects get 250,000 client operations per month at no cost, and Cloud SQL trials last 90 days on Spark or 3 months on Blaze.
• Seed data stays local. The skill marks seed_data.gql as "LOCAL ONLY."
• Unpinned CLI. Each command pulls firebase-tools@latest from npm.

Under Apache 2.0 you can copy and modify the skill as long as you keep the license and notices and mark files you change.

Save it in your Kipt library

Kipt keeps prompts and skills as Markdown in a private GitHub repo you own. Paste the skill's GitHub folder link into the Skills import box to bring in SKILL.md, the examples, templates and the reference/ folder, or paste the repo link and pick it from the list of Firebase skills. Install everything you saved into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Firebase or Google.

Similar skills

• firebase-basics: Firebase CLI, login and project setup.
• Firebase Auth skill: sign-in that feeds the @auth checks in your operations.
• Supabase skill: Supabase's official skill for Postgres, Auth and Edge Functions.
• Postgres best practices skill: Supabase's rules for indexes, RLS, pooling and locking on any Postgres.
• Neon Postgres skill: Neon's skill for serverless Postgres and branch-tested migrations.