跳到正文
cccskills
免费在 GitHub 公开

api-baas-supabase

Supabase backend-as-a-service — Auth, Database, Realtime, Storage, Edge Functions, RLS policies, typed client

中文简介正在准备中。当前显示原文说明。

集成工具SupabasePostgreSQL
查看安装方法

包含的文件(8)

  • SKILL.md15.6 KB
  • examples/auth.md9.3 KB
  • examples/core.md7.2 KB
  • examples/database.md12.2 KB
  • examples/edge-functions.md12.4 KB
  • examples/storage.md8.4 KB
  • metadata.yaml502 B
  • reference.md8.1 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Supabase Patterns

Quick Guide: Use Supabase as your backend-as-a-service for Postgres database, authentication, realtime subscriptions, file storage, and edge functions. Always use the typed client with Database generic, enable RLS on every table, and use the secret key only on the server.


<critical_requirements>

CRITICAL: Before Using This Skill

(You MUST enable Row Level Security (RLS) on EVERY table in an exposed schema — no exceptions)

(You MUST use the Database generic type with createClient<Database>() for type-safe queries)

(You MUST NEVER expose the secret key in client-side code — use the publishable key in browsers, the secret key only on the server)

(You MUST use (select auth.uid()) wrapped in a subquery inside RLS policies for performance)

(You MUST handle all Supabase responses with { data, error } destructuring — never assume success)

</critical_requirements>


Auto-detection: Supabase, createClient, @supabase/supabase-js, @supabase/ssr, supabase-js, auth.uid(), RLS, row level security, realtime, postgres_changes, supabase.auth, supabase.from, supabase.storage, supabase.functions, supabase.channel, edge function, Deno.serve

When to use:

  • Setting up a Supabase client with TypeScript type safety
  • Implementing authentication (email/password, OAuth, magic links, session management)
  • Querying Postgres via the Supabase client (select, insert, update, delete, RPC)
  • Writing Row Level Security policies for data access control
  • Subscribing to database changes in real time
  • Uploading and serving files from Supabase Storage
  • Building serverless functions with Supabase Edge Functions (Deno)

Key patterns covered:

  • Typed client setup with Database generic and environment variables
  • Auth flows: sign up, sign in, OAuth, magic link, session refresh, onAuthStateChange
  • Database queries with filters, joins, RPC calls, and error handling
  • RLS policies: USING vs WITH CHECK, auth.uid(), role-based access
  • Realtime subscriptions via channel().on('postgres_changes')
  • Storage: upload, signed URLs, public URLs, bucket policies
  • Edge Functions: Deno.serve, CORS headers, secrets, Supabase client in functions

When NOT to use:

  • Direct Postgres connections (use a database driver skill instead)
  • Complex server-side ORM patterns (use a dedicated ORM skill)
  • Non-Supabase authentication providers (use dedicated auth skills)

Detailed Resources:

  • For decision frameworks and anti-patterns, see reference.md

Client & Queries:

Authentication:

  • examples/auth.md — Full auth flows, OAuth, magic links, session refresh, middleware protection

Database:

Storage:

Edge Functions:


<philosophy>

Philosophy

Supabase is an open-source Firebase alternative built on Postgres. It provides a complete backend through a combination of Postgres extensions, auto-generated REST/GraphQL APIs, authentication, realtime subscriptions, file storage, and edge functions.

Core principles:

  1. Postgres at the core — Every feature is built on Postgres. RLS policies, auth, and realtime all leverage Postgres primitives. Understanding Postgres is understanding Supabase.
  2. Type safety end-to-end — Generate TypeScript types from your database schema with supabase gen types. Pass the Database generic to createClient for fully typed queries.
  3. Security by default — RLS must be enabled on every table. The publishable key is safe for browsers (RLS enforces access). The secret key bypasses RLS and must never leave the server.
  4. Error as values — Every Supabase method returns { data, error }. Never assume success. Always check error before using data.
  5. Realtime built in — Postgres changes stream over WebSockets via channels. No separate pub/sub infrastructure needed.
  6. Edge-first functions — Edge Functions run Deno at the edge, close to users. Design for short-lived, idempotent operations.

When to use Supabase:

  • Rapid backend development with Postgres, auth, and storage out of the box
  • Projects needing realtime features (chat, notifications, live dashboards)
  • Teams wanting to avoid managing separate auth, database, and storage services
  • Applications that benefit from Row Level Security for multi-tenant data isolation

When NOT to use:

  • Complex server-side business logic requiring a full application server (use Edge Functions for simple cases, a dedicated API for complex ones)
  • Applications needing an ORM with advanced query building (Supabase query builder is powerful but not a full ORM)
  • Offline-first applications requiring complex sync protocols
</philosophy>
<patterns>

Core Patterns

Pattern 1: Typed Client Setup

Always pass the Database generic to createClient for full autocomplete on table names, column names, and return types. Use environment variables for URL and keys.

export const supabase = createClient<Database>(
  SUPABASE_URL,
  SUPABASE_PUBLISHABLE_KEY,
);

Without the generic, typos in table/column names are not caught at compile time. See examples/core.md for browser, server, and admin client setup patterns.


Pattern 2: Error Handling with { data, error }

Every Supabase method returns { data, error }. Always destructure and check error before using data. Never use non-null assertions on data.

const { data, error } = await supabase
  .from("profiles")
  .select("id, username")
  .eq("id", userId)
  .single();
if (error) throw new Error(`Failed to fetch profile: ${error.message}`);

See examples/core.md for the reusable error handler pattern and common mistakes.


Pattern 3: Authentication Flows

Supabase Auth supports email/password (signInWithPassword), OAuth (signInWithOAuth), magic links (signInWithOtp), and phone OTP. Register onAuthStateChange early in the app lifecycle and always clean up with subscription.unsubscribe().

Key gotcha: Do NOT call Supabase methods directly inside onAuthStateChange — use setTimeout(..., 0) to defer.

See examples/auth.md for sign up, sign in, OAuth, magic link, session management, middleware protection, and password reset patterns.


Pattern 4: Database Queries

Use the query builder for type-safe CRUD with filters, joins, ordering, and pagination. Always add .select() after .insert() or .update() to return the affected row.

const { data, error } = await supabase
  .from("posts")
  .select("id, title, author:profiles(username)")
  .eq("published", true)
  .order("created_at", { ascending: false })
  .range(0, PAGE_SIZE - 1);

See examples/database.md for complex queries, upserts, RPC calls, conditional filters, counting, and migrations.


Pattern 5: Row Level Security (RLS) Policies

RLS is the primary security mechanism. Enable it on every table, write separate policies per operation (not FOR ALL), and wrap auth.uid() in a subquery for performance.

alter table public.posts enable row level security;

create policy "posts_select" on public.posts for select to authenticated
using ( published = true or (select auth.uid()) = author_id );

Never trust user_metadata from JWT for access control — it is user-modifiable. See examples/database.md for full CRUD policies, team-based access, and anti-patterns.


Pattern 6: Realtime Subscriptions

Subscribe to database changes via channel().on('postgres_changes', ...). Always unsubscribe on cleanup. DELETE events cannot be filtered — all deletes are received. UPDATE/DELETE payloads need replica identity full for old record data.

const channel = supabase
  .channel("room-messages")
  .on(
    "postgres_changes",
    {
      event: "INSERT",
      schema: "public",
      table: "messages",
      filter: `room_id=eq.${roomId}`,
    },
    (payload) => {
      /* handle */
    },
  )
  .subscribe();

Use for chat, live dashboards, notifications. Avoid for high-frequency data (> 100 updates/sec).


Pattern 7: Storage Operations

Upload files with supabase.storage.from(bucket).upload(). Use getPublicUrl() for public buckets, createSignedUrl() for private buckets with time-limited access. Storage access control uses RLS on storage.objects.

See examples/storage.md for upload, signed URLs, public URLs, image transforms, bucket policies, and signed upload URLs.


Pattern 8: Edge Functions

Use Deno.serve() (not the deprecated serve import). Import supabase-js with npm: prefix: import { createClient } from "npm:@supabase/supabase-js@2". Handle CORS on every response. Use Deno.env.get() for secrets. Forward user JWT for RLS enforcement.

See examples/edge-functions.md for basic functions, authenticated access, shared utilities, webhooks, multi-route "fat functions", and background processing with EdgeRuntime.waitUntil().

</patterns>

<decision_framework>

Decision Framework

Which Supabase Key to Use

Where is the code running?
├─ Browser / Client-side → publishable key (RLS enforced)
├─ Server / API route → publishable key + user JWT (RLS enforced per user)
└─ Admin / Migration script → secret key (bypasses RLS)
    └─ NEVER expose the secret key in client bundles

Auth Method Selection

What auth flow does the user need?
├─ Email + Password → signInWithPassword
├─ Social login (GitHub, Google, etc.) → signInWithOAuth
├─ Passwordless email → signInWithOtp (magic link)
├─ Phone + SMS → signInWithOtp (phone)
└─ SSO / SAML → signInWithSSO (enterprise)

Realtime vs Polling

How fresh must the data be?
├─ Instant (< 1 second) → Realtime subscription (postgres_changes)
├─ Near-instant (1-5 seconds) → Realtime subscription
├─ Periodic (> 5 seconds ok) → Polling with setInterval
└─ On-demand (user refresh) → Re-fetch on action
    └─ High-frequency updates (> 100/sec)?
        ├─ YES → Polling or batch (Realtime has per-subscriber checks)
        └─ NO → Realtime is fine

Storage: Public vs Private Buckets

Who should access the files?
├─ Anyone (public assets, avatars) → Public bucket + getPublicUrl()
├─ Authenticated users only → Private bucket + createSignedUrl()
├─ Specific users (own files) → Private bucket + RLS on storage.objects
└─ Server-only processing → secret key for upload/download

Edge Functions vs Client Queries

Does the operation need server-side logic?
├─ Simple CRUD → Client query with RLS (no edge function needed)
├─ Multi-step / transactional → Edge function or Postgres function (RPC)
├─ Third-party API call → Edge function
├─ Webhook receiver → Edge function
└─ Heavy computation → Edge function with EdgeRuntime.waitUntil() for background work

</decision_framework>


<red_flags>

RED FLAGS

High Priority Issues:

  • Missing RLS on tables — Any table without RLS in an exposed schema is completely open to the public. In January 2025, 170+ apps were found with exposed databases due to missing RLS (CVE-2025-48757).
  • Secret key in client code — The secret key (formerly service_role key) bypasses all RLS. Exposing it in browser bundles gives every user full admin database access.
  • Ignoring { data, error } returns — Accessing data without checking error leads to runtime crashes when operations fail.
  • Using auth.jwt() ->> 'user_metadata' in RLS policies — user_metadata is modifiable by authenticated users via updateUser(). Never use it for access control decisions.

Medium Priority Issues:

  • Using FOR ALL in RLS policies — Separate into SELECT, INSERT, UPDATE, DELETE policies for clarity and auditability.
  • Bare auth.uid() in policies without subquery — Wrap in (select auth.uid()) for up to 94-99% performance improvement per Supabase benchmarks.
  • Not specifying to authenticated or to anon in policies — Without a role, policies apply to all roles, which may expose data unintentionally.
  • Using select("*") everywhere — Fetches all columns including sensitive data. Select only the columns you need.
  • Deprecated serve import in Edge Functions — import { serve } from "https://deno.land/std/http/server.ts" is deprecated. Use Deno.serve().

Common Mistakes:

  • Not adding .select() after .insert() or .update() — Without .select(), these methods return no data (only null).
  • Missing CORS headers in Edge Functions — Browser requests fail without proper CORS headers and OPTIONS handling.
  • Not unsubscribing from Realtime channels — Leaks WebSocket connections and can cause memory issues.
  • Using bare specifiers in Edge Functions — import { createClient } from "@supabase/supabase-js" fails in Deno. Use npm:@supabase/supabase-js@2.
  • Using getSession() to verify auth — getSession() reads from local storage and can be tampered with. Use getUser() for secure server-side verification.

Gotchas & Edge Cases:

  • Realtime DELETE events cannot be filtered — All deletes for a subscribed table are received regardless of filter.
  • Realtime requires replica identity full for old record data — By default, UPDATE and DELETE payloads only include the new record. Set alter table X replica identity full to access payload.old.
  • RLS policies are not applied to Realtime DELETE events — Be cautious about what information DELETE events expose.
  • onAuthStateChange fires on tab focus — SIGNED_IN events fire when a browser tab regains focus, not just on actual sign-in.
  • Do NOT call Supabase methods inside onAuthStateChange callback — This can cause deadlocks. Use setTimeout(..., 0) to defer.
  • Signed URLs expire — createSignedUrl() URLs expire after the specified duration. Signed upload URLs expire after 2 hours.
  • Public bucket URLs bypass RLS — Files in public buckets are accessible to anyone with the URL, regardless of policies.
  • Edge Function cold starts — First invocation after idle period has additional latency. Design "fat functions" (fewer, larger functions) to minimize cold starts.
  • Edge Functions: file writes only on /tmp — The /tmp directory is the only writable path in edge functions.

</red_flags>


<critical_reminders>

CRITICAL REMINDERS

(You MUST enable Row Level Security (RLS) on EVERY table in an exposed schema — no exceptions)

(You MUST use the Database generic type with createClient<Database>() for type-safe queries)

(You MUST NEVER expose the secret key in client-side code — use the publishable key in browsers, the secret key only on the server)

(You MUST use (select auth.uid()) wrapped in a subquery inside RLS policies for performance)

(You MUST handle all Supabase responses with { data, error } destructuring — never assume success)

Failure to follow these rules will create security vulnerabilities, type-unsafe queries, and silent runtime failures.

</critical_reminders>

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Hugging Face Inference SDK patterns for TypeScript/Node.js — InferenceClient setup, chat completion, text generation, streaming, embeddings, image generation, audio transcription, translation, summarization, and Inference Endpoints

中文简介正在准备中。当前显示原文说明。

agents-inc/skills252026年10月17日 更新

LiteLLM proxy server setup, TypeScript client patterns via OpenAI SDK, model routing, fallbacks, load balancing, spend tracking, virtual keys, and production deployment

中文简介正在准备中。当前显示原文说明。

agents-inc/skills252026年10月17日 更新

Serverless GPU compute platform for AI model deployment — web endpoints, GPU functions, model serving, and TypeScript client patterns

中文简介正在准备中。当前显示原文说明。

agents-inc/skills252026年10月17日 更新

Local LLM inference with the Ollama JavaScript client -- chat, streaming, tool calling, vision, embeddings, structured output, model management, and OpenAI-compatible endpoint

中文简介正在准备中。当前显示原文说明。

agents-inc/skills252026年10月17日 更新

Replicate SDK patterns for TypeScript/Node.js -- client setup, predictions, streaming, webhooks, file handling, model versioning, deployments, and training

中文简介正在准备中。当前显示原文说明。

agents-inc/skills252026年10月17日 更新

Together AI SDK patterns for TypeScript — client setup, chat completions, streaming, structured output, function calling, embeddings, image generation, fine-tuning, and OpenAI-compatible endpoints

中文简介正在准备中。当前显示原文说明。

agents-inc/skills252026年10月17日 更新

查看 agents-inc 的全部技能

报告此技能的问题