graphql-apis

GraphQL APIs AI Skill Guide

Overview & Engine Architecture

GraphQL exposes a typed schema (types, queries, mutations, subscriptions) where clients ask for exact fields. Servers resolve fields via resolvers; naive per-field DB access causes N+1 queries. Agents design schemas for product use cases, enforce authz at resolver/field level, batch loads (DataLoader), and treat persisted/allowlisted operations as a production hardening option.

Client query
    |
 GraphQL runtime (parse/validate/execute)
    |
 resolvers (+ DataLoader)
    |
 services / DB

When to use this skill

• Designing or evolving GraphQL schemas
• Implementing resolvers without N+1 pathologies
• Adding authn/authz around sensitive fields
• Pairing GraphQL with Express, FastAPI, or Spring hosts

Operational directives

1. Schema-first (or code-first with schema as contract) - clients depend on stable types.
2. Never trust client-provided IDs without authz checks in resolvers.
3. Batch and cache per-request DB lookups (DataLoader or equivalent).
4. Prefer pagination (connection / cursor) for lists that can grow.
5. Limit query depth/complexity in public APIs; disable introspection in prod if policy requires.

Schema + resolver sketch

type Item {
  id: ID!
  sku: String!
  qty: Int!
}

type Query {
  item(id: ID!): Item
  items(first: Int = 20): [Item!]!
}

type Mutation {
  createItem(sku: String!, qty: Int!): Item!
}

// Pseudocode resolver map
const resolvers = {
  Query: {
    item: (_: unknown, { id }: { id: string }, ctx: Ctx) => ctx.items.byId(id),
    items: (_: unknown, { first }: { first: number }, ctx: Ctx) =>
      ctx.items.list(Math.min(first, 100)),
  },
  Mutation: {
    createItem: async (_: unknown, args: { sku: string; qty: number }, ctx: Ctx) => {
      ctx.requireUser();
      return ctx.items.create(args);
    },
  },
};

Commands

# Depends on stack - examples:
npm run codegen          # GraphQL Code Generator
npx rover graph check    # schema checks when using Apollo tooling