GraphQL Reviewer Skill

GraphQL Reviewer Skill

Purpose

Reviews GraphQL schemas, resolvers, and operations for N+1 problems, query complexity limits, input validation, security best practices, and proper error handling.

When to Use

• GraphQL schema or resolver review requests
• "GraphQL", "N+1", "DataLoader", "query complexity" mentions
• Schema design review
• Projects with .graphql, .gql files
• GraphQL library dependencies (Apollo, Relay, graphql-js)

Project Detection

• .graphql or .gql schema files
• schema.graphql or type-defs.ts
• graphql package in dependencies
• @apollo/​server, graphql-yoga, mercurius dependencies
• @Query, @Mutation, @Resolver decorators (NestJS/​TypeGraphQL)

Workflow

Step 1: Analyze Project

**GraphQL Server**: Apollo Server 4.x / GraphQL Yoga
**Schema**: Code-first / SDL-first
**Language**: TypeScript / JavaScript
**ORM**: Prisma / TypeORM / Drizzle
**Key Features**:
  - DataLoader for batching
  - Query complexity plugin
  - Persisted queries

Step 2: Select Review Areas

AskUserQuestion:

"Which GraphQL areas to review?"
Options:
- Full GraphQL audit (recommended)
- N+1 / DataLoader patterns
- Schema design
- Query complexity / Security
- Error handling
- Input validation
multiSelect: true

Detection Rules

Critical: N+1 Query Problem

// BAD: N+1 problem
// Schema
type Query {
  posts: [Post!]!
}

type Post {
  id: ID!
  author: User!  // N+1 here!
}

// Resolver - fetches author per post
const resolvers = {
  Query: {
    posts: () => db.post.findMany()
  },
  Post: {
    author: (post) => db.user.findUnique({ where: { id: post.authorId } })
    // If 100 posts → 1 + 100 queries!
  }
};

// GOOD: DataLoader for batching
import DataLoader from 'dataloader';

const createLoaders = () => ({
  userLoader: new DataLoader(async (ids: string[]) => {
    const users = await db.user.findMany({
      where: { id: { in: ids } }
    });
    const userMap = new Map(users.map(u => [u.id, u]));
    return ids.map(id => userMap.get(id) ?? null);
  })
});

// Resolver with DataLoader
const resolvers = {
  Post: {
    author: (post, _, { loaders }) => loaders.userLoader.load(post.authorId)
    // Now: 1 + 1 queries (batched)
  }
};

// BEST: Prisma with includes (no N+1)
const resolvers = {
  Query: {
    posts: () => db.post.findMany({
      include: { author: true }  // Single query with JOIN
    })
  }
};

Critical: Excessive Fetching in Resolvers

// BAD: Fetches all fields regardless of query
const resolvers = {
  Query: {
    user: (_, { id }) => db.user.findUnique({
      where: { id },
      include: {
        posts: true,      // Maybe not requested
        comments: true,   // Maybe not requested
        followers: true   // Maybe not requested
      }
    })
  }
};

// GOOD: Use info to select only requested fields
import { GraphQLResolveInfo } from 'graphql';
import graphqlFields from 'graphql-fields';

const resolvers = {
  Query: {
    user: (_, { id }, __, info: GraphQLResolveInfo) => {
      const requestedFields = graphqlFields(info);
      return db.user.findUnique({
        where: { id },
        include: {
          posts: 'posts' in requestedFields,
          comments: 'comments' in requestedFields
        }
      });
    }
  }
};

// BETTER: Use Prisma's select based on GraphQL query
import { PrismaSelect } from '@paljs/​plugins';

const resolvers = {
  Query: {
    user: (_, { id }, __, info) => {
      const select = new PrismaSelect(info).value;
      return db.user.findUnique({ where: { id }, ...select });
    }
  }
};

Critical: Mutation in Query

# BAD: Mutation disguised as Query
type Query {
  incrementViewCount(postId: ID!): Int!  # WRONG! This mutates data
  markAsRead(notificationId: ID!): Boolean!  # WRONG!
}

# GOOD: Mutations for side effects
type Mutation {
  incrementViewCount(postId: ID!): Post!
  markAsRead(notificationId: ID!): Notification!
}

# Query should be idempotent (read-only)
type Query {
  post(id: ID!): Post
  viewCount(postId: ID!): Int!
}

High: Missing Input Validation

// BAD: No validation
const resolvers = {
  Mutation: {
    createUser: (_, { input }) => {
      // input.email could be anything
      return db.user.create({ data: input });
    }
  }
};

// GOOD: Validate inputs
import { z } from 'zod';

const CreateUserSchema = z.object({
  email: z.string().email(),
  name: z.string().min(1).max(100),
  age: z.number().int().min(0).max(150).optional()
});

const resolvers = {
  Mutation: {
    createUser: (_, { input }) => {
      const validated = CreateUserSchema.parse(input);
      return db.user.create({ data: validated });
    }
  }
};

// Schema-level validation (GraphQL)
"""
User creation input
"""
input CreateUserInput {
  email: String! @constraint(format: "email")
  name: String! @constraint(minLength: 1, maxLength: 100)
  age: Int @constraint(min: 0, max: 150)
}

High: No Query Complexity Limit

// BAD: Allows dangerous queries
// Can request: user.friends.friends.friends.friends...

// GOOD: Limit query depth
import depthLimit from 'graphql-depth-limit';

const server = new ApolloServer({
  typeDefs,
  resolvers,
  validationRules: [depthLimit(5)]  // Max 5 levels deep
});

// GOOD: Query complexity plugin
import { createComplexityPlugin } from 'graphql-query-complexity';

const complexityPlugin = createComplexityPlugin({
  estimators: [
    fieldExtensionsEstimator(),
    simpleEstimator({ defaultComplexity: 1 })
  ],
  maximumComplexity: 1000,
  onComplete: (complexity) => {
    console.log('Query Complexity:', complexity);
  }
});

// Schema with complexity hints
type Query {
  users(first: Int!): [User!]! @complexity(multipliers: ["first"], value: 5)
  posts(first: Int!): [Post!]! @complexity(multipliers: ["first"], value: 3)
}

// GOOD: Rate limiting
import { rateLimitDirective } from 'graphql-rate-limit-directive';

const { rateLimitDirectiveTypeDefs, rateLimitDirectiveTransformer } =
  rateLimitDirective();

type Query {
  expensiveQuery: Data! @rateLimit(limit: 10, duration: 60)
}

High: No List Pagination

# BAD: Unbounded list
type Query {
  posts: [Post!]!  # Could return millions!
}

# GOOD: Relay-style pagination
type Query {
  posts(
    first: Int
    after: String
    last: Int
    before: String
  ): PostConnection!
}

type PostConnection {
  edges: [PostEdge!]!
  pageInfo: PageInfo!
  totalCount: Int!
}

type PostEdge {
  node: Post!
  cursor: String!
}

type PageInfo {
  hasNextPage: Boolean!
  hasPreviousPage: Boolean!
  startCursor: String
  endCursor: String
}

# SIMPLER: Offset pagination (for small datasets)
type Query {
  posts(offset: Int = 0, limit: Int = 20): PostPage!
}

type PostPage {
  items: [Post!]!
  totalCount: Int!
  hasMore: Boolean!
}

High: Missing Error Handling

// BAD: Raw error exposure
const resolvers = {
  Query: {
    user: async (_, { id }) => {
      const user = await db.user.findUnique({ where: { id } });
      if (!user) {
        throw new Error('User not found');  // Generic error
      }
      return user;
    }
  }
};

// GOOD: Structured GraphQL errors
import { GraphQLError } from 'graphql';

class NotFoundError extends GraphQLError {
  constructor(resource: string, id: string) {
    super(`${resource} not found`, {
      extensions: {
        code: 'NOT_FOUND',
        resource,
        id
      }
    });
  }
}

const resolvers = {
  Query: {
    user: async (_, { id }) => {
      const user = await db.user.findUnique({ where: { id } });
      if (!user) {
        throw new NotFoundError('User', id);
      }
      return user;
    }
  }
};

// Error formatting plugin
const formatError = (error: GraphQLError) => {
  // Don't expose internal errors
  if (error.extensions?.code === 'INTERNAL_SERVER_ERROR') {
    return new GraphQLError('Internal server error', {
      extensions: { code: 'INTERNAL_SERVER_ERROR' }
    });
  }
  return error;
};

Medium: Internal ID Exposure

# BAD: Exposes database IDs
type User {
  id: Int!  # Sequential, guessable
}

# GOOD: Use opaque IDs
type User {
  id: ID!  # Could be UUID, hashid, etc.
}

// ID encoding/​decoding
import Hashids from 'hashids';
const hashids = new Hashids('secret-salt', 10);

const resolvers = {
  User: {
    id: (user) => hashids.encode(user.dbId)
  },
  Query: {
    user: (_, { id }) => {
      const [dbId] = hashids.decode(id);
      return db.user.findUnique({ where: { id: dbId } });
    }
  }
};

Medium: Missing Non-null Defaults

# BAD: Unnecessarily nullable
type User {
  id: ID          # Should always exist
  email: String   # Required for user
  name: String    # Should be required
  bio: String     # OK to be nullable
}

# GOOD: Clear nullability
type User {
  id: ID!          # Always present
  email: String!   # Required
  name: String!    # Required
  bio: String      # Optional (nullable)
  deletedAt: DateTime  # Optional
}

# For fields that may fail to resolve
type Post {
  id: ID!
  author: User  # Nullable if author deleted
  authorId: ID! # Always has the reference
}

Response Template

## GraphQL Code Review Results

**Project**: [name]
**Server**: Apollo Server 4.x
**Schema**: SDL-first / Code-first

### N+1 / DataLoader

#### CRITICAL
| File | Line | Issue |
|------|------|-------|
| resolvers/​post.ts | 23 | N+1 in author resolver - use DataLoader |
| resolvers/​user.ts | 45 | posts fetched per user without batching |

### Query Complexity / Security
| File | Line | Issue |
|------|------|-------|
| server.ts | 12 | No depth limit configured |
| schema.graphql | 34 | posts query unbounded - add pagination |

### Input Validation
| File | Line | Issue |
|------|------|-------|
| mutations/​user.ts | 56 | No email validation |
| mutations/​post.ts | 23 | Missing input sanitization |

### Error Handling
| File | Line | Issue |
|------|------|-------|
| resolvers/​query.ts | 78 | Raw error thrown - use GraphQLError |

### Schema Design
| File | Line | Issue |
|------|------|-------|
| schema.graphql | 12 | Query with side effect - move to Mutation |
| types/​user.graphql | 8 | Exposes sequential database ID |

### Recommendations
1. [ ] Implement DataLoader for all relationship resolvers
2. [ ] Add depth limit (max 5-7 levels)
3. [ ] Add query complexity plugin (max 1000)
4. [ ] Add pagination to all list fields
5. [ ] Validate all mutation inputs with Zod/​Yup

### Positive Patterns
- Good use of Relay connections for pagination
- Proper error codes in GraphQL errors

Best Practices

1. DataLoader: Always batch relationship resolvers
2. Complexity: Limit depth and complexity
3. Pagination: Cursor-based for large lists
4. Validation: Validate all inputs server-side
5. Errors: Use structured GraphQL errors
6. Security: Rate limit, no introspection in prod

Integration

• schema-reviewer skill: Database schema
• orm-reviewer skill: ORM patterns
• typescript-reviewer skill: TS type safety
• security-scanner skill: API security

Notes

• Based on GraphQL best practices 2024
• Works with Apollo, Yoga, Mercurius
• Supports both SDL and code-first
• Compatible with Prisma, TypeORM, Drizzle