What problem does it solve? GraphQL APIs fail in predictable ways: non-null fields that blank entire screens when one service degrades, offset pagination that duplicates rows, N+1 resolvers firing hundreds of queries per request, and schema changes that silently break shipped clients. This Skill encodes the adjudicated rules for designing, reviewing, and evolving a GraphQL API so these failures are caught before production. ## Core Features & Use Cases - Schema design and review: nullability audits, Relay Connection pagination, error modelling with payload userErrors versus result unions, enums, input objects, and global object identification. - Performance and demand control: per-request DataLoader batching with the keys.length contract, depth and complexity limits, trusted documents, and GET/CDN caching. - Federation and evolution: Apollo Federation 2 directives (@key, @shareable, @override, @requires), composition checks, and additive schema evolution gated by graphql-inspector diffs. - Use Case: A home-feed request fires several hundred SQL statements and one HTTP call per post. The Skill identifies the shared module-level DataLoader leaking data across users, rewrites the batch functions to return keys.length results in order, and adds depth and complexity limits so cyclical queries are rejected at validation time. ## Quick Start Ask the agent to review your GraphQL schema file for nullability, pagination, and error-handling problems using the graphql skill.