Agent Skills: OpenAPI Specification

OpenAPI 3.x specification design, schema patterns, and validation for REST API contracts. Use when creating or maintaining API specifications, generating documentation, or validating API contracts.

UncategorizedID: NickCrew/claude-cortex/openapi-specification

Install this agent skill to your local

pnpm dlx add-skill https://github.com/NickCrew/claude-cortex/tree/HEAD/skills/openapi-specification

Skill Files

Browse the full folder contents for openapi-specification.

Download Skill

Loading file tree…

skills/openapi-specification/SKILL.md

Skill Metadata

Name
openapi-specification
Description
OpenAPI 3.x specification design, schema patterns, and validation for REST API contracts. Use when creating or maintaining API specifications, generating documentation, or validating API contracts.

OpenAPI Specification

Design, validate, and maintain OpenAPI 3.x specifications for REST API contracts. Covers schema patterns, security schemes, versioning, and code generation integration.

When to Use This Skill

  • Creating a new OpenAPI specification from scratch
  • Adding endpoints or schemas to an existing spec
  • Reviewing or validating an API contract for correctness
  • Setting up code generation from OpenAPI definitions
  • Designing reusable schema components for large APIs
  • Implementing security schemes in API specifications
  • Managing API versioning and deprecation

Quick Reference

| Resource | Purpose | Load when | |----------|---------|-----------| | references/spec-patterns.md | Schema patterns, security schemes, validation, versioning, reusable components | Designing or reviewing specs |


Workflow

Phase 1: Design       → Define API overview, resources, and operations
Phase 2: Schema       → Build data models with reusable components
Phase 3: Validate     → Lint and verify the spec for correctness
Phase 4: Integrate    → Generate docs, SDKs, and contract tests

Phase 1: API Design

Start with a high-level design before writing the spec:

  1. Identify resources -- what nouns does this API expose?
  2. Map operations -- what CRUD and custom actions apply to each resource?
  3. Define relationships -- how do resources reference each other?
  4. Plan authentication -- what security schemes are needed?
  5. Set conventions -- naming style, pagination, error format

Spec Skeleton

openapi: 3.1.0
info:
  title: [API Name]
  version: 1.0.0
  description: [What this API does]
paths:
  /resources:
    get:
      summary: List resources
      operationId: listResources
    post:
      summary: Create a resource
      operationId: createResource
  /resources/{id}:
    get:
      summary: Get a resource
      operationId: getResource
components:
  schemas: {}
  securitySchemes: {}

Phase 2: Schema Design

Build data models using components/schemas for reuse:

  • Use $ref to reference shared schemas -- never duplicate definitions
  • Apply allOf for composition, oneOf / anyOf for polymorphism
  • Add example values to every schema and property
  • Use required arrays explicitly -- don't rely on implicit behavior
  • Document nullable fields with type: [string, "null"] (3.1) or nullable: true (3.0)

Phase 3: Validate

Run validation before committing any spec changes:

# Spectral (recommended)
spectral lint openapi.yaml

# Redocly
redocly lint openapi.yaml

# swagger-cli
swagger-cli validate openapi.yaml

Common Validation Issues

| Issue | Fix | |-------|-----| | Missing operationId | Add unique operationId to every operation | | Unused schema | Remove from components or add a $ref | | Missing response description | Add description to every response code | | Path parameter not in path | Match {param} in path with parameter definition | | No 2xx response defined | Add at least one success response per operation |


Phase 4: Integrate

Use the validated spec to generate downstream artifacts:

  • Documentation: Redoc, Swagger UI, Stoplight
  • Client SDKs: openapi-generator, autorest, orval
  • Server stubs: openapi-generator with server templates
  • Contract tests: Schemathesis, Dredd, Prism

Quality Checklist

  • [ ] All paths have operationIds
  • [ ] HTTP methods match resource actions (GET reads, POST creates, etc.)
  • [ ] Every response has a description and schema
  • [ ] Security requirements defined at operation or global level
  • [ ] Examples provided for request and response bodies
  • [ ] Consistent naming conventions (camelCase, snake_case -- pick one)
  • [ ] Deprecation fields set on sunset endpoints
  • [ ] Spec passes linter with zero errors

Anti-Patterns

  • Do not inline schemas -- use $ref to components/schemas for anything reused
  • Do not mix API versions in a single spec file
  • Do not use 200 OK for create operations -- use 201 Created
  • Do not omit error response schemas -- document 4xx and 5xx consistently
  • Do not use additionalProperties: true without clear justification