overlay: 1.0.0
info:
  title: API Evangelist enhancements for the Skai API
  version: 1.0.0
extends: ../openapi/skai-kenshoo-api-openapi.yml
x-provenance:
  generated: '2026-08-12'
  method: generated
  source: >-
    Derived from openapi/skai-kenshoo-api-openapi.yml (Skai's published OpenAPI 3.0.0),
    authentication/skai-kenshoo-authentication.yml, rate-limits/skai-kenshoo-rate-limits.yml
    and lifecycle/skai-kenshoo-lifecycle.yml. Every value below is stated by Skai somewhere in
    its own documentation; the overlay only moves those statements into the machine-readable
    contract where they are missing. The original spec is never mutated.
  rationale: >-
    Skai's spec references a security scheme it never defines, documents a rate limit no
    operation declares, marks a deprecation only in prose, and ships no contact or license
    metadata. Those are the gaps this overlay closes.
actions:
- target: $.info
  description: >-
    Add the developer hub as termsOfService/contact anchor and record the API host. Values
    come from Skai's own developer hub and legal pages.
  update:
    contact:
      name: Skai Developer Hub
      url: https://developers.skai.io/
    termsOfService: https://skai.io/legal/
- target: $
  description: >-
    Define the BearerAuth security scheme the spec already REQUIRES at the root but never
    declares. Skai documents it as an Authorization: Bearer header carrying a JWT access
    token obtained from POST /api/v1/token.
  update:
    components:
      securitySchemes:
        BearerAuth:
          type: http
          scheme: bearer
          bearerFormat: JWT
          description: >-
            Skai access token. Exchange your permanent refresh token and client ID at
            POST /api/v1/token (form-encoded body) for a short-lived JWT, then send it as
            `Authorization: Bearer <access_token>`. Tokens typically last 21600 seconds;
            reuse a token until it expires rather than minting a new one per call. Generate
            the refresh token once at https://login.kenshoo.com/api/dev/refresh-token.
- target: $
  description: >-
    Add the published rate-limit numbers as a reusable, machine-readable extension so an agent
    can plan its call budget. 60 requests/minute and 2,000 requests/hour per USER, as stated in
    the Skai Overview.
  update:
    components:
      x-rate-limits:
        scope: per-user
        windows:
        - {window: 1m, limit: 60}
        - {window: 1h, limit: 2000}
        exhaustion_status: 429
        exhaustion_message: API rate limit exceeded
        retry_after_header: null
        header_names_published: false
- target: $
  description: >-
    Declare a reusable 429 response. The spec declares 429 on zero operations even though the
    limit is documented, so no generated client anticipates it.
  update:
    components:
      responses:
        TooManyRequests:
          description: >-
            Rate limit exceeded — 60 requests per minute or 2,000 requests per hour per user.
            Response headers report the limits and remaining calls for the current
            minute/hour window; Skai does not publish the header names.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ApiResponse'
- target: $.paths['/api/v1/reports/async'].post
  description: >-
    Set the machine-readable `deprecated` flag. Skai marks this operation "(deprecated)" in its
    summary string only, so no linter or generated client surfaces it. The replacement is
    POST /api/v1/reports/async/analysis, which Skai's own summary marks "(recommended)".
  update:
    deprecated: true
    x-replaced-by: asyncAnalysisReport
    x-replacement-path: /api/v1/reports/async/analysis
- target: $.info
  description: >-
    Record the reportable-entity vocabulary and the async-job convention as extensions, so an
    agent can discover them without parsing the prose description.
  update:
    x-reportable-entities: [CAMPAIGN, ADGROUP, KEYWORD, AD, PRODUCT_ASSET, PRODUCT_TARGETING,
      PORTFOLIO]
    x-async-job-convention:
      description: >-
        Operations that can exceed a few seconds return a job_id or execution_id; poll for
        status, then fetch the result file.
      poll_operations: [getJobStatus, getAsyncReportStatus]
      fetch_operations: [getJobResults, downloadAsyncReport]
      callbacks: false
    x-idempotency:
      supported: false
      note: >-
        No idempotency key or request deduplication is published. Retrying a write —
        especially bulkUpdate or an endpoint that can return 207 Multi-Status — may re-apply
        changes that already succeeded.
    x-mcp:
      servers:
      - {name: Reporting MCP, url: 'https://mcp.kenshoo.com/reports-mcp', access: read}
      - {name: Operations MCP, url: 'https://mcp.kenshoo.com/operations-mcp', access: write}
      guide: https://skai-mcp-guide.vercel.app/