generated: '2026-07-26'
method: generated
source: >-
  Packaged Agent Skills generated from openapi/knight-frank-api-v3-openapi.json
  and verified against live anonymous calls to every operation on 2026-07-26.
  Cross-cutting rules cite ../conventions/, ../errors/, ../authentication/ and
  ../data-model/. The harvested contract declares no operationIds, so each skill
  is grounded in the real method+path pairs and references the operationIds
  assigned in ../overlays/knight-frank-api-v3-overlay.yaml.
note: >-
  Knight Frank publishes no AGENTS.md, no skills and no agent documentation —
  both corporate domains return 404 for /llms.txt and /.well-known/ai-plugin.json,
  and robots.txt blanket-disallows a list of AI crawlers. Every skill below
  carries the same caveat: the api-v3 service is live and unauthenticated but is
  internal plumbing offered under no terms, with no support channel and no
  stability commitment.
skills:
  - file: knight-frank-find-office.md
    name: Find a Knight Frank office
    api: openapi/knight-frank-api-v3-openapi.json
    operations: [searchOffices, getOfficeById]
    paths: ['GET /office', 'GET /office/{id}']
  - file: knight-frank-search-content-and-research.md
    name: Search Knight Frank content and research
    api: openapi/knight-frank-api-v3-openapi.json
    operations: [searchAll, searchIntelligenceLab, getIntelligenceLabFacets, searchCmsPages, searchServiceLines]
    paths: ['GET /search', 'GET /intelligencelab', 'GET /intelligencelab/facets', 'GET /cmspage', 'GET /service-lines']
excluded_operations:
  - operation: GET /person, GET /person/autocomplete, GET /person/cms-search
    reason: >-
      Deliberately NOT packaged as an agent skill. These operations return named
      individuals' corporate email, direct-dial and mobile numbers to any caller
      with no credential. Documenting that the exposure exists is legitimate
      research and is recorded in ../review.yml; shipping an agent-ready recipe
      with working pagination for it is not. The finding stays, the tooling does not.
  - operation: POST /telemetry/increment-selected-count
    reason: >-
      Non-idempotent, unauthenticated analytics write with no consumer value.
      Deliberately not packaged as an agent skill.