generated: '2026-09-13'
method: generated
source: https://mcp.globaldata.com/ (Sections 02-10)
summary: >-
  Six packaged Agent Skills for the GlobalData Intelligence Center MCP, one per marquee flow in
  GlobalData's own worked-examples section, plus one orientation skill for the connect-and-discover
  lifecycle that every other flow depends on.
grounding: >-
  There is NO OpenAPI for this provider, so these skills are grounded in GlobalData's published MCP
  tool reference rather than in operationIds. Every tool name used appears verbatim in GlobalData's
  own domain explorer (Section 06/07), and every call sequence mirrors a worked example GlobalData
  publishes (Section 09) or its stated nine-step query lifecycle (Section 08). No tool, parameter,
  endpoint or caveat was invented. Input SCHEMAS are deliberately not reproduced — the live
  tools/list is OAuth-gated and returned HTTP 401, so parameter shapes are shown only where
  GlobalData documents them in prose or in an example.
provider_published_skills:
  found: false
  probes:
    - {url: 'https://mcp.globaldata.com/AGENTS.md', status: 401, note: catch-all, not a gated document}
    - {url: 'https://www.globaldata.com/AGENTS.md', status: 404}
    - {url: 'https://mcp.globaldata.com/skills', status: 401, note: catch-all}
  checked: '2026-09-13'
skills:
  - name: globaldata-connect-and-orient
    file: globaldata-connect-and-orient.md
    summary: >-
      Connect, authenticate, and establish what the subscription actually exposes before spending a
      query. Covers the entitlement-versus-progressive-discovery distinction and the documented
      failure modes (not_entitled, and "tool not found" after a successful reveal).
    tools: [list_domains, get_capabilities, tool_search, discover_capabilities, reveal_advanced, search]
  - name: globaldata-company-profile
    file: globaldata-company-profile.md
    summary: >-
      Identify a company, pull financials and structure, then chain into deals, news, filings, jobs
      and patents on company_id, the primary join key across every content domain.
    tools: [resolve_entities, list_companies, get_company_descriptions, get_company_analytics, list_deals, list_newsarticles, list_filing, list_jobs, list_patents]
  - name: globaldata-deal-activity
    file: globaldata-deal-activity.md
    summary: >-
      Scan M&A, licensing, partnership and JV activity by sector or geography and aggregate it,
      carrying GlobalData's own caveats about undisclosed values and post-close announcements.
    tools: [resolve_entities, list_deals, get_deals_descriptions, get_deals_analytics]
  - name: globaldata-filings-evidence
    file: globaldata-filings-evidence.md
    summary: >-
      Find a filing and extract quotable passages with get_filingsentences rather than downloading
      full text, including the time-limited access URL behaviour on full-document retrieval.
    tools: [list_filing, get_filingsentences, get_filing_document, get_filing_document_text, get_filings_analytics]
  - name: globaldata-hiring-signal
    file: globaldata-hiring-signal.md
    summary: >-
      Read hiring intent from the near-real-time jobs domain and turn it into a trend, with the
      deduplication and thin-bucket salary caveats GlobalData states.
    tools: [resolve_entities, list_jobs, get_job_descriptions, get_jobs_analytics, get_jobs_analytics_multi_view]
  - name: globaldata-news-sentiment
    file: globaldata-news-sentiment.md
    summary: >-
      Track news and social volume and sentiment over time, carrying the model-derived-sentiment
      caveat and the licensed-corpus scope into the report.
    tools: [resolve_entities, list_newsarticles, get_newsarticle_descriptions, get_news_analytics, list_socialmediaposts]
skill_count: 6
cross_cutting_rules_embedded:
  - resolve taxonomy terms before filtering; confidence below 0.65 is dropped into dropped_filters
  - companyName and keyword are free text and must not be resolved
  - at least one selective filter is required or the call is rejected pre-flight as result_set_too_large
  - verify applied_filters before quoting any count; present resolution before data
  - chain domains on company_id, not on company name
  - paginate with pagination.next_page.example_args
  - prefer search(domain, keywords) immediately after a reveal, before the client tool list refreshes
checked: '2026-09-13'