bdr-account-research

The MCP tool owns the data pipeline. This skill owns the rendering.

Given a company website and a tenant ID, the skill calls bdr_account_research once and turns the returned envelope (tenant config, derived use cases, 10-K extracts, LinkedIn footprint, intent signals, AI-scored prospects, plus an inline render_spec) into a self-contained, branded A4 HTML report.

When the agent picks this skill

The user says any of:

• “Generate a report on X.” / “Research account X.”
• “Build a BDR brief for X.”
• “Run account research for X.”
• Mentions a company domain alongside signals, prospects, 10-K, use cases, or tenant.

What the skill teaches the agent

• Call bdr_account_research exactly once. Do not hand-orchestrate Snowflake, signals, or prospecting calls — all of that lives inside the tool.
• Follow the inline render_spec strictly — section order, layout, brand colours, page format (A4 HTML).
• Run the skill’s pre-delivery checklist before delivery.
• Handle follow-up questions by slicing the datasets the tool already pulled via query_datasets, or by calling a narrow typed tool for genuinely new data:
  • query_company_filings for a 10-K section not surfaced by default.
  • query_employee_footprint for a competitor footprint cut.
  • query_intent_signals for a topic outside the tenant’s keyword set.
• Resume cleanly when the orchestrator returns status="still_running" — same arguments, the server-side run id is reused.

Underlying MCP tools

• bdr_account_research — the orchestrator.
• query_datasets — slicing the returned datasets map.
• Typed follow-ups: query_company_filings, query_employee_footprint, query_intent_signals.