Add a practical lead-qualification MCP example graph

[madeinplutofabio]

Summary

The bundled echo-pipeline graph proves the adapter works, but it does not show why anyone should care. A lead-qualification graph is a much better demo: it is a workflow agent hosts can repeat thousands of times, where re-planning in the LLM each time is wasteful.

Acceptance criteria

• Add examples/graphs/lead-qualification/graph.yaml.
• Use mock or local bricks only. No external API dependencies.
• Input shape: { company_url, target_icp, optional notes }.
• Output shape: { company_name, industry, icp_match, score, reasons, next_action }.
• Add an MCP smoke example showing the tools/call input/output (mirror examples/mcp/echo-pipeline-mcp-smoke.md).
• Add the new graph to the bricks list or examples list as needed for discoverability.

Why this matters

This supports the public narrative: NCP turns "agent re-orchestrates the same workflow every call" into "agent invokes one deterministic graph that runs in milliseconds."

Where to read

• examples/graphs/echo-pipeline/graph.yaml for the minimum-viable structure
• examples/graphs/support-routing-stubbed/graph.yaml for a slightly more realistic routing example
• spec/ncp-v0.2.3.md for the graph manifest format