Inventory Audit Workflow

When auditing the local database inventory, follow these steps:

1. Execution Mode Selection (Interactive Choice):
   • Before performing the query, prompt the user using the ask_question tool to select the demo mode:
     • Question: "Which MCP tool execution method would you like to test?"
     • Options:
       1. "(Recommended) Direct Eager Call (mcp_local-db-server_get_inventory_eager)"
       2. "Direct Lazy Call (call_mcp_tool for get_inventory_lazy)"
       3. "Spawn 'self' subagent with Eager Call"
       4. "Spawn 'self' subagent with Lazy Call"
2. Branch A: Direct Eager Call:
   • Query inventory records directly using the native eager MCP tool mcp_local-db-server_get_inventory_eager.
   • Format results into a markdown table (columns: SKU, Item Name, Quantity, Status) and present directly to the user.
3. Branch B: Direct Lazy Call:
   • Check schema if needed, then invoke the proxy meta-tool call_mcp_tool with:
     • ServerName: "local-db-server"
     • ToolName: "get_inventory_lazy"
   • If get_inventory_lazy is not found or in sessions where call_mcp_tool is not available, report this behavior to illustrate lazy loading requirements.
   • Format results into a markdown table and present directly to the user.
4. Branch C: Subagent with Eager Call:
   • Spawn a subagent via invoke_subagent with:
     • TypeName: "self"
     • Role: "Inventory Auditor (Eager)"
     • Prompt: "Query inventory from local-db-server using the eager tool mcp_local-db-server_get_inventory_eager and return the records formatted as a markdown table."
     • Workspace: "inherit"
   • Wait for the reactive message callback, then present the returned table to the user.
5. Branch D: Subagent with Lazy Call:
   • Spawn a subagent via invoke_subagent with:
     • TypeName: "self"
     • Role: "Inventory Auditor (Lazy)"
     • Prompt: "Query inventory from local-db-server using call_mcp_tool for get_inventory_lazy on local-db-server and return the records formatted as a markdown table."
     • Workspace: "inherit"
   • Wait for the reactive message callback, then present the returned table to the user.

Constraints:

• Do NOT view, read, or inspect source code files (e.g. mcp_server.py) or previous transcripts to extract data.
• Do NOT execute shell commands or script files directly via run_command to query data.