Mission: Product-Quality Risk Investigation for Zava — SQL + MCP Agentic Workflow

[NobukoIT]

Mission/open goal Description

Investigate product-quality risk for Zava and produce an evidence-backed recommendation by combining SQL-based quantitative analysis with qualitative insights from documentation, support chats, and vector similarity search.

Harness and model

GitHub Copilot, VS code, Jupyter notebook,

Turn-by-turn journey

A structured agentic workflow was used:

• Enumerated SQL MCP entities and confirmed schema.
• Executed aggregation SQL queries using sqlcmd against PromptathonDb.
• Inspected Docs and SupportChats for qualitative evidence.
• Ran MCP custom tool FindSimilarDocsByDocId to perform vector similarity search.
• Consolidated quantitative and qualitative evidence into an executive brief.

Prompts and agent interactions

• “List available SQL MCP entities.” → MCP describe_entities
• “Show top product categories by revenue.” → SQL via sqlcmd
• “Which SKUs have most support tickets and low satisfaction?” → SQL aggregation on SupportTickets
• “Show docs and run vector similarity for DocId=39.” → MCP find_similar_docs_by_doc_id

MCP tool calls used

• describe_entities — listed: Customer, Doc, Employee, FindSimilarDocsByDocId, Product, SalesOrder, SalesOrderLine, SupportChat, SupportTicket
• read_records — attempted, but direct SQL queries were used for full control
• find_similar_docs_by_doc_id — executed with DocId=39, TopN=5

🧩 Artifacts

• Notebook: artifacts/ZCPTM_issue_notebook.ipynb
  (contains queries, outputs, and executive brief)
• Submission journey: submission_journey.md
• Executive brief (embedded in notebook)

#28

Completion

• Yes, the agent completed the mission or goal.
• No, the agent did not complete the mission or goal.

Bonus work

No response