feat: Add tool annotations for improved LLM tool understanding

[triepod-ai]

Summary

Adds MCP tool annotations (readOnlyHint, destructiveHint, title) to all three MCP tools to help LLMs better understand tool behavior and make safer decisions about tool execution.

Changes

• Added readOnlyHint: true to listTables (read-only table listing)
• Added readOnlyHint: true to describeTable (read-only schema lookup)
• Added destructiveHint: true to executeQuery (can execute INSERT/UPDATE/DELETE queries)
• Added title annotations for human-readable display

Why This Matters

Tool annotations provide semantic metadata that MCP clients can use to:

• Auto-approve safe operations: Clients like Claude Code can automatically approve read-only tools without user confirmation
• Prompt for confirmation: Destructive tools trigger confirmation prompts before execution
• Display meaningful titles: Human-readable names improve UI/UX in client applications
• Enable safer automation: Workflows can distinguish between read and write operations

Tool Summary

Tool	Title	ReadOnly	Destructive
listTables	List Tables	✅	-
describeTable	Describe Table	✅	-
executeQuery	Execute Query	-	✅

Testing

• Code follows mcp-go annotation patterns (WithTitleAnnotation, WithReadOnlyHintAnnotation, WithDestructiveHintAnnotation)
• Using mcp-go v0.41.0 which fully supports annotations
• Annotation values match actual tool behavior

Note: Live verification requires full Anyquery setup with database plugins. The annotation syntax follows the established mcp-go library patterns.

Before/After

Before:

tool := mcp.NewTool("listTables", 
    mcp.WithDescription("Lists all the tables..."))

After:

tool := mcp.NewTool("listTables",
    mcp.WithDescription("Lists all the tables..."),
    mcp.WithTitleAnnotation("List Tables"),
    mcp.WithReadOnlyHintAnnotation(true),
)