feat: Add tool annotations for improved LLM tool understanding

[triepod-ai]

Summary

Adds MCP tool annotations (readOnlyHint, openWorldHint, title) to all 4 tools to help LLMs better understand tool behavior and make safer decisions about tool execution.

Changes

Tool	title	readOnlyHint	openWorldHint
search_notes	"Search Notes"	true	true
get_note_content	"Get Note Content"	true	true
get_note_comments	"Get Note Comments"	true	true
login	"Login"	-	true

• Added readOnlyHint: true to read-only tools (search, get content, get comments)
• Added openWorldHint: true to all tools (they interact with external Xiaohongshu API)
• Added human-readable title annotations
• Bumped @modelcontextprotocol/sdk from ^1.9.0 to 1.11.0 for annotation support
• Fixed TypeScript type errors in stdioLogger.ts for SDK compatibility

Why This Matters

• readOnlyHint enables MCP clients to auto-approve safe operations without user confirmation
• openWorldHint indicates tools access external services, helping clients apply appropriate policies
• title provides human-readable names for better tool display

Testing

• Build passes (npm run build)
• tools/list returns annotations for all 4 tools
• Annotation values match actual tool behavior

Before/After

Before:

{
  "name": "search_notes",
  "description": "根据关键词搜索笔记"
}

After:

{
  "name": "search_notes",
  "description": "根据关键词搜索笔记",
  "annotations": {
    "title": "Search Notes",
    "readOnlyHint": true,
    "openWorldHint": true
  }
}

🤖 Generated with Claude Code