Single tool 'select' has a reasonable description (89 chars) and complete input schema with types and descriptions. However, the schema lacks critical details: no output schema documentation, no error handling guidance, no parameter constraints (e.g., min/max for options array), and no indication of side effects (opens browser, creates temp server). The tool name 'select' is generic and does not start with a clear action verb, 'open_selector' or 'multi_select_reorder' would be clearer. Parameter 'options' lacks format specification (what structure do items need?). No evidence of idempotency, retry safety, or recovery guidance. The tool appears to be a UI wrapper rather than a data-manipulation tool, which is unusual for MCP and raises questions about agent composability.

• Rename 'select' to 'open_multi_select_dialog' or 'prompt_user_selection' to clarify the action and distinguish from generic selection operations.
• Document the output schema explicitly: 'Returns {selected: [string], ordered: [string], cancelled: bool, error?: string, mode: string}' so LLMs know what fields to expect.
• Add format constraints to 'options' parameter: 'Array of strings or objects with {id: string, label: string, description?: string}. If strings, they are used as both id and label.'
• Add error handling guidance: 'If the browser fails to open, set MULTI_SELECT_REORDER_DIRECT_TTY=1 to use terminal selection instead. Timeout after 1 hour if user does not submit.'
• Document side effects in the tool description: 'Opens a short-lived local web page (browser) for interactive selection. Requires a graphical environment or TTY. Not suitable for headless/automated agents.'