An MCP server exposing one tool, `ask_user_question`, that asks the user a multiple-choice clarifying question and blocks until they answer. It is a stand-in for Claude Code's built-in AskUserQuestion, which GitHub Copilot has no equivalent of.
Single tool with well-structured schema and comprehensive description. The ask_user_question tool has a complete JSON Schema with nested object definitions, proper type constraints (minItems/maxItems), and detailed parameter descriptions. Description is 280+ chars and explains WHAT (ask clarifying questions), WHEN (different readings lead to different work), and constraints (1-4 questions, 2-4 options each). Tool name is clear and verb-based. However, no output schema is documented, the tool's response structure is not formally specified, forcing LLMs to infer what fields to expect. Error handling guidance is absent; no recovery instructions for timeout or invalid input scenarios.
Ask the user 1-4 clarifying questions, each with 2-4 suggested options, and wait for their answer. Use when different readings of the request would lead to materially different work. Put the option you recommend first and suffix its label with "(recommended)". Every question also offers a free-text "Other" answer. Do not use this to ask whether to proceed, or for choices that have a conventional default you could pick yourself.
Output schema not documented. Tool description and code show the tool returns user answers, but the response structure (fields, types, format for multiSelect vs single-select) is not formally specified. LLMs cannot plan downstream actions without knowing what data they receive.
No error handling guidance. Tool can timeout (30-minute window) or receive invalid input, but error responses do not guide the LLM on recovery. Missing: retryable vs fatal classification, actionable error messages, and fallback strategies.
STDIO transport only. Server is not remotely accessible and cannot be used by hosted MCP clients. Limits deployment to local/CLI-only scenarios.
No tool annotations. Tool does not declare readOnlyHint, destructiveHint, or idempotentHint. The ask_user_question tool is read-only (does not modify state), but this is not formally signaled to clients.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Elicitation fallback not documented. Code comments note that clients without elicitation capability degrade to text output, but the degraded response format and LLM guidance for that path are not specified.