Embedded semantic `guide` tool + lazy/dynamic tool exposure for rmcp MCP servers
The 'guide' tool has a well-defined schema with proper types and descriptions for both parameters (query: string, top_k: integer with minimum constraint). The tool description is clear and actionable (97 chars), explaining both the discovery function and the no-query listing mode. However, the server exposes only ONE tool, a meta-discovery tool rather than domain-specific capabilities. The guide tool itself is well-crafted, but it is a wrapper/framework tool, not a functional tool that performs real work. Output schema is not documented in the source code provided. The tool is read-only (no state mutation), which is appropriate. No error handling guidance is visible in the schema or description.
Discover this server's tools by asking in natural language; returns which tool to use and how. Call with no query for a compact listing of all capabilities.
Output schema not documented. The 'guide' tool returns matched tools and their schemas, but the response structure is not declared in the tool definition or visible in source. LLMs cannot plan downstream calls without knowing what fields to expect.
No error handling guidance. If a query matches no tools or top_k is invalid, the tool description does not explain what the LLM should do next (retry, refine query, etc.).
This is a framework/wrapper tool, not a domain tool. The server provides no actual functional tools, only a discovery mechanism. Agents cannot accomplish real work; they can only discover what tools *could* be available if wrapped by this framework.
Parameter 'top_k' has a minimum constraint (1) but no maximum. Unbounded integers can cause performance issues or excessive output. Specify a reasonable upper bound (e.g., 50).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |