A Rust-based MCP server implementation supporting tools, prompts, and resources with shared state management
This MCP server has severe definition quality gaps. Only 2 tools are defined (get_forecast, do_nothing). get_forecast has a structured input schema with types but lacks parameter descriptions (latitude, longitude, temperature_multiplier all have null descriptions). The do_nothing tool is particularly problematic, it has a trivial, non-descriptive name and serves no practical purpose. Both tool descriptions are minimal (24 and 21 characters respectively), falling short of the 10-1024 character guidance and LLM optimization targets. No output schema is documented despite tools returning Vec<mcp_schema::PromptContent>. Error handling is absent, no guidance on what failures mean or how to recover. The tool registry code shows proper schema generation via schemars, but parameter descriptions are not populated in the actual tool definitions (ForecastParams struct has no doc comments). This is a proof-of-concept server, not production-grade.
Do absolutely nothing
Get weather forecast for a location
All input parameters lack descriptions. latitude, longitude, and temperature_multiplier have null/missing description fields, violating the critical rule that every parameter must have a description for LLM inference.
Tool descriptions are under 25 characters (get_forecast: 24 chars, do_nothing: 21 chars). These descriptions lack context for LLM selection: they do not explain WHAT the tool does, WHEN to use it, or what it returns.
do_nothing is a meaningless tool name that violates naming conventions. It does not start with a standard action verb (get, create, update, delete, search, list, send). The tool description 'Do absolutely nothing' provides no actionable guidance. This tool should be removed entirely, it adds no value and confuses LLM decision-making.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 34 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 25 | - | v1 |
No output schema is documented. Tools return Vec<mcp_schema::PromptContent>, but callers have no schema to understand what fields to expect. LLMs cannot plan downstream chaining without knowing the response structure.
No error handling or recovery guidance. The do_nothing tool uses `pending().await` which never resolves, this will hang indefinitely without timeout, error message, or guidance. get_forecast returns random temperatures and can fail silently if state.lock() panics.
Parameter temperature_multiplier defaults to 1.0 but has no minimum/maximum bounds documented. Unbounded multipliers could produce absurd temperature values (e.g., 1000x, -1000x) that break real-world usage. No constraint guidance for the LLM.
get_forecast returns mutable shared state (Arc<Mutex<State>>) to track history, but this state is never reset and is not documented to the LLM. The tool modifies state (pushes to history), this should be explicitly noted in the description so the LLM knows the tool has side effects beyond the return value.