Loading ToolBench
Weather MCP server providing access to National Weather Service alerts and forecasts, plus system monitoring tools
Server has 3 tools with basic descriptions and parameter schemas, but significant quality gaps. run_shell_command is a destructive tool with minimal safety guardrails and no error guidance. get_alerts and get_forecast have adequate descriptions (60-80 chars) but lack output schema documentation. Parameters have type definitions but descriptions are minimal (10-20 chars). No tool annotations (readOnlyHint/destructiveHint). Error handling returns unstructured strings rather than actionable recovery guidance. Tool composition is reasonable but output formatting is opaque to the LLM.
Get active weather alerts for a state.
Get weather forecast for coordinates.
Run a shell command and return the output.
run_shell_command lacks destructive operation safeguards. No confirmation step, dry-run mode, or permission gate. Description does not warn of irreversible consequences. Agents can execute arbitrary shell commands without friction.
Output schemas not documented. get_alerts and get_forecast return formatted strings; LLM cannot parse structure or extract fields for downstream tool calls. Responses should be structured JSON with typed fields.
Parameter descriptions are too brief (10-20 chars). 'Two-letter state code (e.g., CA, NY)' is acceptable but 'Latitude coordinate' and 'Longitude coordinate' lack context on valid ranges, precision, or coordinate system (WGS84 assumed but not stated).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | <=2025-11-25 | v2 |
Error responses are unstructured strings ('Unable to fetch alerts or no alerts found'). No recovery guidance. LLM cannot determine if error is retryable, user-fixable, or fatal. Should return structured error with actionable next steps.
No tool annotations. run_shell_command should have destructiveHint=true. get_alerts and get_forecast should have readOnlyHint=true. Annotations enable client-side safety warnings and agent planning.