Loading ToolBench
Single tool with basic structure. Tool name lacks action verb convention, description is minimal, and parameter descriptions are present but lack depth. Input schema properly typed with Zod and optional parameter correctly marked. No output schema documented. Basic error handling present but no recovery guidance. Server registers 1 tool (helloTool) alongside 3 resources. Tool definitions are directly visible in src/tools/helloTool.ts with explicit server.tool() registration.
Tool to echo a message or respond with "Hello, World!"
Tool name lacks action verb. 'helloTool' does not start with a verb (get, create, send, etc.) that signals the action. Should be 'echo_message' or 'get_greeting' to match LLM expectations.
Tool description is too brief (58 chars) and lacks WHEN and WHY context. Current: 'Tool to echo a message or respond with "Hello, World!"' Should explain: what it does, when to use it vs other tools, what it returns, and any prerequisites.
Output schema is not documented. The tool returns a content array with text objects, but LLMs are not told what fields to expect in responses. Without documented return types, downstream tool chaining and output parsing is guesswork.
Parameter description lacks format and usage context. 'Message to echo back. If not provided, returns a default greeting.' does not clarify max length, character restrictions, or when echoing vs default greeting is chosen. Should state: 'Free-form string (max 1000 chars). If empty, server returns default greeting "Hello, World!"'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 47 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 31 | - | v1 |
No error recovery guidance. Error handling returns error text but does not guide LLM on what to do next. If an exception occurs, the LLM receives only the exception message with no retry guidance or alternative actions.