Loading ToolBench
Enterprise SDK traits for ADK-Rust MCP servers — health checks, tool metadata, risk classes, credential binding, and registry compatibility
This is an SDK/framework, not a production MCP server. The 5 test tools have minimal descriptions (10-30 chars), no parameter descriptions, and incomplete schemas. Only 'echo' has a visible input schema with a typed parameter. The 'danger', 'json_status', 'json_error', and 'read' tools lack input schemas entirely in the source. Descriptions are too brief to guide LLM tool selection. No error handling guidance, no output schema documentation, and no security/permission metadata visible. The macro-generated server structure obscures actual tool registration details.
Protected identity test
Echo a value
Return a legacy JSON text error
Return a legacy JSON text result
Read a value
Four of five tools (danger, json_status, json_error, read) have no visible input schema in source code.
All tool descriptions are under 30 characters and lack context for LLM selection. 'Echo a value', 'Protected identity test', 'Return a legacy JSON text result' do not explain WHEN to use each tool or what it returns.
No parameter descriptions visible for any tool. The 'echo' tool has a 'value' parameter with type 'string' but no description explaining what kind of value it accepts or what the tool does with it.
No output schemas documented. LLMs cannot infer what fields to expect from tool responses. The test code shows tools return strings, but no structured schema is declared.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2026-07-28+ | v2 |
Tool names 'json_status' and 'json_error' are vague and do not start with action verbs. 'json_status' could mean 'get_status_as_json', 'check_json_status', or 'return_json_status'. Ambiguous names cause LLM confusion.