Loading ToolBench
Trust, reputation, and economic accountability layer for MCP
Two tools with minimal schema documentation and generic descriptions. tools/list lacks input schema entirely (score 0); tools/call has partial schema (name, arguments as strings/objects) but no type constraints, enums, or validation guidance. Descriptions are functional but under 100 chars and lack context on when to use each tool or what happens on failure. No parameter descriptions beyond bare names. No output schema documented. Error handling returns JSON-RPC codes but lacks recovery guidance (e.g., 'trust evaluation failed' does not explain why or how to remediate). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Security model (OAuth + trust evaluation) is present but not reflected in tool descriptions or permission declarations.
Invokes a tool with trust evaluation before proxying to upstream servers
Returns trust-annotated tool list without requiring authentication
tools/list has NO input schema. Tool accepts no parameters but this is not formally declared.
tools/call input schema lacks type constraints and parameter descriptions. 'name' and 'arguments' are documented as string/object but no enum, pattern, or validation rules. LLM cannot infer valid tool names or argument structure.
No output schema documented for either tool. tools/list returns trust-annotated tool list but structure is inferred from code, not declared. tools/call proxies upstream result but response format is opaque to LLM.
Error responses lack recovery guidance. 'Trust evaluation failed' (code -32001) does not explain why (KYA level too low? reputation insufficient? delegation depth exceeded?) or what the agent should do next.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | <=2025-11-25 | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). tools/list is read-only; tools/call is destructive/write. These should be declared in tool metadata for agent safety.