Loading ToolBench
MCP server for accessing Currents test run data and analytics
Single tool 'get-run' has a basic verb-noun name and minimal description (47 chars). Input schema is present with type but parameter lacks description. No output schema documented. No error handling guidance. Tool returns raw JSON dump without structure or pagination. Missing parameter descriptions, output documentation, and error recovery patterns.
Get run details for a defined run ID
Parameter 'runId' lacks description. LLMs cannot infer whether to pass a UUID, numeric ID, or string identifier without explicit guidance.
No output schema documented. Tool returns raw JSON.stringify() of API response without structure, field descriptions, or pagination info. LLMs cannot plan downstream operations.
Tool description is 47 characters, below the 50-200 char baseline for LLM-optimized descriptions. Does not explain WHEN to use this tool, what data it returns, or prerequisites.
Error handling returns generic 'Failed to retrieve run data' with no recovery guidance. LLM cannot determine if the runId was invalid, the API is down, or the run doesn't exist.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
API credentials hardcoded in source code. Bearer token 'FlLFehSQxlNkBlh7C2X2E1RT43IiHNpF3vD578vVzQJUWHHuvsifXHkkQgTz3mtg' is exposed in repository and will be logged in agent traces.