Loading ToolBench
A MCP server for Mineru document analysis
Single tool with adequate naming but significant gaps in description depth and schema documentation. Tool name 'analyze_document_mineru' follows verb_noun convention. Description is 60 chars, below the 194-char baseline for A+ tools. Input schema visible with proper types and enums, but output schema is undocumented (returns string, no structure). No error handling guidance for LLM recovery. No parameter descriptions for 'file_url' beyond type. Missing idempotency hints, timeout behavior, and rate-limit guidance despite calling external API.
Analyze document using Mineru v4 API and return content.
Output schema undocumented. Tool returns string but LLM cannot infer structure (JSON vs markdown). No guidance on field extraction or downstream tool chaining.
Tool description too brief (60 chars vs 194-char baseline). Lacks WHEN to use, prerequisites (MINERU_TOKEN env var), and what to expect. LLM cannot determine if this is the right tool for document analysis.
Parameter 'file_url' lacks description. No guidance on supported formats (PDF, etc.), URL validation, or size limits. Parameter 'output_type' has enum but no explanation of when to use 'json' vs 'fullmd'.
Error handling returns plain strings ('Error: MINERU_TOKEN...', 'Create Task Failed...') without recovery guidance. LLM cannot determine if error is retryable, user-fixable, or fatal.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
No idempotency or timeout documentation. Tool polls for up to 20 minutes (600 retries × 2s). LLM unaware of long-running nature; may timeout or retry unnecessarily. No dry-run or confirmation for destructive operations (though this is read-only).