Loading ToolBench
MCP Server for generating DOCX documents from text input with BHK formatting support. Integrates with Mistral agents to create formatted Word documents and provides download links.
This MCP server exposes 2 tools with significant gaps in definition quality. Tool naming is descriptive (verb_noun style), but critical issues exist: (1) generate_docx_document lacks comprehensive input validation guidance in the schema, (2) parameter descriptions are minimal, (3) output schema is not formally documented, (4) error handling is generic without recovery guidance, (5) no tool annotations (readOnlyHint/destructiveHint). The generate_docx_document tool returns plain text messages rather than structured JSON, violating the response-shaper pattern. check_service_health is a READ_ONLY tool but has no documentation of what constitutes failure or what the user should do if the service is unhealthy. Both tools lack the specificity and LLM-optimization baseline of 50-200 character descriptions.
Überprüft ob der DOCX Generator Service betriebsbereit ist.
Generiert ein DOCX-Dokument aus dem bereinigten Diktat-Text und gibt einen Download-Link zurück. Dieser Service erstellt formatierte Word-Dokumente für juristische Texte. Verwenden Sie dieses Tool SOFORT, wenn der Nutzer die Erstellung eines Word-Dokuments wünscht.
Missing structured output schema documentation for both tools. generate_docx_document returns plain text (markdown-like format) instead of structured JSON with typed fields (filename, download_url, expires_at). check_service_health returns free-form text status.
generate_docx_document parameter 'text' lacks constraints and validation guidance. Description only says 'Der Text für das Dokument (Pflichtfeld)', no mention of accepted length, required format, character restrictions, or what constitutes invalid input. LLMs cannot infer these constraints.
Error handling is generic and non-actionable. generate_docx_document returns 'Fehler bei der Dokumentgenerierung' without telling the LLM what to do next (retry? ask user for different input? check service health first?). No error classification or recovery guidance.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 47 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 33 | - | v1 |
check_service_health lacks meaningful output schema. Returns text like '✅ Service ist betriebsbereit', no structured indication of disk space, memory, or specific failure reasons. LLM cannot reason about what's actually wrong.
No tool annotations declared (readOnlyHint, destructiveHint, idempotentHint). generate_docx_document is destructive (creates files, schedules deletion) and creates side effects, but the tool definition does not advertise this.
Descriptions are below optimal length for LLM decision-making. generate_docx_document description is ~180 chars (acceptable), but check_service_health is only ~59 chars, below the 100-char minimum. Baseline for A+ tools is 50-200 chars; these descriptions lack strategic context.
No pagination or result limit controls documented. generate_docx_document always generates exactly one file, but if the tool were to support batch operations, it would need pagination. Current single-file design is adequate but lacks future-proofing.