Loading ToolBench
MCP server for creating and manipulating Microsoft Office documents (Word, Excel, PowerPoint, Email) from Markdown and structured data
This server provides 5 document-generation tools with generally good naming (verb_noun pattern) and moderately detailed schemas. However, descriptions are inconsistent in quality and depth, parameter descriptions lack actionable constraints, and output schemas are not documented. Most tools have schemas visible but lack comprehensive parameter guidance. The server addresses a specialized domain (Office doc creation) well but falls short of production-grade quality due to missing output documentation and limited error guidance.
Create an unsent email draft (EML) using a Mustache HTML template
Create a PowerPoint presentation from structured slides and upload it
Create an XML file from the provided XML content. Validates that the content is well-formed XML before saving
Convert Markdown with tables to Excel workbook and upload it
Convert Markdown to Word document, save to memory and upload
Output schemas not documented. Tools return file buffers or upload results, but LLMs cannot see what structure or metadata is returned. This prevents planning downstream operations and extracting critical IDs or references.
Parameter descriptions lack actionable constraints. E.g., 'Header text (optional)' does not specify length limits, formatting, or valid content. 'Slide models' is vague, no description of the expected structure. LLMs cannot validate inputs without explicit constraints.
Error handling and recovery guidance missing. No documentation of what happens on malformed input, file upload failures, or API errors. LLMs receive no direction on whether to retry, ask the user, or skip the operation.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 61 | 2026-07-28+ | v2 |
create_presentation 'slides' parameter description is vague. References 'Slide models, or dicts in the current or previous spelling' without explaining the schema. LLMs cannot construct valid slide structures without detailed field documentation.
create_xml_file description ('Create an XML file...Validates that...') is minimal and does not explain when to use this vs create_eml or other tools, or what happens on validation failure.
create_eml 'priority' parameter uses free-form string instead of enum. Description says 'low, normal, or high' but does not enforce it. LLMs may pass invalid values like 'urgent' or 'medium', causing silent failures or defaults.
No tool idempotency documentation. When called with identical parameters, will these tools produce the same output (same filename, identical content) or generate new files with different names? This affects agent retry logic.