Loading ToolBench
https//.aamardokan.online POS MCP Server, AI automation and Add chatbot services
Server has critical definition gaps. Tool descriptions are present but minimal (10-50 chars, well below 194-char baseline). Input schemas exist for most tools but lack parameter descriptions entirely, LLMs cannot infer what 'pipeline' or 'collection' mean without explicit guidance. No output schemas documented. Error handling is absent; tools throw raw errors with no recovery guidance. Security concerns: createOrder accepts arbitrary orderData without validation. Tool naming is acceptable (verb-first) but parameter naming is inconsistent ('collection' vs 'collectionName'). No tool annotations (readOnlyHint, destructiveHint). Overall, definitions are skeletal, suitable for internal use, not production agents.
Create a new sales order in the database
Get the schema of a collection by retrieving the first document
List all collections in the MongoDB database
Run a MongoDB aggregation pipeline on a collection
Validate MongoDB aggregation pipeline for forbidden stages
No parameter descriptions. Input schemas define types but omit descriptions for all parameters (e.g., 'pipeline' has no guidance on structure, format, or constraints). LLMs cannot infer parameter meaning from names alone.
No output schemas documented. Tools return raw MongoDB results (arrays, objects) with no schema guidance. LLMs cannot plan downstream calls or extract required fields (e.g., does listCollections return 'name' or 'collectionName'?).
Minimal tool descriptions (10-50 chars vs. 194-char baseline). Descriptions lack WHAT, WHEN, and WHY context. E.g., 'Validate MongoDB aggregation pipeline for forbidden stages' does not explain when to call it or what happens if validation fails.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
No error handling or recovery guidance. Tools throw raw errors ('Unknown method', 'Pipeline contains forbidden stages') with no actionable next steps for LLMs. No error categorization (retryable vs. fatal).
createOrder accepts unvalidated orderData object. No schema enforcement, no required field checks, no type validation. Agents can pass malformed orders; tool silently accepts them, risking data corruption.