Loading ToolBench
An MCP server that provides a unified Todoist assistant capable of creating tasks and fetching task lists based on natural language prompts using LLM intent detection
Single tool 'todoist_assistant' has critical definition gaps. Tool name lacks action verb prefix (should be 'create_or_list_tasks' or split into separate tools). Description is adequate (127 chars) but parameter schema is severely underdeveloped: single 'prompt' string with minimal guidance. No output schema documented. Tool conflates two distinct operations (create + list) violating single-responsibility principle. No error recovery guidance, no parameter constraints, no pagination support for list results. Input validation and error handling exist in implementation but are not reflected in tool definition.
All-in-one Todoist assistant that can create tasks or fetch task lists based on natural language prompts
Tool name 'todoist_assistant' lacks action verb prefix. Should start with verb (create_, list_, search_) to signal intent to LLM. Current name is ambiguous and does not follow verb_noun convention.
Single tool conflates two distinct operations: task creation and task listing. Violates single-responsibility principle. Should be split into 'create_task' and 'list_tasks' so agent can compose them independently.
Input schema is minimal: single 'prompt' parameter with no constraints, enums, or format guidance. LLM must infer intent from free-form text, increasing hallucination risk. No parameter validation rules documented.
No output schema documented. Tool returns unstructured string responses. LLM cannot parse task IDs, due dates, or other structured fields for downstream chaining. Breaks tool composition.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 36 | 2026-07-28+ | v2 |
No pagination support for list_tasks operation. If user has 100+ tasks, all returned in single response, wasting tokens and risking context window exhaustion. No limit parameter or cursor documented.