MCP server for managing tasks with add, retrieve, update, and delete operations
This MCP server provides 7 task management tools with solid naming conventions and mostly complete schemas. All tools follow verb_noun naming (add_new_task, get_task_by_id, etc.) which is excellent. Tool descriptions are present and reasonably clear. However, there are significant gaps: (1) Output schemas are completely undocumented, tools return plain strings instead of structured objects, forcing LLMs to parse free-text responses; (2) parameter descriptions are minimal in some cases; (3) no error handling guidance or recovery paths; (4) destructive operations (remove_task, clear_completed_tasks) lack confirmation patterns; (5) response output does not follow the Structured Output pattern, all tools return formatted strings rather than JSON objects with typed fields. The server lands in the Fair (C) to Poor (D) range due to these structural gaps, despite solid naming.
Add a new task to the to do list
Remove all completed tasks from the list.
Mark a specific task as completed using its ID.
List tasks from the database. Optional status filter: 'pending' or 'completed'
Retrieve a single task by its ID.
Delete a task from the list permanently.
Update the title or description of an existing task.
Output schemas completely undocumented. All tools return plain strings (e.g., 'Task created with ID: xyz' or formatted multi-line text) instead of structured JSON objects with typed fields. LLMs cannot extract data reliably and must parse free-form text, wasting tokens and inviting errors.
Destructive operations lack confirmation or dry-run patterns. remove_task and clear_completed_tasks execute immediately with no chance for the agent to review or confirm before data is permanently deleted.
No error guidance or recovery paths. When operations fail (task not found), the tool returns plain error strings like 'Error: Task with ID xyz not found.' LLMs receive no actionable guidance on what to do next, should they retry, call search, or ask the user?
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 52 | <=2025-11-25 | v2 |
| 2026-03-09 | D | 57 | - | v1 |
get_all_tasks accepts an enum status_filter but does not support pagination (limit, offset, page). Returning all matching tasks in one call risks exploding output and context window. Rubric baseline expects pagination for list tools.
Parameter descriptions are sparse. 'task_id' descriptions say 'The ID of the task to retrieve', acceptable but minimal. No guidance on format or how LLMs should obtain task_id values. Title and description in add_new_task lack usage guidance.
clear_completed_tasks has no input parameters (empty schema). While functional, the return type is also undocumented string, making its output semantics opaque to LLMs.