The server has 5 tools with basic structure, but significant gaps in descriptions and parameter documentation. All tools are properly named with verb_noun convention (add-task, list-tasks, complete-task, remove-task, update-task). Input schemas are present but minimally documented, parameter descriptions exist but are terse. Output schemas are not documented. Error handling and recovery guidance are absent. The server lacks idempotency documentation and does not distinguish between retryable vs destructive operations in error responses.
Add custom metadata (key-value pairs) to a task by ID.
Add a new task to the todo list. Description must be plain text. Priority must be a single uppercase letter (A-Z). Contexts must start with @, projects with +.
Mark a task as completed by its 1-based ID.
Delete a task by its 1-based ID.
Filter tasks by specific criteria (priority, context, project).
List all tasks, optionally filtered by priority, context, project, or metadata.
Search for tasks containing a query string.
Sort tasks by priority, creation date, or completion date.
All tool descriptions are critically under-length (8-33 chars vs 194 char baseline). Examples: 'Add a new task to the todo list' (31 chars), 'List all tasks from the todo list' (33 chars). These descriptions provide minimal LLM guidance on WHEN to call each tool, WHAT it returns, or HOW it differs from similar tools. Per hard scoring rule, descriptions under 20 characters cannot exceed 40 points.
Missing output schema documentation. Source code shows tools return results (e.g., 'Task added successfully', list of tasks), but no schema is visible defining output structure, fields, or types. LLMs cannot predict downstream field availability or plan chained tool calls without documented return types.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 0 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 30 | - | v1 |
No error handling or recovery guidance. Tools that perform destructive operations (remove-task, complete-task) have no documented error cases, recovery paths, or confirmation patterns. Per pattern:confirmation-request, irreversible operations should support dry-run or confirmation. No error responses visible that guide the LLM what to do next (e.g., 'Task index 5 out of bounds, list-tasks first to verify available indices').
Minimal parameter descriptions. The 'task' parameter in add-task is described as 'The task description to add', and 'taskIndex' in complete-task/'remove-task' as 'The index of the task to complete/remove', these are terse and lack format guidance. No mention of valid range for taskIndex (0 to count-1?), character limits for task strings, or what happens on edge cases.
No idempotency documentation. Tools like complete-task and remove-task do not state whether they are idempotent (calling twice with same input). Agents retry on ambiguous failures, if remove-task on index 3 fails and the agent retries, will it delete a different task due to index shift? This is unspecified and violates pattern:idempotent-operation.
No tool annotations for risk levels. The metadata shows WRITE and DESTRUCTIVE risk labels, but source code does not show destructiveHint or readOnlyHint annotations in tool registration. MCP supports toolAnnotations for clients to highlight risky operations, this server does not implement them, preventing clients from warning users before execution.
Numeric parameter 'taskIndex' lacks range/constraint documentation. No description states valid range (0 to list size?), what happens if index is negative, out of bounds, or a float. Per pattern:constrained-input, numeric parameters must declare min/max and format. Free-form numbers invite LLM errors.