MCP server for HR leave management with employee leave balance tracking, leave application, and history management
LeaveManager has basic tool definitions but significant quality gaps. All 4 tools have descriptions and input schemas, but descriptions are generic and lack LLM-optimization guidance. Parameters lack detailed constraints (e.g., date format, employee_id format). Output schemas are undocumented, responses are plain strings rather than structured objects. Error handling is minimal (only 'not found' cases). The get_greeting tool is off-domain (greeting vs leave management). No tool annotations, no pagination support, no recovery guidance in errors.
Apply leave for specific dates (e.g., ["2025-04-17", "2025-05-01"])
Get a personalized greeting
Check how many leave days are left for the employee
Get leave history for the employee
Output schemas undocumented. All tools return plain strings instead of structured objects. LLMs cannot extract typed fields for downstream tool chaining.
Parameter descriptions lack format constraints. 'employee_id' and 'leave_dates' have no format specification (e.g., 'ISO 8601 date string', 'employee ID format: E###'). LLMs will guess at valid formats.
Error handling provides no recovery guidance. 'Employee ID not found' does not suggest next steps (e.g., 'Try searching available employees'). Errors are bare strings, not structured with actionable guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
apply_leave lacks confirmation/dry-run pattern. Destructive write operation (deducts leave balance) has no confirmation step. Agents cannot preview impact before committing.
get_greeting is off-domain. A greeting tool does not belong in a leave management system. Dilutes tool clarity and wastes agent reasoning cycles.