MCP server for evaluating product return and exchange eligibility based on purchase date, current date, and return reason
This MCP server has severe definition quality issues. The single tool 'evaluate_eligibility' has a reasonable input schema and acceptable description, but lacks critical production-grade qualities: no output schema documentation, no error handling guidance, no parameter descriptions, no type hints in parameters, and no tool annotations. The implementation is a bare Python class with no MCP framework integration visible. Tool names lack action verbs (should be 'check_return_eligibility' or 'evaluate_return_eligibility'). Parameters are present in the code but lack formal descriptions in the schema definition.
Evaluates whether a product return or exchange is eligible based on purchase date, current date, and reason. Returns eligibility status, days since purchase, applicable restocking fee, and RMA number if eligible.
No output schema documented. Tool returns a dict with keys 'eligible', 'days_since_purchase', 'restocking_fee', 'rma_number', but LLMs cannot plan downstream actions without knowing the structure and types of returned fields.
Input parameter descriptions missing or trivial. Schema shows parameter types (string, string, string) but no descriptions explaining what 'reason' should contain, what formats are valid, or example values.
Tool name lacks action verb. 'evaluate_eligibility' is passive; should be 'check_eligibility', 'verify_return_eligibility', or 'assess_return_eligibility' to align with LLM action-inference patterns. Action verbs (get, create, update, check, verify) appear in 90% of production tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2026-07-28+ | v2 |
No error handling or recovery guidance. Code has no validation for invalid dates, malformed input, or edge cases. If an LLM passes an unparseable date string, the tool will crash with a ValueError. No guidance for the agent to retry or correct input.
Tool description is 203 characters, acceptable length, but lacks context on WHEN to use this tool, WHAT it returns, and HOW to interpret results. No mention of the 30-day return window, restocking fee conditions, or RMA format.
No tool annotations (readOnlyHint, idempotentHint, etc.). Tool is read-only and idempotent but these properties are not declared. LLMs cannot infer safety properties and may incorrectly reason about retry semantics.
Hardcoded business logic (30-day limit, $9.99 fee, RMA format) has no documentation. LLMs cannot understand the rules and may make incorrect eligibility predictions in real scenarios. Should document these as constraints in tool description.