Verifies citations and checks the provenance of AI-generated responses against provided sources
This MCP server has severe quality gaps across definition dimensions. The single tool 'verify_citations' has a basic description but lacks critical schema details, parameter descriptions, and output documentation. The implementation shows a mock client that returns hardcoded values rather than actual verification logic. No input validation, error handling, or structured output schema. The tool definition is visible but incomplete, parameters lack type information in the schema, descriptions are minimal, and output structure is not formally documented.
Verifies citations and checks the provenance of AI-generated responses against provided sources
Input schema lacks parameter type definitions and descriptions
Tool description is too generic and does not explain when or why to use it
No output schema documented, LLMs cannot plan downstream tool calls or extract results
No error handling or recovery guidance, what should the LLM do if verification fails?
Implementation returns hardcoded mock values instead of performing actual verification
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | 2026-07-28+ | v2 |
Parameters 'ai_response_text' and 'citations' lack descriptions explaining their purpose and format
No validation of input, citations array format is not constrained (are they URLs, IDs, or free text?)
Return type missing, no formal documentation of 'overall_accuracy_score', 'verified_citations_count', 'unverified_claims' structure