MCP server for reviewing financial documents against regulatory frameworks (MiFID II, SEC 17a-4, FINRA 4511) and providing compliance findings with remediation priorities
Single tool 'execute_skill_action' has a reasonable name (verb_noun pattern) and adequate description (95 chars, within 10-1024 range). Input schema is present with three parameters, all typed and described. However, critical gaps severely limit production readiness: (1) NO output schema documented, LLMs cannot plan downstream calls or extract results; (2) NO error handling guidance, failures return nothing actionable; (3) NO parameter constraints (enums, ranges, patterns) despite 'regulation_frameworks' accepting only specific values (MiFID_II, SEC_17a-4, FINRA_4511) and 'institution_type' accepting only known types; (4) tool definition is inferred from client.py mock implementation, not explicitly registered in mcp_server.py, cannot verify actual MCP registration. The mcp_server.py prints a static JSON response with no actual tool invocation logic, suggesting this is a stub/demo, not a working implementation.
Execute compliance review action for financial documents against regulatory frameworks
No output schema documented. LLMs cannot infer what fields the tool returns (compliance_findings, overall_compliance_score, remediation_priority, etc.), forcing them to guess or make wasteful follow-up calls.
Input parameters lack enum constraints despite accepting only specific values. 'regulation_frameworks' should declare enum ['MiFID_II', 'SEC_17a-4', 'FINRA_4511']; 'institution_type' should declare enum ['investment_bank', 'broker_dealer', 'advisor']. Free-form strings invite hallucinated invalid values.
No error handling guidance. Tool description does not explain what happens on invalid input, missing document, or unsupported regulation. No recovery hints for LLM (e.g., 'If regulation not recognized, try: MiFID_II, SEC_17a-4, FINRA_4511').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 41 | 2026-07-28+ | v2 |
Tool definition not explicitly registered in mcp_server.py. The mcp_server.py only prints a static JSON response; actual tool invocation logic is missing. Tool existence is inferred from client.py mock, not verified in the MCP server implementation.
Parameter 'document_text' description lacks format guidance. Should specify: expected length, encoding (UTF-8), supported file types (PDF text extract, plain text), or maximum size. LLMs may pass binary data or oversized documents.