Single tool with critical definition gaps. Tool name 'evaluate_policy_compliance' is acceptable (verb_noun pattern), but input schema is completely absent from the MCP registration. The tools/list response shows only name and description, no input schema, parameter types, or descriptions. The client.py implementation reveals parameters (text_content, category) exist, but they are NOT exposed in the MCP protocol layer. Description is 95 chars (within 10-1024 range) but generic and lacks WHEN/WHY context. No output schema documented. No error handling guidance. No tool annotations. STDIO transport caps protocol readiness at 50.

• Add inputSchema to tools/list response with full JSON Schema: {"type": "object", "properties": {"text_content": {"type": "string", "description": "Advertising or e-commerce text to evaluate (max 5000 chars). Defaults to example if omitted."}, "category": {"type": "string", "enum": ["Consumer Electronics", "Health & Wellness", "Financial Services", "Food & Beverage"], "description": "Product category for compliance context. Defaults to 'Consumer Electronics'."}}, "required": []}
• Expand tool description to 150-200 chars: 'Evaluates advertising and e-commerce text for FTC/FDA regulatory compliance. Detects prohibited medical claims, unsubstantiated guarantees, and trademark abuse. Returns compliance score (0-100) and remediation actions. Use before publishing marketing copy.'
• Document output schema in description or as a separate field: {"evaluation_id": "string", "category": "string", "compliance_score": "number (0-100)", "compliance_status": "enum: COMPLIANT_APPROVED|FLAGGED_MODERATION_REQUIRED|REJECTED_POLICY_BREACH", "violations_detected_count": "number", "remediation_actions": "array of {violation_category, flagged_token, severity, remediation}"}
• Add input validation and error responses: if text_content exceeds 5000 chars, return {"error": "Text too long (max 5000 chars)"}. If category is invalid, return {"error": "Invalid category. Must be one of: Consumer Electronics, Health & Wellness, Financial Services, Food & Beverage"}.
• Add tool annotations to indicate read-only behavior: include "readOnlyHint": true in tool definition to signal this tool does not modify state and is safe to call multiple times.