Three well-named, read-only tools with clear descriptions (194 - 392 chars, within baseline). All tools have explicit input schemas with typed properties and descriptions. sanctions_screen has comprehensive parameter documentation (name, list, type, country, limit); sanctions_entry documents uid and entity_number; sanctions_lists is minimal but appropriate for a no-arg tool. Output shaping is present (shapeMatch function strips internal fields, caps alt_names to 8, addresses to 3, remarks to 300 chars). Error handling returns structured {error, message} objects with actionable guidance. Main gaps: no output schema documentation in tool definitions, no enum constraints on list/type/country parameters (LLMs may hallucinate invalid values), no pagination support despite potentially large result sets, and no tool annotations (readOnlyHint, idempotentHint) in the schema.

• Add enum constraints to inputSchema for list, type, and country parameters. E.g., 'list': {type: 'string', enum: ['SDN', 'SSI', 'CMIC', ...], description: '...'}. This prevents LLM hallucination and self-documents valid options.
• Document output schemas in tool definitions. Add a 'returns' field or update descriptions to specify: sanctions_screen returns {query, matches_found, clear, note, matches: [{uid, name, score, type, list, list_name, programs, alt_names, country, addresses, title, remarks, source_list_url}]}.
• Add pagination to sanctions_screen: accept offset/page_size parameters (default 10, max 50) and return total_count and next_cursor. This enables agents to iterate through large result sets without context explosion.
• Add tool annotations to inputSchema: set readOnlyHint: true and idempotentHint: true for all three tools. This signals to agents that these calls are safe to retry and do not modify state.
• Validate country parameter against ISO 3166-1 alpha-2 codes in the description. E.g., 'country': {type: 'string', description: '2-letter ISO 3166-1 alpha-2 country code (e.g., CN, RU, IR). Invalid codes return no matches.'}.