USWDS MCP has 9 tools with complete input schemas and descriptions. Naming follows verb_noun convention (search_, get_, generate_, validate_, recommend_). Descriptions are present (avg ~100 chars) but lack LLM-optimized guidance on WHEN to use each tool vs alternatives. Parameters are typed with Zod but descriptions are minimal (e.g., 'Component slug or name' lacks format/constraint detail). Output schemas are not documented, LLMs cannot predict response structure. Error handling returns JSON but lacks recovery guidance. No tool annotations (readOnlyHint, destructiveHint). Composition is sound: tools are single-responsibility and chainable (search → get_component). Security is adequate for read-only tools.

• Add explicit output schema documentation to each tool. Define the structure of 'results', 'findings', 'summary', and nested objects. Use TypeScript interfaces or JSON Schema in comments.
• Enhance parameter descriptions with format/constraint detail. E.g., 'slug_or_name: Component slug (lowercase, hyphen-separated, e.g. text-input) or display name (e.g. Text Input). Both are accepted and case-insensitive.'
• Add readOnlyHint: true to all 9 tool schemas to signal safety to agents and enable filtering for read-only-only agent configurations.
• Improve error messages with recovery guidance. E.g., 'Component not found: button. Try search_uswds(query="button") to find similar components, or call get_component("text-input") for a known component.'
• Clarify tool selection in descriptions. E.g., recommend_uswds_structure: 'Use this to plan site structure and component hierarchy. For HTML generation, use generate_uswds_page.'