Three well-named, read-only tools with clear descriptions (150 - 200 chars each) and complete input schemas. All parameters are typed and described. Tool names follow verb_noun convention (search_cases, get_case, cases_by_term). Schemas include proper type constraints and required fields. However, output schemas are not documented, LLMs cannot predict response structure for downstream chaining. Error handling is minimal (generic catch-all). No pagination metadata (total_count, next_cursor) in search results despite accepting limit parameters. Missing tool annotations (readOnlyHint). Parameter descriptions lack format/range details (e.g., term year bounds, docket format pattern).

• Document output schema for each tool. Specify fields returned by mapCaseSummary (id, name, docket_number, citation, term, question, description, oyez_url, href) with types and descriptions. For get_case, document full case object structure (facts, conclusion, decision, justices, advocates, oral_argument_audio).
• Add explicit parameter constraints: term must be 1791 - 2024 (or current year); docket_number must match pattern \d{2}-\d{3}; limit must be 1 - 30 (search_cases) or 1 - 40 (cases_by_term). Include these in parameter descriptions, not just schema.
• Return pagination metadata: include total_count and has_more in search_cases and cases_by_term responses so LLMs know if results are truncated and can plan multi-page retrieval.
• Improve error handling: categorize errors (e.g., 'not_found', 'rate_limited', 'invalid_input') and include recovery guidance. Example: 'Term 2025 not found. Available terms: 1791 - 2024. Try a recent term like 2024.'
• Add tool annotations to schema: set readOnlyHint=true and idempotentHint=true for all three tools to signal safety to agents and enable more aggressive caching/retry strategies.