Server has 9 tools with clear naming (verb_noun pattern) and detailed descriptions (100-300 chars each). All tools have input schemas with type definitions. However, output schemas are not formally documented in the code, descriptions mention return types but JSON Schema output definitions are absent. Parameter descriptions are present and specific. Tool annotations are correctly applied. Main gaps: no pagination support for list tools, no error handling guidance, and some descriptions are overly verbose with implementation details rather than LLM-optimized intent statements.

• Add formal JSON Schema output definitions for all tools. Document return types as structured objects with typed fields. E.g., list_case_types should return {case_types: string[], descriptions: {[key: string]: string}, acronym_to_case_type: {[key: string]: string}}.
• Declare case_type parameter as an enum in get_case_type_format and create_case_folder. Replace 'One of the slugs returned by list_case_types()' with explicit enum constraint to prevent hallucinated values.
• Add error handling guidance to tool descriptions. E.g., create_case_folder: 'Returns the absolute path on success. If case_type is invalid, returns error with valid options. If folder already exists, returns existing path (idempotent).'
• Condense get_agent_instructions and get_case_type_format descriptions to 100-150 chars. Move orchestration logic to a separate 'orchestration_guide' resource or tool. Descriptions should state intent, not implementation.
• Add pagination to list_case_types and list_case_artifacts. Include limit (default 20, max 100) and offset parameters. Return total_count and has_more fields so agents can iterate efficiently.