ClimaSafeAI MCP server has critical definition gaps. Tool descriptions are present but generic and lack LLM-optimized guidance on WHEN to use each tool. Input schemas are visible but lack proper type definitions for complex parameters (e.g., 'df' in dataframe_analysis is typed as 'object' with no structure). No output schemas documented. Parameter descriptions are minimal (avg ~40 chars, well below 72-char baseline). No error handling guidance. Tool names lack clear verb-noun structure (e.g., 'cache' and 'cicd' are nouns, not actions). No security considerations visible for write-risk tools (agent_installer, cache, cicd, data_io). Composition issues: agent_installer performs multiple steps (clone, validate, register) that should be separate tools.

• Rename tools with action verbs: 'cache' → 'create_cache' or 'setup_cache'; 'cicd' → 'generate_cicd_workflow'; 'api' → 'extract_api_endpoints'. Follow verb_noun convention.
• Document output schema for every tool. E.g., agent_installer should return {installed_path: string, validation_status: 'valid'|'invalid', registry_entry: object}.
• Expand parameter descriptions to 50 - 150 chars with format/constraint hints. E.g., 'repo_url: Git repository URL or GitHub shorthand (user/repo). Must be publicly accessible or authenticated via SSH key.'
• Add error handling guidance to descriptions. E.g., 'If validation fails, call validate_agent(path) separately to diagnose. Retryable on network errors; fatal on malformed structure.'
• Split agent_installer into three tools: clone_agent(repo_url, destination), validate_agent(path), register_agent(path). Allows agent to handle each step independently and retry selectively.