Server has 9 tools with clear verb-noun naming (figma_get_*, figma_execute_code, figma_job_status, figma_rollback, figma_list_files). All tools have descriptions (10-100 chars, mostly 50-80 range). Input schemas are present and properly typed with enums for format/scale parameters. However, descriptions are terse (e.g., 'Fetch the full Figma file structure' lacks WHEN/WHY context). Output schemas are not documented, LLMs cannot predict response structure. Error handling is absent from tool definitions. The figma_execute_code tool is high-risk (arbitrary JS execution) but lacks confirmation/dry-run pattern. Parameter descriptions are minimal; 'format' enum is documented but 'scale' lacks range constraints. No pagination guidance for list_files despite potential large result sets.

• Document output schemas for all tools. Example: figma_get_file returns {file_id: string, name: string, pages: [{id: string, name: string, nodes: [...]}], ...}. Include field types and descriptions.
• Expand tool descriptions to 80-150 chars with WHEN/WHY guidance. E.g., 'Fetch a specific node from a Figma file by ID. Use this when you know the node ID; use figma_get_file to explore the full structure first.'
• Add confirmation step to figma_execute_code: require explicit user approval before executing arbitrary JS. Return a 'requires_confirmation' response with a summary of what will execute.
• Add range constraints to numeric parameters. E.g., scale: {type: 'number', minimum: 0.1, maximum: 4.0, description: 'Export scale factor (0.1 - 4.0, default 1)'}.
• Add pagination to figma_list_files: limit (1 - 100, default 20), offset, and return total_count. Document in description: 'Returns up to 20 files per call; use offset to paginate.'