The server defines 7 tools with complete input schemas and descriptions in English. However, descriptions are uniformly generic (10-15 words each), lack actionable context about when/why to use each tool, and do not explain error recovery. Parameter descriptions exist but are minimal. Output schemas are not documented. No error handling guidance provided. The tools are read-only with good naming conventions (verb_noun pattern), but lack the depth required for production LLM integration. Naming is clear and distinct across tools. The schema quality is moderate, all tools have proper JSON Schema with types and enums where appropriate, but descriptions are too brief to meet LLM optimization standards (target 50-200 chars; actual ~50-80 chars per tool).

• Expand each tool description to 100-150 characters explaining WHAT it does, WHEN to use it (vs. similar tools), and any prerequisites. Example: 'Export a single Figma node as an image. Use this when you need one asset; use export_multiple_images for batch operations. Requires node-id in the Figma URL. Returns image URL and file info.'
• Document output schemas for all tools. Create a schema definition showing returned fields, types, and meanings. Example for get_file_info: {file_id: string, name: string, owner: {id: string, name: string}, last_modified: ISO8601_string}.
• Add error handling guidance to each tool. Return errors like 'Invalid URL: must match https://www.figma.com/file/{fileId}/... with ?node-id parameter. Got: {actual_url}' instead of raw parse errors.
• Clarify tool distinctions. Merge overlapping tools: consolidate get_figma_image, get_node_images, and extract_node_elements into a single export_node_asset tool with an 'assetType' parameter (image|svg|elements). Or document exactly when each is preferred.
• Add dependency hints in descriptions. E.g., 'If you only have a file URL, call get_file_info first to list available nodes, then pass node-id to this tool.' This prevents wasted calls.
• Document pagination, rate limits, and result caps. If export_multiple_images is limited to 100 nodes per call, say so. If there is a 60-call-per-minute rate limit, document it and suggest batch strategies.
• Specify parameter constraints in descriptions. For 'scale': 'A multiplier from 0.01 to 4. Values outside this range are rejected with error code INVALID_SCALE.' For 'format': 'PNG is default and supports transparency. SVG is vector (smallest). PDF is vector with text (best for print).'