Safari MCP server has 13 tools with consistent naming (verb_noun pattern) and complete input schemas. However, descriptions are minimal (10-50 chars, well below the 194-char baseline), parameter descriptions lack depth, and output schemas are entirely undocumented. No error handling guidance, no recovery paths, and no tool annotations. Naming is clear but descriptions fail to explain WHEN to use each tool or what it returns, forcing LLMs to guess. This is typical of community servers that prioritize schema completeness over LLM-optimized guidance.

• Expand tool descriptions to 100-200 chars. For each tool, answer: What does it do? When should the LLM call it? What does it return? Example: 'Get browser console logs (errors, warnings, info) captured since session start. Use this to debug page behavior or find error messages. Returns array of {timestamp, level, message}. Optional: filter by logLevel (DEBUG|INFO|WARNING|SEVERE) or text search.'
• Document output schemas for all tools. Add a 'Returns' section to each description or a separate outputSchema field. E.g., safari_get_console_logs returns [{timestamp: string, level: string, message: string}]; safari_take_screenshot returns {data: base64, mimeType: 'image/png'}.
• Add tool annotations to the tool definitions. Mark safari_clear_console_logs and safari_clear_network_logs with destructiveHint=true; mark all get_* and list_* tools with readOnlyHint=true; mark safari_start_session with idempotentHint=true if re-calling with the same sessionId is safe.
• Enhance parameter descriptions with constraints and examples. E.g., sessionId: 'Unique session identifier (alphanumeric, 1-64 chars). Created by safari_start_session. Invalid IDs return error 'Session not found'.' For logLevel enum, add: 'Filter logs by severity. ALL returns all levels; SEVERE returns only errors.'
• Add error handling guidance to tool descriptions. E.g., safari_navigate: 'If navigation fails (timeout, invalid URL), returns error with reason. Retry with a different URL or call safari_close_session and start a new session.' This guides LLM recovery behavior.