Playwright MCP has minimal tool definitions with critical gaps. All 6 tools have names and basic schemas, but descriptions are extremely terse (10-30 chars, well below the 50-200 char baseline for LLM-optimized tools). Parameter descriptions are sparse or missing entirely. No output schemas documented. Error handling is generic. The server lacks composition guidance and parameter constraints (enums, ranges). Tool names like 'browser_close' and 'browser_wait' are generic and don't clearly convey intent to LLMs. No tool annotations (readOnlyHint, destructiveHint) despite clear WRITE vs READ_ONLY risk classifications in the metadata.

• Expand all tool descriptions to 50-200 chars. Example: 'browser_close' → 'Close the current browser page and release resources. Call this when done with the page to clean up memory and connections.'
• Document output schemas for all tools. Example: browser_console_messages should return {messages: [{level: string, text: string, timestamp: number}], count: number}.
• Add parameter descriptions for all inputs. Example: 'time' in browser_wait should be 'Duration in seconds (1-3600). Used to pause execution for network delays or animations.'
• Implement tool annotations in server.ts setupRequestHandlers(). Map WRITE tools to destructiveHint=true, READ_ONLY tools to readOnlyHint=true. This requires MCP SDK 1.10+ capability support.
• Add input validation and error guidance. Example: browser_file_upload should validate paths exist and return 'File not found: /path/to/file. Check the absolute path and permissions.' instead of generic errors.