Critical failures across all 45 tools. No output schemas documented. No error handling guidance. Tool names lack clarity (e.g., 'execute_shell' vs 'run_shell', 'browser_click' vs 'browser_interact'). No parameter descriptions, types, or constraints. This server is not production-ready for agent use.

• Add complete input schemas to all 43 built-in tools. Each parameter must have type (string, number, boolean, array, object), description (50-100 chars), and constraints (enum, minLength, pattern, minimum, maximum). Example: read_file should have 'path' (string, required, 'File path to read'), 'encoding' (string, enum: ['utf-8', 'ascii'], default 'utf-8').
• Replace generic descriptions with LLM-optimized text (50-200 chars). State WHAT the tool does, WHEN to use it, and any prerequisites. Example: 'read_file' → 'Read file contents from disk. Use this to inspect code, config, or data files. Returns full content or error if file not found or unreadable.'
• Document output schemas for all tools. Specify returned fields, types, and structure. Example: list_files should return {files: [{name: string, path: string, size: number, isDir: boolean}], total: number}. This enables agents to chain calls and extract required IDs.
• Add error handling guidance to destructive tools (delete_file, clear_memory, execute_shell, execute_plan). Implement dry-run or confirmation steps. Return actionable errors: 'File not found at /path/to/file. Available files in /path: [list]. Did you mean one of these?' instead of bare 404.
• Add parameter descriptions and type constraints to browser_* and pg_* tools. Example: browser_click needs 'selector' (string, CSS selector), 'timeout' (number, 1-30000 ms, default 5000). pg_query needs 'sql' (string, SQL statement), 'params' (array, query parameters), 'timeout' (number, 1-60000 ms).