TideSurf provides 20 browser automation tools with basic schemas and descriptions, but falls short of production quality. Tool names are action-verb based (navigate, click, scroll) which is good. However, descriptions are minimal (10-50 chars typically), parameter descriptions are sparse or missing, and output schemas are not documented. Many tools lack detail on what they return, making it difficult for LLMs to chain calls. Error handling guidance is absent. The server exposes a JavaScript evaluation tool (evaluate) without clear security boundaries or input validation guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications in the metadata.

• Expand tool descriptions to 100-200 chars. Include WHEN to use (e.g., 'click: Activate a button or link by node ID. Use after get_state to identify interactive elements.') and WHAT it returns.
• Document output schemas for all tools. E.g., get_state returns {dom: string, nodeMap: {[id]: {tag, text, attrs}}, ...}; screenshot returns {data: base64, format: 'png', ...}.
• Add tool annotations: mark read-only tools (screenshot, get_state, search, classify, wait) with readOnlyHint=true; mark destructive tools (navigate, click, type, fill, select, press, hover, focus, clear, evaluate, download, back, forward, reload) with destructiveHint=true.
• Enhance parameter descriptions with constraints and examples. E.g., 'id: Node ID from get_state nodeMap (string, 1-50 chars, alphanumeric+underscore)' and 'expression: JavaScript expression (no fetch/eval/import; max 500 chars; returns JSON-serializable value)'.
• Add error recovery guidance. E.g., 'If element not found, call get_state first to verify node ID exists. If node ID is stale, re-run get_state to refresh the DOM map.'