Scoring was not performed

• Document the complete output schema for each tool. E.g., list_tabs should return [{tabId: number, title: string, url: string}, ...]. extract_active_tab should specify whether content is returned as a single string, object with {title, content, url}, or other structure.
• Convert 'format' parameter to an enum: {"type": "string", "enum": ["markdown", "json", "mcp"], "description": "Output format. markdown: readable prose. json: structured object. mcp: MCP-formatted response. Defaults to markdown."}
• Add error recovery guidance to each tool description. E.g., 'If extraction fails due to CSP or auth, the response will include an error_code and suggested_action. Common codes: csp_blocked (suggest list_tabs to pick another tab), auth_required, timeout_exceeded (retry).'
• Expand parameter descriptions with type hints and examples: 'tabId: number. The Chrome tab ID. Obtain from list_tabs() response. Example: 1234'
• Add a section to extract_url description explaining the timeout behavior and data size limits. E.g., 'Opens URL, waits up to 10s for load, extracts content up to 5MB, then closes the tab. Returns null if timeout.'
• Document idempotency: confirm whether calling extract_active_tab twice with the same format returns identical results, or whether state changes.
• Add an example of the combined batch output format (Markdown with separators, JSON Lines) since site/app/guides/batch/page.js references it.