Gmail MCP search results blow context; jq the saved file instead

Gmail MCP search_threads with pageSize 50 returns 50-80KB JSON that exceeds the tool output limit; the harness saves it to a tool-results file. Do not retry with smaller pages, just jq the saved file: extract .threads[].messages[] as date|sender|subject|snippet lines, filter out newsletters by sender regex, and pull threadId per subject for targeted get_thread calls. Use messageFormat PLAIN_TEXT on get_thread and grep bodies for URLs with grep -o https?://. For hunting when a public webpage vanished, combine this with Wayback CDX fl=timestamp,statuscode over all captures to get the exact month it flipped from 200 to 404, and check the raw id_ snapshot HTML for API calls that would load content Wayback text would miss.