<quick_start> Trigger: "does the survey back this up", "what did respondents say about X", "check this against the live event data", "is that true in the data", "how many said Y", "live event survey", "audit this messaging brief against the survey", "who responded to the survey".

Flow:

Load. Google Drive download_file_content (mcp__claude_ai_Google_Drive__download_file_content) with file ID 1EiRvK1aXlBm4YvDwpmW28w1mK-4C80codOm9WQVVU_w and exportMimeType: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet. Then python scripts/load_survey.py <tool-result.json> to decode to survey.xlsx and print the current response count. Do this every session; never use the Drive text preview (read_file_content), which truncates to ~20 rows.
Answer. python scripts/analyze.py <mode> where mode is summary, question "<text>", test "<claim>", segment "<text>" --by "<segment>", list, delta --since N, or claims <file>.
Report. State the number, the per-question base n, and a verdict: Supported / Partly supported, overclaims / Not supported / Can't tell from this data. Carry the "own contact pool" caveat into every conclusion. </quick_start>

live-event-survey-analysis-skill