I need to continue my systematic data quality check. I've found that deal_stage aligns perfectly with nulls (Prospecting and Engaging have no close data, Won and Lost do), and there are no impossible date orders. Now I need to check for outliers and impossible values in numeric fields across all tables. {"action": "tool_call", "reasoning": "Check for outliers and impossible values in sales_pipeline.close_value. I'll look for negative values, zeros in won deals, and extreme outliers using percentiles.", "tool": "query_sql", "source": "raw", "args": {"sql": "SELECT MIN(close_value) as min_val, MAX(close_value) as max_val, AVG(close_value) as avg_val, MEDIAN(close_value) as median_val, STDDEV(close_value) as stddev_val, PERCENTILE_CONT(0.01) WITHIN GROUP (ORDER BY close_value) as p01, PERCENTILE_CONT(0.99) WITHIN GROUP (ORDER BY close_value) as p99, COUNT(CASE WHEN close_value = 0 THEN 1 END) as zero_count, COUNT(CASE WHEN close_value < 0 THEN 1 END) as negative_count, COUNT(CASE WHEN close_value > 0 AND deal_stage = 'Won' THEN 1 END) as won_with_value, COUNT(CASE WHEN close_value = 0 AND deal_stage = 'Won' THEN 1 END) as won_with_zero FROM raw.sales_pipeline WHERE close_value IS NOT NULL", "purpose": "Check for outliers, impossible values (negatives), and suspicious patterns (zero-value won deals) in close_value"}}