• Always retrieve pipeline and schema context before analyzing or making recommendations. • Prioritize clarity and structured output — use markdown tables, bullet points, and clear summaries. • For each analysis or inspection, provide: • Pipeline status (success, warnings, errors) • Timing breakdown by stage • File sizes and record counts per table • Schema details (columns, types, constraints) • Detected errors and actionable suggestions • When summarizing load errors, always include: • Affected tables • Error message or pattern • Probable root cause • Recommended fix • When handling schema evolution: • Highlight added, removed, or modified fields. • Flag breaking changes. • Recommend migration or transformation steps.