Here's your **ETL Pipeline Auditor** — a fully interactive chat interface powered by the auditor system prompt you provided. https://claude.ai/public/artifacts/ac2164a6-67b2-484d-9148-bb94c6400be2 **What it does:** The auditor runs as a persistent AI-powered session that follows the full 8-phase analysis protocol. You can feed it your pipeline in three ways: - **Tree-First (Mode 1)** — paste the output of `tree` or `find` and it will identify the 5–10 highest-priority files to request next - **Manual Paste (Mode 3)** — paste one or more file contents as code blocks for direct, line-level analysis - **Hybrid (Mode 4)** — mix both; it automatically separates `{Direct}` findings from `{Speculative}` ones **What comes back:** Each analysis response follows the full structured format — operating mode declaration, granularity declaration, dependency map, coupling metrics table (Ca/Ce/I/Fan-In/Fan-Out/Blast Radius), provenance-tagged findings, refactor action plan with risk tiers, and pipeline execution impact notes. **Quick-start suggestions** — four pre-filled prompts on the landing screen — and **follow-up chips** after each response let you drill into priority files, request concrete refactor code, or ask specifically about HIGH-risk changes.