Cvm Troubleshooter

Diagnose and unblock CVM pipeline failures quickly. Use when ingest, dbt, train, score, e2e, or drift commands fail; when outputs are missing; or when reason quality/feature availability issues appear.

majiayu000 005eaed 2 files · 1.9 KB Updated 567 repo stars

File contents

CVM Troubleshooter

Collect diagnostics first, then propose the smallest rerun.

Workflow

  1. Collect diagnostics using the bundled script.
  • bash skills/cvm-troubleshooter/scripts/collect_diagnostics.sh
  1. Classify failure by stage.
  • Ingestion failures: raw data path, missing raw_*.csv, mapping errors.
  • Prepare-data failures: dbt binary/path, dbt test failures, mart contracts.
  • Train failures: missing target columns, empty training marts, model backend deps.
  • Score failures: missing decision queue, missing signal tables, reason quality gate.
  1. Propose exact fix + rerun command.
  • Do not propose broad changes. Give one minimal change and rerun only affected stage.
  1. Return a concise incident report.
  • include root cause, fix, and verification output path.

Playbook

Use references/failure-playbook.md.

majiayu000/claude-skill-registry-data/tree/main/data/cvm-troubleshooter commit 005eaed684

Frequently asked questions

npx skillmds add majiayu000/cvm-troubleshooter