1. Gather job run information from dbt Cloud logs or the MCP Admin API.
2. Classify the error type as infrastructure, code/compilation, or data/test failure.
3. Investigate the root cause based on the error type, checking warehouse status, git history, and data.
4. If the root cause is found, create a new branch and implement the fix.
5. Add a test to prevent recurrence of the issue.
6. Create a pull request with a description of the issue, root cause analysis, and fix.
7. If the root cause is not found, create a findings document outlining the investigation process and next steps.