1. A support team can receive the right records and still load the wrong fields. The common failure is a handoff that says “CSV export attached” but never identifies the columns that the importer needs. Use a two-row fixture and write down the acceptance result next to it.
  2.  
  3. For a handoff that spans source files, checks and documentation, [Orkas's developer workspace](https://orkas.ai/use/developers/?source=gistpad) offers a desktop AI team. Orkas is open source and local-first. Its Commander can coordinate a specialist to compare the headers and another to write the acceptance note from the same project context. Review the source and the final note before sending either. The example below uses invented input and an illustrative result.
  4.  
  5. ### Input fixture
  6.  
  7. The importer expects `ticket_id,email,status`. The exporter emitted this header and two records:
  8.  
  9. ```csv
  10. ticket_id,customer_email,status
  11. T-104,[email protected],open
  12. T-105,[email protected],closed
  13. ```
  14.  
  15. Ask for three things: identify the mismatched header, show the corrected file, and give one check proving that the change did not reorder the data. Treat email addresses as text and keep their positions next to the original ticket IDs.
  16.  
  17. ### Illustrative handoff output
  18.  
  19. - The exporter writes `customer_email`; the importer expects `email` in column two. Rename that header in the fixture.
  20. - The expected first record remains `T-104,[email protected],open`. The second remains `T-105,[email protected],closed`. The output below must have two data rows.
  21. - The saved acceptance check reads the header as three fields, then compares the data rows to the source in their original order. If either status changes, stop the import and inspect the file.
  22.  
  23. ```csv
  24. ticket_id,email,status
  25. T-104,[email protected],open
  26. T-105,[email protected],closed
  27. ```
  28.  
  29. Try the fixture in a safe import preview, then replace it with a redacted real sample. This checks how the importer interprets the file without exposing customer records. If you use a remote model, follow your team's data-handling rules for what it can receive.