A backup is tested when its exported data opens
A successful archive command only proves that an archive was written. Test a small, non-sensitive export in a separate directory and compare its contents. Our accompanying text fixture was actually archived, restored and read locally; it does not test your application or backup service.

1. Create an export through the application’s documented route. Keep the original intact and record the application/version and export format. A database copied while it is being written may not be a valid application export.
2. List the archive before restoring. Reject unexpected absolute paths, parent-directory traversal, links and unrelated files. Restore only into a new empty test directory; never overwrite the working profile as an experiment.
3. Open the restored data with the intended reader. Check one specific record, Unicode characters and attachments where applicable. Compare hashes as a byte check, but also verify that the reader can interpret the format.
4. Record what was tested and what remains unknown: versions, encryption keys, external dependencies and a full-service recovery may be separate exercises. Keep the test evidence with the export and choose the next review date.

Actual fixture: A UTF-8 text record with a title, “café”, “研究” and the line value=42. No user profile or private data is included.
Executed check: Python standard-library tarfile: create .tar.gz, inspect members, extract the one expected regular file into a new directory, compare SHA256, read UTF-8, assert value=42.
Pass criteria: Exactly one expected regular member; no links or unsafe path; restored bytes equal source; text decoding and the record assertion succeed.
Scope limit: This proves the supplied tiny text export round trip. It does not prove a browser profile, NixOS system, encrypted backup or application database can be recovered.
Your restore record: Export tool/version __; format __; archive members __; isolated target __; opened record __; attachment check __; missing key/dependency __; next test __.

Actual fixture: ____________________
Executed check: ____________________
Pass criteria: ____________________
Scope limit: ____________________
Your restore record: ____________________
Download the worksheet and reproduce the small fixture before adapting the procedure. Keep a failed open as a failure even when the checksum matches: the format or reader may be wrong.

Python — tarfile extraction and inspection — https://docs.python.org/3/library/tarfile.html