HireThomas, Inc. — CSV duplicate-flagging execution handoff
FICTIONAL DEMONSTRATION / AI-authored company sample.
Synthetic, nonpersonal data; not client work.

Executed on 2026-09-29 with Python 3.13.15, ARM64, macOS.
The exact four files in csv-duplicates-sample.zip were extracted and tested.
No third-party Python packages were required.

Commands (from the directory containing the four extracted files):
python3 -m unittest -v tests
python3 script.py fixture.csv output.csv
The interpreter above was Python 3.13.15. Choose a new output path.

Unit-test result: exit status 0; all 15 tests passed.
test_cli_fixture_exact_expected_output (tests.FlagCSVTests.test_cli_fixture_exact_expected_output) ... ok
test_duplicate_headers_rejected (tests.FlagCSVTests.test_duplicate_headers_rejected) ... ok
test_existing_output_not_overwritten (tests.FlagCSVTests.test_existing_output_not_overwritten) ... ok
test_flag_header_collision_rejected (tests.FlagCSVTests.test_flag_header_collision_rejected) ... ok
test_formula_leading_values_literal (tests.FlagCSVTests.test_formula_leading_values_literal) ... ok
test_header_only_file (tests.FlagCSVTests.test_header_only_file) ... ok
test_invalid_utf8_cli_error (tests.FlagCSVTests.test_invalid_utf8_cli_error) ... ok
test_missing_and_empty_headers_rejected (tests.FlagCSVTests.test_missing_and_empty_headers_rejected) ... ok
test_quoted_commas_and_newlines (tests.FlagCSVTests.test_quoted_commas_and_newlines) ... ok
test_repeated_rows_retained_in_order (tests.FlagCSVTests.test_repeated_rows_retained_in_order) ... ok
test_same_path_rejected (tests.FlagCSVTests.test_same_path_rejected) ... ok
test_unclosed_quote_rejected (tests.FlagCSVTests.test_unclosed_quote_rejected) ... ok
test_unicode_and_embedded_crlf_cr_preserved (tests.FlagCSVTests.test_unicode_and_embedded_crlf_cr_preserved) ... ok
test_whitespace_case_and_numeric_strings_distinct (tests.FlagCSVTests.test_whitespace_case_and_numeric_strings_distinct) ... ok
test_wrong_row_lengths_rejected_without_output (tests.FlagCSVTests.test_wrong_row_lengths_rejected_without_output) ... ok

----------------------------------------------------------------------
Ran 15 tests in 0.065s

OK

Standalone CLI result: exit status 0; standard error empty.
Standard output:
rows=7 duplicate_groups=2 repeated_rows=2 flagged_rows=4

Actual output.csv matched expected.csv byte for byte (140 bytes).
Seven records retained in original order; two duplicate groups,
two repeated rows beyond the first occurrence, four flagged rows.

SHA-256 of the exact downloadable files:
2774e7d479e05a04795ca23c9c0113b277afb14db9ba60da30cc7e46e4c56923  script.py
b998ae01de70bc3b9205d418434e59e1b2a8e21ac37c4ede229d6d5873a2c0c9  tests.py
c2cd7da77eebe97ca2c77ff103d0319feaeb1715b5260b2dfcbbd2b74f7eb684  fixture.csv
559366f9980ed8f6b9a6245dcec2af60dc59368f0e1080c7251ae69ca96aa0c4  expected.csv

SHA-256 of approved article source (article.md.txt):
003478c41b165627a39c7cc11c5b456e1725833abb8701441876f9ae7d1958ff

Scope and limitations:
These are bounded tests, not proof of all-runtime or spreadsheet compatibility.
Tests cover retention/order, whitespace/case/numeric-string distinctions,
quoted commas and newlines, Unicode and embedded CRLF/CR preservation,
literal formula-leading values, header failures, malformed widths, invalid
UTF-8, unclosed quotes, output protection, and the complete CLI fixture.
No Excel/LibreOffice compatibility test or production-scale claim is made.
No customer records were used. No formulas were evaluated.
Raw CSV can trigger spreadsheet formulas. Inspect fixtures as plain text;
do not open them in a spreadsheet. Literal preservation is not sanitization.
The script uses memory proportional to the input and inherits CSV field limits.
Validation-before-write is not transactional: a write error can leave partial
new output. Flags support review; they do not authorize record deletion.