Fix verification skill: from repeated-save failure to passing tests

The starter function appended an article ID every time it was saved. A second save of the same article produced a duplicate. The expected behavior was to keep first-seen order, ignore blank IDs, and leave the caller’s list unchanged.

python3 -B -m unittest -v

On Python 3.14.5, the unchanged starter failed two checks: saving an existing ID again and repeating the save several times. Four adjacent checks already passed. The command exited with status 1.

def save_article(saved, article_id):
    """Return a new reading list, preserving first-seen order."""
    result = list(saved)
    if article_id and article_id not in result:
        result.append(article_id)
    return result

The correction copies the input and appends only a nonblank ID that is not already present. The repeated-save tests assert observable output, so they fail the old implementation and pass the corrected one.

test_blank_id_is_ignored ... ok
test_empty_list ... ok
test_input_is_not_mutated ... ok
test_new_article_preserves_order ... ok
test_repeated_calls ... ok
test_repeated_save_is_idempotent ... ok
Ran 6 tests
OK (exit 0)

The runner copied the starter into a temporary directory, ran the test file, replaced only the function with the corrected solution, and reran the same command. The test suite was unchanged between the two runs. The project requires no third-party Python packages or network service.

Download the developer practice project and reproduce the failure before reading the solution. In your handoff, record the command, environment, failing assertion, changed behavior, passing result, and remaining limits. Do not substitute “looks correct” for the original regression check.

The assistant manually applied the published fix-verification procedure to a small deterministic Python fixture on September 10, 2026. Six cases were tested, not arbitrary input types, concurrent writes, persistence, or a production repository. There was no independent evaluator, repeated model-run comparison, or automatic skill-selection benchmark.

Read the skill and its package notes · Download the practice project