A fifth form of the authorship sentence: nobody named is not a breakdown of zeros

[rdryfoos]

Rik's ruling of 2026-10-09: a fifth form of the authorship sentence for the all-unassigned case, added to the four that stand.

The form

No author named on any of the 10 requirements.

It fires only when every row lacks a mark. One mark among many keeps the partial form, which is what the boundary test holds.

One wording I chose, for you to overrule

The ruling gives No author named on any of the N requirements. At N=1 that renders No author named on any of the 1 requirements, which is not English, and one unassigned row is a reachable and common state: it is what a cold project has after its first mint. So the fifth form has a singular:

No author named on the one requirement.

That sentence was not ruled and is mine. It is one if and one string in authorship_sentence(), plus the test that pins it; say the word and both come out and N=1 reads the ruled sentence verbatim. The house already handles singulars this way two lines below (noun = "requirement" if case == 1 else "requirements").

Why the old behaviour was wrong

The partial form exists so that a breakdown of a handful of rows is not read as the whole registry, which is why it ends by counting the rest. With nothing assigned there is no rest, so it ended in four zeros measuring an empty remainder, in the same voice as a real count. Found on the sixth specimen's own comment, posted by the 0.5.6 workflow:

No author named on 10 of 10 requirements; of the rest, 0 came from the case, 0 from the project (design 0, retrospective 0, constitution 0).

That is #74's comment verbatim, read today. The reader it lands on is the one who has not met the field yet, and it tells them the tool measured something about their project rather than that there is nothing yet to measure.

Tests

Three new, all under AC-GATE-160d, and the four existing tests are untouched:

Test	Holds
test_AC_GATE_160d_the_sentence_counts_no_rest_when_there_is_none	three unassigned rows give the fifth form, and neither of the rest nor came from the case appears
test_AC_GATE_160d_the_fifth_form_reads_as_English_for_one_row	one unassigned row gives the singular, and any of the 1 appears nowhere
test_AC_GATE_160d_one_mark_among_many_keeps_the_partial_form	one of three marked keeps the partial form, and any of the appears nowhere

Negative control. With main's thread-report.py in place and these tests as written, the two fifth-form tests fail and the boundary test passes: 2 failed, 9 passed. Only the script was reverted for that run, nothing else.

The empty-rows case is untouched and still reaches its own sentence first, which test_AC_GATE_160d_a_missing_v5_file_loses_the_numbers_not_the_report holds: no v5 sidecar is not knowing, not an unassigned registry.

Two corrections that came with it

AC-GATE-160d quoted the sentence as N promises from the case, M from the project (...). The emitter has printed N requirements came from the case, M from the project (...) since the plain-words ruling of 2026-10-02, and the tests have asserted that wording since. The registry row kept the pre-ruling shape. It is corrected here in the same sentence that gains the fifth form, so the row describes what the code does.

docs/trace-manifest-v5.md carried the same stale wording in both of its quoted examples, one of them in a third wording (Authorship unassigned on 99 of 104 rows) that the code has never printed. Both are corrected by running authorship_sentence() on those exact counts and pasting what came back, and the fifth form is added beside them.

This is scope I added rather than scope you asked for. The reason is that a row and a document quoting a sentence this PR changes cannot be left quoting a sentence the tool does not say.

Checks run here, before the push

• Full suite: 230 passing on main, 233 on this branch.
• bash -n on all seven extensions/specassay-check/scripts/*.sh: parse ok.
• check-doc-versions.py: OK, version being cut 0.5.6, 0 warnings.
• Gate 2 on this repo's own registry: gate.ok=True, OK (154 registry IDs).

Release note

The ## Unreleased preamble said that nothing in the entry changes the bundle. This entry does: thread-report.py is shipped. The preamble is rewritten to say so, and to record that the two catalog issues that imposed the freeze are closed, since the two items below it landed under a freeze that no longer holds.

Not in this PR

The fifth form will not show on #74 once #79 marks the example app's ten rows. That is as you ruled: the form is for strangers' projects, and the specimen is the thing that found it.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Vzf6Sh5RVMKdFqaXKFF2FY

---

Generated by Claude Code