v0.82 CLOSEARM lane (#1484): the conjunction two gates never read, and a population of 32

[avrabe]

R14 — the conjunction two gates never read

status: implemented is simultaneously R3's only remedy for "the status under-reports
shipped work" and authorised_close_set's permission to retire the issue. v0.81's
JESSDIVERGE3 and WIDEFILE4 each delivered a refutation, said so in their own
verified-by, and were set claiming for R3's sake — which silently placed an externally
reported live blocker in the authorised close set. An operator caught it by running the
closure gate before tagging. No gate did.

R14 now reds an artifact whose verified-by: opens with an INCOMPLETE verdict while its
status claims the outcome holds and no issue-scope: is declared at all.

Why a declaration, not outlives

R11 reds a non-delivery disposition: beside a claiming status. R3 reds a non-claiming
status whose done-when evidence exists. So for a lane that did real work and refuted its own
premise, claiming + prose-only refutation + no disposition is the forced configuration —
there is no legal third option. Demanding outlives would red RQ-77-PROSEBLIND, whose
verified-by opens REFUTED and whose issue is genuinely resolved. Either value satisfies
R14; what it forbids is leaving the issue's fate inferable only from English.

R3 and R11 are unchanged.

Potency — paired mutants, both directions

mutant	tests killed
rule condition forced False	the 2 red-first tests
not _iscope14 dropped (over-fires)	the 2 green-path tests
as shipped	7/7 OK, live gate rc=0

On real data R14 reds RQ-77-PROSEBLIND, and reds both v0.81 artifacts when their
outlives is stripped — the near-miss, reproduced.

The vocabulary is imported, not mirrored

INCOMPLETE and leading_verdict come from verdict_prose_check under an assertIs —
identity, not equality, because an equal-but-separate set passes today and drifts tomorrow.
A second copy of the verdict sets would be the #667 defect one tier down. Asserted: zero
INCOMPLETE = { literals in the gate.

RQ-77-PROSEBLIND gains issue-scope: closes, the truthful value and the first closes
beside a claiming status — so R12's body was read before adding the field, not after a red.

VALUELESSKEY: the population, derived (#1458)

With ast, not a grep — because the artifact's own git grep 'str(fields.get(' undercounts
the class it scopes, missing str(art.get(, str(a.get(, str(fl.get( and the tuple form:

class	sites
reads a YAML artifact field (the live class)	14
reads API JSON / config	13
test fixture, cannot surprise	2
other (2 reach a field via a tuple)	3
total	32

11 of the 32 sit in status_evidence_check.py itself, two in R12's own path where a bare
issue-scope: becomes the invented string "none". R14 is not in the population — it
writes or "" before str(), verified by re-running the census after it landed. The lane's
deliverable is a derivation plus a tripwire, not a 32-site sweep.

Also

Two module docstrings made raw, clearing the last two SyntaxWarnings in scripts/
(invalid escapes that become SyntaxError in a future Python); census re-run to 0 with a
planted control proving the detector still fires. CLOSEARM's refuting command re-pointed at
the right module, and its done-when corrected after the design changed.

All 20 CI-invoked test scripts pass in CI's exact invocation form.

Refs #1484 #1458 #1319

🤖 Generated with Claude Code

https://claude.ai/code/session_01YJK5LZZEkV5smCY1jKn18L

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!