[Polichinel]

Closes #190. Closes epic #181.

The acceptance criteria, run as a script

ruff check .                                      clean
pytest -q                                         307 passed, 3 skipped, 41 xfailed, 5 known pyarrow (C-72)
git diff --stat origin/main -- tests/fixtures/    empty
CONTRACT_VERSION                                  "1.5"          (unmoved all epic)
grep -rn PLATFORM-001 --include=*.py              0
live `"unknown"` constants in gaul_lookup         0              (AST-checked, not grepped)
grep -rn "/home/\|/Users/" tests/ scripts/ pkg/   0
views_pipeline_core importers                     1
register                                          74 / 15 open / 59 resolved

What closed

	by	proven by
C-71	S1	a check over both entry validators, so the pair cannot drift again
C-43, C-59, C-61	S2	test_gaul_lookup_fidelity.py, 26 tests
C-03	S4	43 self-referential tests → 14 against the real gate
C-60	S5	artifact rebuilt, values byte-identical, stamp unchanged
C-57	S6	four checks, each mutation-proven
C-46	S7	grep -rn "/home/" → 0 · Cluster K closed
C-74	S10	scan coverage 6 → 17 files
C-22	S8	written for the delivery that exists

Plus: #158's rename finished and the broken URL it created in ADR-013 §7d repaired (Erratum E2); #154's undone half completed across four living docs; #15 superseded; a cross-repo pin re-taken after #196.

Open 24 → 15, all classified: six on the 3.0.0 publish, five owned upstream, four deliberately deferred.

ADR-014, and the case against writing it

The argument for writing nothing was strong for most of this. A contributor who violates the register's closing-condition convention already gets a failure naming the exact strings and the fix, at the moment of violation — better teaching than an ADR nobody opens.

What that leaves unrecorded is the two clauses no test carries:

• §4 a deferral names a trigger and an owner, or it is not a deferral
• §5 a change that names a record disposes of that record in the same change

An ADR for the mechanisable rules would be ceremony. One for the un-mechanisable ones is the only vehicle available.

§5 says outright that no test enforces it and none is proposed, and records the attempt so nobody repeats it. The evidence that it lapses is this epic's own record: C-71 was fixed in the morning and sat filed open all day while eight further stories shipped, and the CIC lagged three consecutive stories — caught by review every time, by anything mechanical never.

The other three clauses are the guard discipline the epic paid for: mutation-prove or it is decoration (§2); prefer a false negative to a false alarm, and when a guard cries wolf check the matching before the scope (§3); and existence is not reachability — a pin that existed, whose files existed at it, which passed every check written at the time, and had never reached main.

What did NOT get done — stated at length, on purpose

A closeout reporting only successes is the defect this epic exists to fix.

1. Cluster M untouched, correctly. Six entries, two Tier 2, one upstream publish. No engineering here moves them.
2. The CI question is decided in writing but unowned. Three gated cross-repo checks run nowhere automatic. Recommendation and trigger are in C-46's residual. A decision awaiting an owner, not a task awaiting effort.
3. Withdrawal of a bad delivery is chosen and deliberately unbuilt — needs an ADR-013 amendment plus views-faoapi work, and FAO's answer on audit requirements decides whether it is wanted at all.
4. Two questions are with the UN FAO (Pre-Release Note 07, B.1 and B.2).
5. The deploy gate's xfail tuning was left alone — S7 fixed how the checkout is found, not what the gate asserts.
6. Local Python is 3.10 against a declared >=3.11 floor, so three checks skip locally and run in CI.

Verification

The block at the top is reproducible; every line is a command.

🤖 Generated with Claude Code