[Polichinel]

Closes #184. Finishes #158. Part of epic #181.

What #158 actually left behind

It was closed on 2026-08-01 as "cosmetic, non-blocking". It was neither.

Four citations survived in .py. The one that mattered sits inside the message raised by appwrite_env.assert_env_declared — what an operator reads when a delivery refuses to launch, pointing them at a retired contract name in a repo they do not own, on the day their run failed.

And the rename broke a URL. It was applied as a blanket find-and-replace, which rewrote the filename inside the link in ADR-013 §7d:

.../docs/ADRs/platform/the Appwrite Seam Contract_identity_secrets_configuration_contract.md

A path containing spaces. It has never resolved. The upstream file has since been renamed to appwrite_seam_contract.md independently, so the link was wrong twice over.

Both §7d links now pinned

/blob/main/ → /blob/b54928f/. §7d claimed "referenced by URL, never by copy" — true in letter, false in effect while an unpinned link drifts underneath it. It did drift: the registry moved 1.3.0 → 1.4.0 on 2026-08-02, the same day.

Recorded as Erratum E2 in the post-adoption record, matching E1's shape. contract_version stays 1.5; no wire artifact touched.

The ban

test_the_retired_contract_name_is_gone_from_code scans the package, the tests, and scripts/, reports file:line, and excludes docs/ADRs and reports/ — an ADR describing a 2026-07 decision correctly says what the thing was called in 2026-07. Code has no such excuse.

Review findings — all five addressed

F1 (warning) — the ban scanned two of three code roots. scripts/build_gaul_lookup.py (builds the committed GAUL artifact) and scripts/build_wire_fixture.py (generates the §10 fixture) were unguarded. Clean today, so it passed — which is exactly register C-74's defect, registered ninety minutes earlier: a guard whose declared scope exceeds its actual scan. Committing it again inside the epic that exists to close it would have been indefensible. Widened, and re-mutation-tested at the new width:

['scripts/build_gaul_lookup.py:1']

F2 — the module's pinned URL now names the version (v1.4.0), matching §7d. A pinned URL does not rot but it does go stale; the version is what makes that legible.

F3 — test_secret_env_names_follow_the_platform_naming_rule → ..._the_seam_contract_naming_rule. The comment had been corrected; the function name still echoed the retired contract.

F4 — Erratum E2 added to the post-adoption record, so the change appears in the changelog rather than only inline at §7d.

F5 — self-exclusion now compares resolved paths, not basenames.

Caught mid-story

The ban test failed on my own docstring — a parenthetical I had just written into appwrite_env.py naming the retired contract in order to explain its retirement. Removed. That explanation belongs in the ADR, which now carries it.

Forward note for S6 (#187)

The registry is v1.4.0 as of today (amended = "2026-08-02", contract §5.9). #187's body specifies SEAM_CONTRACT_VERSION = "1.3.0" and is stale before it is written. b54928f now appears in two files that must move together — making that pair checkable is precisely S6's job.

Verification

ruff check .                                                    # clean
pytest -q                                                       # 313 passed, 44 xfailed, 5 known pyarrow (C-72)
grep -rn "PLATFORM-001" --include=*.py views_postprocessing/ scripts/ tests/   # 0 (excl. the banning file)
grep -rnoE "https://github\.com/[^)]*" docs/ | grep " "         # 0 — no space-bearing URLs
grep -rn "views-appwrite/blob/main" docs/                       # 0 — all pinned
git diff --stat origin/main -- tests/fixtures/                  # empty

Both pinned targets verified present in the views-appwrite checkout, and b54928f confirmed a real commit there.

🤖 Generated with Claude Code