test(spec): the transactions tombstone pin discriminates its own subject (#19078)

Fixes #18996

Clause-②: no

## What was wrong

`packages/spec/src/data/driver.test.ts` pinned the `transactions`
retired-key prescription with a regex that required only that `METHOD
PRESENCE` and `beginTransaction` appear *somewhere* in the thrown
message:

```
DriverCapabilities\.transactions.*removed.*METHOD PRESENCE.*beginTransaction.*Delete the key
```

PR #18890 rewrote that prescription's opening from "Transaction use is
gated on METHOD PRESENCE — `driver.beginTransaction` …" to "Transaction
use is gated on the DRIVER'S DECLARATION, no longer on METHOD PRESENCE
alone: `engine.transaction()` asks `driverSupportsTransactions(driver)`
…". The old regex matches **both**, so reverting the prescription to the
now-false sentence kept this test green: the pin could not fail on the
one shape it exists to catch. Its title — "points at method presence" —
was in the old world too.

The shipped prose is true today; what was missing is the guard that
keeps it that way. This is coverage debt, not a contract defect.

## What changed

One regex and one title, in one test file. The new anchor is the pair of
things #18890 made load-bearing — the **negation** (`no longer on METHOD
PRESENCE`) and the **predicate that replaced presence**
(`driverSupportsTransactions`):

```
DriverCapabilities\.transactions.*removed.*no longer on METHOD PRESENCE.*driverSupportsTransactions.*beginTransaction.*Delete the key
```

`beginTransaction` and `Delete the key` stay, so nothing the old pin
covered is dropped.

The card suggested anchoring on `DECLARATION`; both discriminate on the
tree measured below, and the negation was chosen because it is the
literal contradiction of the false claim — a future sentence cannot say
"gated on METHOD PRESENCE" and "no longer on METHOD PRESENCE" at once,
whereas a bare mention of a declaration can coexist with a
presence-gated claim.

**⛔ Deliberately not a ban on the phrase.** `driver.zod.ts` carries a
*second* "gated on METHOD PRESENCE" sentence, for `schemaSync`, and that
one is true of the runtime today. It is untouched, and so is
`driver.zod.ts` as a whole — this PR's diff is one `.test.ts` file.

## Evidence — the acceptance leg is an ablation, not "the new pin
passes"

**1. Ablation (decisive).** `node scripts/ablation-replace.mjs` in WRAP
mode over `packages/spec/src/data/driver.zod.ts`: the prescription's
opening clause was replaced with the **verbatim** pre-#18890 false
sentence, read out of history at
`5ba2ec3ca4^:packages/spec/src/data/driver.zod.ts:266-267`, with the
rest of the prescription left intact.

- mutation landed, proven on disk, not by an exit code: anchor `1 -> 0`,
replacement `0 -> 1`, blob `46daf837a26a -> 9268d7835908`
- the new pin on that false tree: **RED** — `Tests 1 failed | 57 passed
(58)`, the failing case being the one under repair
- restore proven byte-for-byte: blob back to `46daf837a26a`, equal to
the path's blob at HEAD, `git diff HEAD` empty
- the new pin on the restored tree: **GREEN** — `Tests  58 passed (58)`

**2. Lit control — the defect itself, measured first-hand, on that same
false tree.** A probe that reads the thrown message and applies both
regexes reported:

```
PROBE_REGEX_IN_SYNC_WITH_TEST_FILE: true
PRESCRIPTION_OPENING: Transaction use is gated on METHOD PRESENCE — `driver.beginTransaction` (`engine.transaction()`, ADR-0034 ambient transactions). …
OLD_REGEX_MATCHES: true
NEW_REGEX_MATCHES: false
```

The OLD pin stays green on a tree whose prescription is false. That is
the blindness this card names, reproduced here rather than quoted.

**3. Dark control — the trap.** `driver.test.ts`: **58 passed before, 58
passed after**. `driver.zod.ts` is byte-identical to the base (`git
hash-object` = `46daf837a26a…` = its blob at `0ec81857aa`), so
`schemaSync`'s true sentence cannot have moved. One honest
qualification: the `transactionsUnsupported.*NOT this key restored`
assertion lives in the *same* `it()` block, so on the false tree it was
**NOT MEASURED** — vitest aborts a case at its first failed assertion.
It is green on the restored tree, inside the 58.

## Verification

| leg | result |
|:---|:---|
| `pnpm --filter @objectstack/spec test` | `Test Files 491 passed
(491)`, `Tests 14299 passed (14299)` |
| `pnpm --filter @objectstack/spec typecheck` | exit 0 (`tsc --noEmit` +
`check:scripts-typecheck` + `check:test-typecheck`) |
| derived gate families | `node scripts/pm/dispatch-gates.mjs --commands
--repo objectstack-ai/objectstack` → 76 commands; **72 exit 0**, 4 exit
3 |
| gate reconciliation | `--ran` with exit codes recorded: 76 derived, 72
run, **4 NOT-MEASURED, 0 UNRUN** |
| repo-wide lint | `eslint . --no-inline-config` exit 0 over **6866
files**, 0 errors / 0 warnings, at `bada58f3c2` |

The four exit-3 gates each print `PREREQUISITE NOT MET … nothing was
measured` — they read built output of packages this diff does not touch
and need a full monorepo build: `check:doc-formula-expressions`,
`check:dual-build-cjs-loads`, `check:lean-entry-closure`,
`check:type-check-debt`. Not findings, not passes. CI builds fresh and
runs them.

No narrowing is claimed for lint: the population above is the whole
repo, run at the final commit of this branch.

## Changeset — `skip-changeset` measured, not assumed

`npm pack --dry-run --json` in `packages/spec` at this commit lists
**2039** files. Positive control `src/data/driver.zod.ts` is
**present**; `src/data/driver.test.ts` is **absent**, and **zero**
`.test.ts` files ship at all (`files[]` publishes `src/**/*.zod.ts`,
never `*.test.ts`). Nothing any released package actually ships moves,
so this diff publishes nothing. `Clause-②: no`.

## Acceptance notes

- The generic `it.each(RETIRED_BITS)` pin in the same block asserts only
`DriverCapabilities.BIT` … `removed` … `Delete the key` (the bit name
interpolated per case). That is weak **by design** — it pins that each
tombstone exists and prescribes, not what any one sentence claims — so
it is not a second instance of this defect. Noted, not filed.
- The other two hand-written prescriptions in that block (`streaming`,
`queryFilters`) anchor on content distinctive to their current
sentences, and neither sentence has been rewritten. Noted, not filed.

---
_Generated by [Claude
Code](https://claude.ai/code/session_01JbZnqu8bt6YqfJsr9vaFb3)_

Co-authored-by: Claude <noreply@anthropic.com>

Author: os-bill

packages/spec/src/data/driver.test.ts

@@ -121,9 +121,16 @@ describe('[#4634] the 31 inert capability bits are tombstoned, not stripped', ()
     );
   });
 
-  it('the transactions prescription points at method presence, and says the live bit is not it restored', () => {
+  it('the transactions prescription points at the DECLARATION gate, not method presence alone, and says the live bit is not it restored', () => {
+    // [#18996] Anchored on what #18890 made load-bearing: the NEGATION and the
+    // predicate that replaced method presence. An anchor on `METHOD PRESENCE`
+    // alone matched this prescription's PREVIOUS, now-false opening ('gated on
+    // METHOD PRESENCE — `driver.beginTransaction`') exactly as happily as the
+    // true one, so it could not fail on the one shape it exists to catch. ⛔ Not
+    // a ban on the phrase — `schemaSync`'s own 'gated on METHOD PRESENCE'
+    // sentence is true of the runtime and stays.
     expect(() => DriverCapabilitiesSchema.parse({ transactions: true })).toThrow(
-      /DriverCapabilities\.transactions.*removed.*METHOD PRESENCE.*beginTransaction.*Delete the key/s,
+      /DriverCapabilities\.transactions.*removed.*no longer on METHOD PRESENCE.*driverSupportsTransactions.*beginTransaction.*Delete the key/s,
     );
     // [#18063] The trap this sentence exists to close: a reader who sees a live
     // `transactionsUnsupported` and concludes the retired key came back. It did