fix(pm): stop check-widening-tells T2 keying on a continuation operator, and price a false positive honestly (#16944)

T2 read two accidental variables as widening evidence, and #16822 measured
both against real diffs:

  * a bare-string line was a closed-set member whenever the NEXT line did not
    begin `+ '…'`. An eight-fragment prose argument produced exactly one tell —
    its first fragment — and moving that fragment onto the calling line made
    the identical string stop being a tell. The continuation operator is now
    read on both sides, so the two spellings agree.
  * `CLOSED_SET_OPENER` fired on the constructor keyword whether the rewrite
    widened the set or narrowed it. An opener whose list opens on a later line
    declares no member, so when the same hunk removes an opener with the
    identical binding prefix, the set already existed and its members — still
    read one line each — decide.

Both decline only on evidence the hunk carries; with no evidence the tell
still fires. Neither recovers block state and neither claims direction.

The docblock's "the remedy is one word in the claim comment — never a weakened
rule here" is replaced by the real price: the exit-0 condition is "no tell, or
the declaration is not `no`", so the only word that clears a false tell writes
a widening that did not happen into a governance ledger. Its true half — never
trade a loud failure for a quiet one — is restated as the standing rule. The
sibling's twin of the same sentence is corrected with it; C5's exit condition
does not move.


Claude-Session: https://claude.ai/code/session_012GKcPZbMoGq7WPzKLfRBTU

Co-authored-by: claude[bot] <claude[bot]@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

Author: claude[bot]

scripts/pm/check-clause2-carriers.mjs

@@ -1432,9 +1432,15 @@ export function pairWidening(pair, repo) {
  * enqueue gate before the card may enqueue -- the diff is a fact; the card's
  * semantics were a prediction", and until #16448 that gate was a human reading.
  *
- * ⚠️ A TELL, never a proof, in BOTH directions: a false positive costs the
- * author one word in the claim comment, and a false negative is the cost the
- * ruling accepted when it took the directional reading. So this row never
+ * ⚠️ A TELL, never a proof, in BOTH directions: a false negative is the cost
+ * the ruling accepted when it took the directional reading, and a false
+ * POSITIVE is repaired in the matcher — ⛔ not paid for by the author. This
+ * file priced it at "one word in the claim comment" until #16822 measured the
+ * price: the exit-0 condition is "no tell, OR the declaration is not `no`", so
+ * the only word that clears a false tell is `Clause-②: no` → `yes` — a
+ * widening recorded in a governance ledger that did not happen, and afterwards
+ * indistinguishable from one that did. ⛔ Nothing about this row is relaxed by
+ * saying so: the exit stays non-zero and stays a hard block. So this row never
  * asserts that the diff widens -- it asserts that the diff has the SHAPE of one
  * that does, while the claim says it does not, and names the file:line so the
  * author can answer with the file open.