The script that catches stale documentation had a stale claim of its own

check_doc_claims.sh's header opened "the project has 471 regression tests" -- the
v1.6.1 figure, wrong since the count reached 574. The claim was doing no work, so it
is gone rather than updated to a number that will go stale again: the rationale holds
whatever the count is.

The one remaining 471 in tools/ is deliberate. check_doc_counts.sh quotes "471 tests
across 26 suites (as of v1.6.1)" as the EXAMPLE of the convention it enforces -- a
count carrying its version stamp, versus a bare number. It is self-consistent and
illustrates the rule rather than asserting a current fact.

Swept the rest: run_regression.sh carries no style or count references,
check_error_handling.sh's cross-reference to the other four scripts is accurate, and
run_test.sh's usage is unchanged. Nothing else in tools/ went stale today.

Gates: all five 0.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: ironsheep

tools/check_doc_claims.sh

@@ -6,8 +6,8 @@
 # *strings* — the sample transcripts, banners, and status lines that documents
 # quote as tool output.
 #
-# Why this exists: the project has 471 regression tests and every one asserts on
-# behavior. None compares a description to the behavior it describes. So a stale
+# Why this exists: the project's regression suites assert on behavior, and every one
+# of them does. None compares a description to the behavior it describes. So a stale
 # sample transcript has exactly one detection mechanism — someone reading it — and
 # if that reading only happens at the release gate, findings arrive under time
 # pressure. This script makes the comparison mechanical.