Read this board's subjects rather than counting them beside the switch

[iderex]

The issue this belongs to

Closes #400

What changed

The header of .github/workflows/shared-hygiene.yml no longer counts this
board's commit subjects. The command that answers the count stands where the
numbers were, the correction says what they said and when each stopped being
true, and the paragraph that follows says what deciding the input would take.

The input itself is unchanged, and so is every trigger, permission, job and pin
in the file.

What failure it prevents

The switch subject_names_issue: false is an input to a workflow pinned by hash
in another repository, and the only thing on this board saying why it is set that
way was that count. A reader deciding whether the rule can be turned on read a
measurement taken on 2026-08-23 and never re-taken, and took this board for one
that never names an issue in a subject.

Both halves were wrong. Thirty-one of the last fifty-two authored subjects are in
exactly the shape the sentence said none of them were:

git log origin/main --no-merges --format='%s' -52 | grep -c '\[#[0-9]'
31
git log origin/main --no-merges --format='%h %ad %s' --date=short --reverse | grep '\[#[0-9]' | head -1
3774826 2026-09-05 State the decision and not why the record exists, for records from now on [#304]

And the half about the twelve most recent was refuted at the commit that wrote
it, by that commit:

git log 41e930d --no-merges --format='%h %ad %s' --date=short -12 | grep '#[0-9]'
41e930d 2026-08-23 Call the shared hygiene check beside this board's own gate (#178)
git log origin/main --no-merges --format='%s' -12 | grep -c '\[#[0-9]'
6

It was found by reading the header of every workflow file on this board against
the tree, after the same defect was repaired in two documents earlier today.

Evidence

The commit being pushed:

git rev-parse HEAD
4e06b1b308f57013221db49d3c8bdde7d409f360
git diff --stat origin/main...HEAD
 .github/workflows/shared-hygiene.yml | 29 +++++++++++++++++++++++------
 1 file changed, 23 insertions(+), 6 deletions(-)

Neither claim stands in the present tense any more, and the command that answers
them is in the file:

git grep -c 'not one is in the square brackets' HEAD -- .github/workflows/shared-hygiene.yml ; echo "exit=$?"
exit=1
git grep -c 'carry no reference at all' HEAD -- .github/workflows/shared-hygiene.yml ; echo "exit=$?"
exit=1
git grep -n 'no-merges' HEAD -- .github/workflows/shared-hygiene.yml
HEAD:.github/workflows/shared-hygiene.yml:19:#     git log origin/main --no-merges --format='%s' -52 | grep -c '\[#[0-9]'

The input is where it was:

git grep -n 'subject_names_issue' HEAD -- .github/workflows/shared-hygiene.yml
HEAD:.github/workflows/shared-hygiene.yml:69:      subject_names_issue: false

The file still parses:

python -c "import yaml; yaml.safe_load(open('.github/workflows/shared-hygiene.yml', encoding='utf-8')); print('yaml ok')"
yaml ok

What this does not cover

• Which subject the shared rule reads. The called workflow is pinned by hash in
  another repository and I did not read it, so whether subject_names_issue
  judges a pull request title or a commit subject is not settled here, and the
  edited comment says so rather than assuming it. The two populations answer
  differently: nine of the last fifty-two merged pull request titles carry a
  bracketed reference against thirty-one of the authored subjects.
• Whether the input should be true. Twenty-one of the last fifty-two authored
  subjects carry no bracketed reference, turning the rule on would judge every
  pull request here, and nothing in this change asks for that.
• Any mechanism. Nothing on this board reads a count written into a workflow
  comment, so the next one is refused by nothing and would be found by a reader
  again. No check is added, and none is asked for.
• The rest of the header. The pin, the concurrency group and the argument for
  calling the shared check beside the local gate are untouched.

Who has read it

Nobody other than me has read this change. The evidence above stands in place of
a second reading rather than beside one.