The verdict says how much of the documentation it actually covered

[melbinjp]

The README promises this, and until now it was only half true:

a clean report over two files in a project with three hundred cannot be mistaken for a clean
report over three hundred

The coverage note printed in the header. The verdict printed forty lines below it and said
Nothing contradicted. 149 claims checked on its own. The line anyone quotes from a CI log is
the last one.

The measurement that prompted it

I ran docproof over nine public repositories and compared what it read against what was there:

repo	read	not read	coverage
zeroclaw	209	55	79%
paperclip	145	231	39%
sentry-react-native	8	18	31%
tutti	92	234	28%
hive	54	162	25%
cherry-studio	120	418	22%
gsd-core	307	1218	20%
opensre	26	102	20%
langwatch	11	372	3%

972 of 3,782, or 25.7%.

Then I re-ran two of them with --docs '**/*.md'. langwatch went from 1 broken to 19.
cherry-studio went from 23 to 111. Those runs were never clean. They were narrow, and only
the header said so.

What changed

Report.render() takes read and unread and appends one sentence to the verdict:

1 broken, 149 checked, 497 not judged.
This judged 11 of 383 documentation file(s). 372 were never read, so this verdict
covers 2% of the documentation in this project.

Attached to the broken verdict as well as the clean one, because "19 broken" over a fifth
of a tree misleads exactly as much as "nothing contradicted" over a fifth. A project whose
documentation was entirely read gains no sentence at all, because a reassurance printed on
every clean run is the kind of noise that teaches a reader to skip the whole block.

Three tests, 193 to 196, and the README now shows the real output.

One of those tests failed the first time and deserved to. I built a fixture with no deleted
path and asserted the run was broken. make_repo's own docstring is emphatic that a later
removal is what separates real drift from an illustration, and I had ignored it. Rewritten with
documented_before plus deleted.

This does not change the default scope, which is deliberate and argued for in find_docs. It
changes what a narrow run is allowed to sound like.