Four skill documents organised by subject, replaced by six organised by task, which is the whole finding.
by subject (what failed)
architecture.md 2,100 words. read partially,
or not at all.
conventions.md 1,400 words.
testing.md 1,800 words.
by task (what works)
add-an-endpoint.md 180 words
add-a-migration.md 140
write-a-test.md 220
add-a-value-object.md 160
review-a-change.md 200
debug-a-failure.md 190
same information, a tenth of the length at the point of
use.
A document loaded because a task started is read; a document loaded because a subject is relevant is skimmed. The task documents cross-reference the subject ones for detail, which means the long documents still exist and are read by people rather than by tools.