fix(core): capture judge/evaluator token usage in reports#63

Merged

cemde merged 8 commits into

mainfrom

fix-eval-cost-tracking

May 23, 2026

Commits on May 23, 2026

test: add failing repro for judge usage dropped from reports (#60)

cemde
and
claude
committed
test: tighten judge-usage repro per code review (#60)

cemde
and
claude
committed
fix(core): collect usage after evaluate() so judge tokens are captured (#60)

cemde
and
claude
committed
fix(core): tighten step-comment cleanup per code review (#60)

cemde
and
claude
committed
test: assert agent usage survives evaluator and execution failures (#60)

cemde
and
claude
committed
test: tighten failure-path regression tests per code review (#60)

cemde
and
claude
committed
docs: changelog entry for judge usage fix (#60)

cemde
and
claude
committed
docs: fill in PR number for changelog entry

cemde
and
claude
committed