Phase 4: make Tests-fix empirical (stash-and-run), not judged #47

Template gap

The Phase 4 rubric item "Tests-fix: a test exists that would fail without the fix" is scored by judgment today. That reasoning can be wrong, producing a false-negative regression test that passes even with the bug present (commonly when the "after" state is indistinguishable from the "before" — e.g. a "don't overwrite" test that writes back identical content).

Rule to add to the template (Phase 4 Tests-fix item / Phase 8 regression gate)

Tests-fix (empirical, not judged): temporarily revert the fix (git stash push -- <src files>), run the new/changed tests, and confirm they FAIL; then restore (git stash pop) and confirm they PASS. A regression test that passes with the fix stashed is a false-negative — use distinct/sentinel data so the failure mode is observable.

Language- and repo-agnostic; the false-negative-regression-test trap is universal.

Consolidates (close on merge)

• example-|-dev-loop#5 — strengthen Tests-fix rubric with stash-and-run verification

Consolidated upstream from per-loop self-audit findings. drafted by Claude on behalf of Daniel Stephenson