add-regression-test

A regression test verifies a bug fix by forcing a failing pre-fix run and a passing post-fix run. It reproduces the failing behavior using git stashing or reverts, then restores the fix and confirms the test passes against the corrected behavior.

Two-state verification reproduces the failing behavior before the fix, then restores the fix and confirms the test passes. It uses git stashing, reverts, or temporary worktrees to prove the test would have failed on the pre-fix code path.

Regression tests can be generated for staged, committed, or working-tree bug fixes. The process uses the fix's git state to verify the test fails before the change and passes after it across common test harnesses and project conventions.

No specific test runner is required. The process selects a compatible test runner based on your project conventions, such as colocated test files or test folders, to wire the test with symptom-based naming.

Regression test assertions should target externally observable outcomes like responses, DB state, emitted events, or rendered DOM. This prevents tests from breaking on internal implementation details and ensures they capture the actual bug symptom.

If a regression test passes before the fix, it likely asserts internal implementation details rather than the bug symptom. The test must reproduce the failure deterministically by targeting externally observable outcomes to force a pre-fix failure.