add-regression-test

A regression test proves a bug exists by failing on pre-fix code and passing on post-fix code. This Skill derives the exact bug input and observable outcome from your fix diff, then uses safe reproduction modes like stashing or temporary worktrees to verify the red and green states.

Writing a regression test for a git commit requires extracting the failing input from the commit diff, asserting on externally observable behavior, and verifying the test runs red on the pre-fix state and green on the post-fix state before wiring it into your project conventions.

Yes, you can use a temporary git worktree to reproduce the pre-fix state. This Skill selects safe reproduction modes like stashing, reverting via worktree, or checking out pre-fix hunks to force a failing-state run and confirm the test is red before the fix.

Regression testing works with working trees, staged changes, local commits, and squashed commits. The Skill targets code fixes across these git states by selecting an appropriate reproduction mode to verify the test fails before and passes after the fix.

Your regression test passes when the bug is present because it is not asserting on the correct externally observable outcome. A true regression test must assert on observable behavior like HTTP, DB, UI, or command output, and be verified red on the pre-fix code.

To fit regression tests into project conventions, name the test after the specific bug symptom for long-term maintainability and place it in the nearest project test-file convention. Run the full test suite after wiring it in to ensure everything is green.