/heal-tests

Auto-fixes broken selectors and timing issues in a failing Playwright spec file. Real app bugs are skipped — those get escalated as bug reports, not silently fixed.

Usage

/heal-tests [<spec-file>]

All command combinations

Every --tool and --id option in one place → Command Combinations

Examples

# Heal a specific spec file
/heal-tests tests/generated/login.spec.ts

# Heal all generated specs
/heal-tests

What it does

For each failing test, the agent:

1. Classifies the failure:

   • Selector issue — element moved or renamed in the DOM
   • Timing issue — element not ready when the action runs
   • Logic mismatch — test expectation doesn't match current app behavior
   • Real app bug — the app is broken, not the test

2. Auto-fixes selector and timing issues, then re-runs to verify

3. Flags logic mismatches for human review

4. Logs real bugs to your issue tracker

When to use it

• After a UI change broke selectors
• After upgrading a component library that changed class names or attributes
• After the pipeline flags flaky tests that you suspect are timing-related

What it won't do

/heal-tests will not silently "fix" a test by weakening its assertions to make it pass. If the app broke, the test failure is correct — /heal-tests will surface that as a bug, not paper over it.