Initial commit

commands/fix-flaky.md

@@ -0,0 +1,88 @@
+# Fix Flaky Test Command
+
+## Description
+
+Analyze and fix flaky (intermittently failing) Playwright tests. Identifies race conditions, improves waiting strategies, and ensures test stability.
+
+## Usage
+
+```
+/fix-flaky [test-name]
+```
+
+## Parameters
+
+- `test-name` - Name or path of the flaky test (required)
+
+## Examples
+
+```
+/fix-flaky dashboard.spec.ts
+```
+
+```
+/fix-flaky "should load user profile"
+```
+
+```
+/fix-flaky tests/checkout.spec.ts:42
+```
+
+## Instructions for Claude
+
+When this command is invoked:
+
+1. **Invoke the test-debugger and test-maintainer skills** for analysis
+
+2. **Gather information**:
+   - Test file and name
+   - Failure pattern (how often it fails)
+   - Error messages when it fails
+   - Environment (local vs CI)
+   - Which step usually fails
+
+3. **Identify flakiness causes**:
+   - Race conditions
+   - Missing waits
+   - Hardcoded timeouts
+   - Network dependencies
+   - Test isolation issues
+   - Environment differences
+
+4. **Common fixes**:
+   - Add explicit waits before interactions
+   - Replace waitForTimeout() with proper waits
+   - Wait for network to settle
+   - Wait for specific API responses
+   - Increase timeouts for slow operations
+   - Improve test isolation
+   - Add retry configuration
+
+5. **Apply improvements**:
+   - Update test code
+   - Add proper waits
+   - Fix race conditions
+   - Ensure test isolation
+   - Configure retries if needed
+
+6. **Verify stability**:
+   - Run test multiple times (10+ times)
+   - Test in different environments
+   - Check in CI environment
+   - Monitor for continued flakiness
+
+## Error Handling
+
+- If test not found, ask for correct path
+- If can't reproduce failure, ask for more details
+- If multiple issues, fix most impactful first
+- Warn if test design is fundamentally flaky
+
+## Notes
+
+- Flakiness is usually caused by timing issues
+- Never use waitForTimeout() - use explicit waits
+- Wait for network/animations to complete
+- Ensure tests are isolated and don't share state
+- Run tests 10+ times to verify fix
+- Consider enabling retries for legitimately slow operations