Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.
MANDATORY: Use the Playwright Healer Agent for Reproduction
Always use the Playwright healer agent to run and reproduce failing tests. The healer provides richer diagnostics than plain yarn playwright test — it can debug step-by-step, inspect the live UI, and collect detailed failure context automatically.
Note: The Playwright healer agent is currently supported in OpenCode and Claude Code only. In Cursor or other tools without Playwright agent support, skip the healer initialization and use the "Fallback: Direct Execution" method below instead.
Healer Initialization
If not already initialized in this session, initialize the healer agent in e2e-tests/:
cd e2e-tests
# For OpenCode
npx playwright init-agents --loop=opencode
# For Claude Code
npx playwright init-agents --loop=claude
Generate the .env file by passing the --env flag to local-test-setup.sh:
cd e2e-tests
source local-test-setup.sh <showcase|rbac> --env
To regenerate (e.g. after token expiry), re-run the command above.
Project Selection
When running specific test files or test cases, use --project=any-test to avoid running the smoke test dependency. The any-test project matches any spec file without extra overhead:
yarn playwright test <spec-file> --project=any-test --retries=0 --workers=1
Running via Healer Agent
Invoke the healer agent via the Task tool:
Task: "You are the Playwright Test Healer agent. Run the following test to reproduce a CI failure.
Working directory: <path>/e2e-tests
Test: <spec-file> --project=any-test -g '<test-name>'
Run: set -a && source .env && set +a && npx playwright test <spec-file> --project=any-test --retries=0 --workers=1 -g '<test-name>'
If the test fails, examine the error output, screenshots in test-results/, and error-context.md.
Report: pass/fail, exact error message, what the UI shows at the point of failure."
Fallback: Direct Execution
If the healer agent is unavailable (e.g., in Cursor), run tests directly:
cd e2e-tests
yarn playwright test <spec-file> --project=any-test --retries=0 --workers=1
Examples:
# A specific spec file
yarn playwright test playwright/e2e/plugins/topology/topology.spec.ts --project=any-test --retries=0 --workers=1
# A specific test by name
yarn playwright test -g "should display topology" --project=any-test --retries=0 --workers=1
Headed / Debug Mode
For visual debugging when manual investigation is needed:
If the first run passes (doesn't reproduce the failure), run multiple times to check for flakiness:
cd e2e-tests
# Run 10 times and track results
PASS=0; FAIL=0
for i in $(seq 1 10); doecho"=== Run $i ==="if yarn playwright test <spec-file> --project=any-test --retries=0 --workers=1 2>&1; then
PASS=$((PASS + 1))
else
FAIL=$((FAIL + 1))
fidoneecho"Results: $PASS passed, $FAIL failed out of 10 runs"
Result Classification
Consistent Failure
Definition: Fails every time (10/10 runs fail)
Action: Proceed to e2e-diagnose-and-fix skill
Confidence: High — the fix can be verified reliably
Flaky
Definition: Fails some runs but not all (e.g., 3/10 fail)
Action: Proceed to e2e-diagnose-and-fix skill, focus on reliability improvements
Typical causes: Race conditions, timing dependencies, state leaks between tests, external service variability
Cannot Reproduce
Definition: Passes all runs locally (0/10 fail)
Before giving up, try running the entire Playwright project that failed in CI with CI=true to simulate CI conditions (this sets the worker count to 3, matching CI):
cd e2e-tests
CI=true yarn playwright test --project=<ci-project> --retries=0
Replace <ci-project> with the project from the CI failure (e.g., showcase, showcase-rbac). This runs all tests in that project concurrently, which can expose race conditions and resource contention that single-test runs miss.
If the full project run also passes, stop and ask the user for approval before skipping this step. Present the reproduction results and the list of possible environment differences. Do not proceed to diagnose-and-fix without explicit user confirmation.
Investigation: Check environment differences between local and CI:
Cluster version: CI may use a different OCP version (check the cluster pool version)
Image version: CI may use a different RHDH image
Resource constraints: CI clusters may have less resources
Parallel execution: CI runs with 3 workers; the full project run above simulates this
Network: CI clusters are in us-east-2 AWS region
External services: GitHub API rate limits, Keycloak availability
Artifact Collection
Playwright Traces
After a test failure, traces are saved in e2e-tests/test-results/:
# View a trace
yarn playwright show-trace test-results/<test-path>/trace.zip
HTML Report
Never use yarn playwright show-report — it starts a blocking HTTP server that will hang the session. Instead, the HTML report is generated automatically in playwright-report/ after test runs. To view it, open playwright-report/index.html directly in a browser, or use Playwright MCP to navigate to it.
Screenshots and Videos
On failure, screenshots and videos are saved in test-results/<test-path>/:
test-failed-1.png — Screenshot at failure point
video.webm — Full test recording (if video is enabled)
Test Project Reference
Refer to the e2e-fix-workflow rule for the Playwright project → config map mapping.
