feat: add autonomous E2E CI failure fix workflow with skills and Playwright agents

Add a 7-phase skill-based workflow for autonomously investigating and fixing
failing E2E CI tests. Includes /fix-e2e command, Playwright Test Agent
definitions (healer/generator/planner), and supporting rules.

Skills: parse-ci-failure, setup-fix-branch, deploy-rhdh, reproduce-failure,
diagnose-and-fix, verify-fix, submit-and-review.

Managed via rulesync (skills feature added) and synced to OpenCode, Claude
Code, and Cursor.

Author: zdrapela

.claude/commands/fix-e2e.md

@@ -0,0 +1,147 @@
+---
+description: >-
+  Autonomously investigate and fix a failing RHDH E2E CI test. Accepts a Prow
+  job URL or Jira ticket ID. Deploys RHDH, reproduces the failure, fixes the
+  test using Playwright agents, and submits a PR with Qodo review.
+---
+# Fix E2E CI Failure
+
+Autonomous workflow to investigate, reproduce, fix, and submit a PR for a failing RHDH E2E test.
+
+## Input
+
+`$ARGUMENTS` — A Prow job URL, Jira ticket ID, or Jira URL:
+- **Prow URL**: `https://prow.ci.openshift.org/view/gs/...`
+- **Jira ticket ID**: `RHIDP-XXXX`
+- **Jira URL**: `https://redhat.atlassian.net/browse/RHIDP-XXXX`
+
+## Workflow
+
+Execute the following phases in order. Load each skill as needed for detailed instructions. If a phase fails, report the error and stop — do not proceed blindly.
+
+### Phase 1: Parse CI Failure
+
+**Skill**: `parse-ci-failure`
+
+Parse the input to extract:
+- Failing test name and spec file path
+- Playwright project name
+- Release branch (main, release-1.9, etc.)
+- Platform (OCP, AKS, EKS, GKE)
+- Deployment method (Helm, Operator)
+- Error type and message
+- local-run.sh job name parameter
+
+**Decision gate**: If the input cannot be parsed (invalid URL, inaccessible Jira ticket), report the error and ask the user for clarification.
+
+### Phase 2: Setup Fix Branch
+
+**Skill**: `setup-fix-branch`
+
+Create a feature branch based on the correct upstream release branch:
+
+```bash
+git fetch upstream <release-branch>
+git checkout -b fix/e2e-<test-description> upstream/<release-branch>
+```
+
+If a Jira ticket was provided, include the ticket ID in the branch name:
+`fix/RHIDP-XXXX-e2e-<test-description>`
+
+### Phase 3: Deploy RHDH
+
+**Skill**: `deploy-rhdh`
+
+Deploy RHDH to a cluster using `e2e-tests/local-run.sh`:
+
+```bash
+cd e2e-tests
+./local-run.sh -j <job-name> -t <image-tag> -s
+```
+
+Use deploy-only mode (`-s`) to skip automated test execution — we'll run the specific failing test manually.
+
+Select the image tag based on the release branch:
+- `main` → `next`
+- `release-1.9` → `1.9`
+- `release-1.8` → `1.8`
+
+After deployment completes, set up the local test environment:
+```bash
+source e2e-tests/local-test-setup.sh <showcase|rbac>
+```
+
+**Decision gate**: If deployment fails, the `deploy-rhdh` skill has error recovery procedures. If deployment cannot be recovered after investigation, report the deployment issue and stop.
+
+### Phase 4: Reproduce Failure
+
+**Skill**: `reproduce-failure`
+
+Run the specific failing test to confirm it reproduces locally:
+
+```bash
+cd e2e-tests
+yarn playwright test <spec-file> --project=<project> --retries=0 --workers=1
+```
+
+**Decision gates**:
+- **Consistent failure**: Proceed to Phase 5
+- **Flaky** (fails sometimes): Proceed to Phase 5, focus on reliability
+- **Cannot reproduce** (passes every time after 10 runs): Report that the failure cannot be reproduced locally, list possible environment differences, and ask the user how to proceed
+
+### Phase 5: Diagnose and Fix
+
+**Skill**: `diagnose-and-fix`
+
+Analyze the failure and implement a fix:
+
+1. **Classify the failure**: locator drift, timing, assertion mismatch, data dependency, platform-specific, deployment config
+2. **Use Playwright Test Agents**: Invoke the healer agent (`@playwright-test-healer`) for automated test repair — it can debug the test, inspect the UI, generate locators, and edit the code
+3. **Follow project conventions**: Use semantic selectors, Page Object Model, component annotations, proper utility classes
+4. **Cross-repo investigation**: If the issue is in deployment config, use Context7 or Sourcebot to search `rhdh-operator` and `rhdh-chart` repos
+
+**Decision gate**: If the analysis reveals a product bug (not a test issue):
+1. Mark the test as `test.fixme()` with a descriptive comment
+2. Report the product bug (update Jira ticket if applicable)
+3. Proceed to Phase 6 with the `test.fixme()` change
+
+### Phase 6: Verify Fix
+
+**Skill**: `verify-fix`
+
+Verify the fix:
+1. Run the fixed test once — must pass
+2. Run 5 times — must pass 5/5
+3. Run code quality checks: `yarn tsc:check`, `yarn lint:check`, `yarn prettier:check`
+4. Fix any lint/formatting issues
+
+**Decision gate**: If the test still fails or is flaky, return to Phase 5 and iterate.
+
+### Phase 7: Submit PR and Handle Review
+
+**Skill**: `submit-and-review`
+
+1. **Commit**: Stage changes, commit with conventional format
+2. **Push**: `git push -u origin <branch>`
+3. **Create PR**: Against `redhat-developer/rhdh`. Determine the GitHub username from the fork remote: `git remote get-url origin | sed 's|.*github.com[:/]||;s|/.*||'`. Then use `gh pr create --repo redhat-developer/rhdh --head <username>:<branch> --base <release-branch>`
+4. **Trigger Qodo review**: Comment `/agentic_review` on the PR
+5. **Wait for review**: Poll for Qodo bot comments (check every 60s, up to 10 minutes)
+6. **Address feedback**: Apply valid suggestions, explain rejections
+7. **Monitor CI**: Watch CI checks with `gh pr checks`
+
+### Final Report
+
+After all phases complete, produce a summary:
+
+```
+E2E Fix Summary:
+- Input: <Prow URL or Jira ticket>
+- Test: <spec file> (<playwright project>)
+- Branch: <fix branch> → <release branch>
+- Root cause: <classification and description>
+- Fix: <what was changed>
+- Verification: <X/X passes>
+- PR: <PR URL>
+- CI Status: <PASS/PENDING/FAIL>
+- Qodo Review: <status>
+```