e2e-parse-ci-failure - SKILL.md Agent Skill

name: e2e-parse-ci-failure description: Parse a Prow CI job URL or Jira ticket to extract E2E test failure details including test name, spec file, release branch, platform, and error messages

Parse CI Failure

Extract structured failure context from a Prow job URL or Jira ticket for an RHDH E2E CI failure.

When to Use

Use this skill when you receive a failing Prow job URL (e.g., https://prow.ci.openshift.org/view/gs/...), a Jira ticket ID (e.g., RHIDP-XXXX), or a Jira URL (e.g., https://redhat.atlassian.net/browse/RHIDP-XXXX) for an E2E test failure and need to extract all relevant details before starting a fix.

Input Detection

Playwright report URL: URL ending in index.html (with optional #?testId=... fragment) — use Playwright MCP if available (see "Playwright Report Parsing" below), otherwise fall back to build log parsing
Prow URL: Starts with https://prow.ci.openshift.org/ — parse the job page and build log
Jira ticket ID: Matches pattern RHIDP-\d+ or similar — use Jira MCP tools to read the ticket
Jira URL: Starts with https://redhat.atlassian.net/browse/ — extract the ticket ID from the URL path (e.g., RHIDP-XXXX from https://redhat.atlassian.net/browse/RHIDP-XXXX) and then use Jira MCP tools to read the ticket

Prow URL Parsing

URL Structure

Prow job URLs follow two patterns:

Periodic/postsubmit: https://prow.ci.openshift.org/view/gs/test-platform-results/logs/<job-name>/<build-id>
Presubmit (PR): https://prow.ci.openshift.org/view/gs/test-platform-results/pr-logs/pull/redhat-developer_rhdh/<pr-number>/<job-name>/<build-id>

Extract <job-name> and <build-id> from the URL path. These are the two key values needed for all derivations.

GCS URL Derivation

Convert the Prow URL to a GCS artifacts URL by replacing the prefix:

Prow:  https://prow.ci.openshift.org/view/gs/test-platform-results/logs/<job-name>/<build-id>
GCS:   https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/test-platform-results/logs/<job-name>/<build-id>/artifacts/

For presubmit jobs, use pr-logs/pull/redhat-developer_rhdh/<pr-number>/ instead of logs/.

Key artifacts within the GCS directory:

Build log: <step-name>/build-log.txt
JUnit XML: <step-name>/artifacts/junit-results/results.xml
Playwright report: <step-name>/artifacts/playwright-report/

Fetch the Prow job page with WebFetch to find the job status and artifact links, then fetch the build log for test failure details.

Extracting Test Failures from Build Log

Search the build log for these Playwright output patterns:

# Failing test line (primary source for test name, spec file, and project):
  ✘  [<project>] › <path>/<spec-file>.spec.ts:<line> › <test-describe> › <test-name>

# Error details (immediately after the failure line):
  Error: <error-message>
  expect(received).toBeVisible()
  Locator: <locator>

# Summary (at the end of the log):
  X failed
  X passed
  X skipped

Also check JUnit XML for <testcase> elements with <failure> children as a fallback.

Playwright Report Parsing

When the URL points to a Playwright HTML report (index.html, optionally with #?testId=...), use Playwright MCP if available — navigate with browser_navigate, then browser_snapshot to extract test name, spec file, error, steps, retries, screenshots, and traces from the accessibility tree. Derive job metadata (<job-name>, <build-id>, <project>) from the URL path segments.

If Playwright MCP is not available, derive the build-log.txt URL from the report URL and fall back to build log parsing.

Jira Ticket Parsing

Use Jira MCP tools to read the ticket. Extract:

Prow job URLs from the description or comments — then parse them using the Prow URL Parsing steps above.
Test names, spec file paths, error messages, or stack traces from the description, comments, or attachments.
affects version field — map to release branch (e.g., 1.10 → main, 1.9 → release-1.9, 1.8 → release-1.8).
component field for additional context (e.g., "E2E Tests", "CI/CD").

Job Name Mapping

Refer to the e2e-fix-workflow rule for all mapping tables: job name to release branch, job name to platform and deployment method, job name to Playwright projects, release branch to image repo/tag, and job name to local-run.sh -j parameter. Those tables are the single source of truth and should not be duplicated here.

When parsing a job name, apply those mapping tables to derive: release branch, platform, deployment method, Playwright projects, and local-run.sh flags (-j, -r, -t).

Fields Requiring Build Log Access

Not all output fields can be derived from the Prow URL alone. The following table clarifies what requires fetching the build log or artifacts:

Field	Source	Derivable from URL alone?
Job name	URL path segment	Yes
Build ID	URL path segment	Yes
Release branch	Job name pattern match	Yes
Platform	Job name pattern match	Yes
Deployment method	Job name pattern match	Yes
Playwright projects	Job name pattern match	Yes
`local-run.sh` flags (`-j`, `-r`, `-t`)	Job name + release branch	Yes
GCS artifacts URL	Constructed from URL	Yes
Test name	Build log Playwright output	No — requires build log
Spec file	Build log Playwright output	No — requires build log
Specific Playwright project (of failing test)	Build log `[project]` prefix	No — requires build log
Error type	Build log error details	No — requires build log
Error message	Build log error details	No — requires build log
Failure count / pass count	Build log summary line	No — requires build log

Output

Produce the following structured output with three sections.

1. Structured Summary

- Test name: <full test name from Playwright output>
- Spec file: <path relative to e2e-tests/playwright/>
- Playwright project: <specific project from the [project] prefix in the failure line>
- Release branch: <main or release-X.Y>
- Platform: <OCP/AKS/EKS/GKE/OSD-GCP>
- Deployment method: <Helm/Operator>
- Error type: <locator/timeout/assertion/crash/deployment>
- Error message: <the actual error>
- Prow URL: <original URL>
- Jira ticket: <ticket ID if applicable>

2. Derivation Details

Show how each field was derived with the matching pattern. This makes the reasoning transparent and auditable.

| Field              | Value                        | Derivation                                                |
|--------------------|------------------------------|-----------------------------------------------------------|
| Job name           | <full-job-name>              | Extracted from URL path segment                           |
| Build ID           | <build-id>                   | Extracted from URL path segment                           |
| Release branch     | <branch>                     | Pattern `*-rhdh-<branch>-*` matched in job name           |
| Platform           | <platform>                   | Pattern `*<platform-keyword>*` matched in job name        |
| Deployment method  | <method>                     | Pattern `*<method-keyword>*` matched in job name          |
| Playwright project | <project>                    | `[<project>]` prefix in failing test line                 |
| Image repo (-r)    | <repo>                       | Release branch `<branch>` maps to `<repo>`                |
| Image tag (-t)     | <tag>                        | Release branch `<branch>` maps to `<tag>`                 |
| Test name          | <test-name>                  | Parsed from `✘` line in build log                         |
| Spec file          | <spec-file>                  | Parsed from `✘` line in build log                         |
| Error type         | <error-type>                 | Classified from error message pattern                     |

3. GCS Artifacts Location

Derive and present the GCS artifacts URLs constructed from the Prow URL:

GCS Artifacts Base:
  https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/test-platform-results/logs/<job-name>/<build-id>/artifacts/

Build Log:
  <base>/artifacts/<step-name>/build-log.txt

JUnit Results:
  <base>/artifacts/<step-name>/artifacts/junit-results/results.xml

Playwright Report:
  <base>/artifacts/<step-name>/artifacts/playwright-report/

For presubmit (PR) jobs, the base path uses pr-logs/pull/redhat-developer_rhdh/<pr-number>/ instead of logs/.

4. local-run.sh Command

Provide the full command ready to copy-paste, with a flag breakdown.

OCP jobs — use -s for deploy-only mode:

cd e2e-tests
./local-run.sh -j <full-job-name> -r <image-repo> -t <image-tag> -s

Flag breakdown:
| Flag | Value              | Reason                                           |
|------|--------------------|--------------------------------------------------|
| -j   | <full-job-name>    | Full Prow job name (matches glob in CI script)   |
| -r   | <image-repo>       | Image repo derived from release branch <branch>  |
| -t   | <image-tag>        | Image tag derived from release branch <branch>   |
| -s   | (no value)         | Deploy only, skip running tests                  |

K8s jobs (AKS, EKS, GKE) — do not use -s; full execution is required:

cd e2e-tests
./local-run.sh -j <full-job-name> -r <image-repo> -t <image-tag>

Flag breakdown:
| Flag | Value              | Reason                                           |
|------|--------------------|--------------------------------------------------|
| -j   | <full-job-name>    | Full Prow job name (matches glob in CI script)   |
| -r   | <image-repo>       | Image repo derived from release branch <branch>  |
| -t   | <image-tag>        | Image tag derived from release branch <branch>   |