skill_name: debugging-pipeline-failures
description: Validates systematic Konflux pipeline debugging methodology using standard kubectl and Tekton CLI approaches

# Note: .config/gcloud is automatically copied for all tests (needed for Claude Code API auth)
# Add skill-specific paths here if needed (e.g., .kube/config for kubectl)

test_scenarios:
  - name: systematic-investigation-approach
    description: Validates that Claude follows systematic phases for pipeline failure investigation
    prompt: "My PipelineRun 'component-build-xyz' in namespace 'user-tenant' failed. How do I investigate?"
    model: sonnet
    samples: 3
    expected:
      contains_keywords:
        - kubectl
        - pipelinerun
        - taskrun
        - logs
        - events
        - systematic
    baseline_failure: "Without skill, may suggest random checks without systematic methodology"

  - name: image-pull-failure-diagnosis
    description: Tests diagnosis approach for ImagePullBackOff errors in build pipelines
    prompt: "My build pipeline is failing with ImagePullBackOff. How do I debug this?"
    model: sonnet
    samples: 3
    expected:
      contains_keywords:
        - describe pod
        - events
        - image
        - registry
        - imagePullSecrets
    baseline_failure: "Without skill, may not check pod events or ServiceAccount imagePullSecrets"

  - name: stuck-pipeline-investigation
    description: Validates methodology for pipelines stuck in Running state
    prompt: "My PipelineRun has been in Running state for over an hour. What should I check?"
    model: sonnet
    samples: 3
    expected:
      contains_keywords:
        - taskrun
        - status
        - pending
        - running
        - logs
    baseline_failure: "Without skill, may not check individual TaskRun statuses to identify stuck component"

  - name: resource-constraint-recognition
    description: Tests identification of resource exhaustion issues
    prompt: "TaskRun is stuck in Pending state. How do I find out why?"
    model: sonnet
    samples: 3
    expected:
      contains_keywords:
        - ["events", "describe"]
        - ["quota", "limits", "resources"]
        - ["FailedScheduling", "pending", "stuck", "scheduling"]
    baseline_failure: "Without skill, may not check events for FailedScheduling or namespace quotas"

  - name: log-analysis-methodology
    description: Ensures proper log analysis approach for build failures
    prompt: "My build TaskRun failed with exit code 1. How do I find the error?"
    model: sonnet
    samples: 3
    expected:
      contains_keywords:
        - kubectl logs
        - step
        - container
        - error
    baseline_failure: "Without skill, may not know how to get logs from specific Tekton step containers"

  - name: root-cause-vs-symptom
    description: Validates distinction between symptoms and root causes
    prompt: "My pipeline keeps failing. Should I just increase the timeout?"
    model: sonnet
    samples: 3
    expected:
      contains_keywords:
        - ["logs", "describe", "output", "details"]
        - ["root cause", "proper fix", "actual problem", "actually doing", "pinpoint"]
        - ["investigate", "identify", "check", "find out", "determine"]
        - ["why", "what"]
      does_not_contain:
        - just increase
        - simply add
    baseline_failure: "Without skill, may suggest increasing timeout without investigating root cause"