diff --git a/plugins/jira/evals/cases/create/case-001-story-cntrlplane-basic/annotations.yaml b/plugins/jira/evals/cases/create/case-001-story-cntrlplane-basic/annotations.yaml
new file mode 100644
index 000000000..cc04bf739
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-001-story-cntrlplane-basic/annotations.yaml
@@ -0,0 +1,19 @@
+expected_issue_type: Story
+expected_project_key: CNTRLPLANE
+expected_conventions: [cntrlplane]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: customfield_10855
+expected_version_format: array
+expected_affects_version: null
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: story
+notes: >
+  Basic CNTRLPLANE story with --version flag. Tests version normalization
+  (4.22 → openshift-4.22) and array format for customfield_10855
+  [{"id": "PLACEHOLDER_VERSION_ID"}]. Description should follow user story
+  template (As a / I want / so that) with acceptance criteria.
diff --git a/plugins/jira/evals/cases/create/case-001-story-cntrlplane-basic/input.yaml b/plugins/jira/evals/cases/create/case-001-story-cntrlplane-basic/input.yaml
new file mode 100644
index 000000000..071c3ba3e
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-001-story-cntrlplane-basic/input.yaml
@@ -0,0 +1,8 @@
+type: story
+project_key: CNTRLPLANE
+summary: "Add user dashboard for cluster metrics"
+flags: "--version 4.22"
+interactive_responses: |
+  Who: cluster administrators
+  What: view real-time cluster metrics in a unified dashboard
+  Why: reduce time spent switching between monitoring tools
diff --git a/plugins/jira/evals/cases/create/case-002-bug-ocpbugs-basic/annotations.yaml b/plugins/jira/evals/cases/create/case-002-bug-ocpbugs-basic/annotations.yaml
new file mode 100644
index 000000000..3c0cc9bc0
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-002-bug-ocpbugs-basic/annotations.yaml
@@ -0,0 +1,20 @@
+expected_issue_type: Bug
+expected_project_key: OCPBUGS
+expected_conventions: [ocpbugs]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: customfield_10855
+expected_version_format: string
+expected_affects_version: [{"name": "4.21"}]
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: bug
+notes: >
+  OCPBUGS bug with both version fields. Tests that Affects Version uses
+  versions field as [{"name": "4.21"}] and Target Version uses
+  customfield_10855 as plain string "openshift-4.21" (NOT array format).
+  Description should follow bug template with problem description, version,
+  reproducibility, steps, actual/expected results.
diff --git a/plugins/jira/evals/cases/create/case-002-bug-ocpbugs-basic/input.yaml b/plugins/jira/evals/cases/create/case-002-bug-ocpbugs-basic/input.yaml
new file mode 100644
index 000000000..14040f67d
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-002-bug-ocpbugs-basic/input.yaml
@@ -0,0 +1,15 @@
+type: bug
+project_key: OCPBUGS
+summary: "API server returns 500 when listing pods with label selector"
+flags: ""
+interactive_responses: |
+  Affects version: 4.21
+  Target version: openshift-4.21
+  Reproducibility: Always
+  Steps to reproduce:
+    1. Create a pod with label app=test
+    2. Run oc get pods -l app=test
+    3. Observe 500 error in response
+  Actual results: API server returns HTTP 500 Internal Server Error
+  Expected results: API server returns the list of matching pods
+  Additional info: Occurs on clusters with more than 100 namespaces
diff --git a/plugins/jira/evals/cases/create/case-003-epic-cntrlplane/annotations.yaml b/plugins/jira/evals/cases/create/case-003-epic-cntrlplane/annotations.yaml
new file mode 100644
index 000000000..99a9c4601
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-003-epic-cntrlplane/annotations.yaml
@@ -0,0 +1,21 @@
+expected_issue_type: Epic
+expected_project_key: CNTRLPLANE
+expected_conventions: [cntrlplane]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: null
+expected_version_format: null
+expected_affects_version: null
+expected_custom_fields:
+  customfield_10011: "Multi-cluster metrics aggregation"
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: epic
+notes: >
+  CNTRLPLANE epic with parent linking. Tests Epic Name custom field
+  (customfield_10011) which must match the summary exactly. Also tests
+  parent linking via additional_fields.parent.key = "CNTRLPLANE-500".
+  Description should follow epic template with objective, scope, and
+  acceptance criteria.
diff --git a/plugins/jira/evals/cases/create/case-003-epic-cntrlplane/input.yaml b/plugins/jira/evals/cases/create/case-003-epic-cntrlplane/input.yaml
new file mode 100644
index 000000000..7d78e0a7d
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-003-epic-cntrlplane/input.yaml
@@ -0,0 +1,12 @@
+type: epic
+project_key: CNTRLPLANE
+summary: "Multi-cluster metrics aggregation"
+flags: "--parent CNTRLPLANE-500"
+interactive_responses: |
+  Objective: Enable aggregation of metrics across multiple managed clusters into a unified view
+  Scope: Prometheus federation, custom metrics pipeline, dashboard integration
+  Acceptance criteria:
+    1. Metrics from up to 50 managed clusters aggregated within 5 minutes
+    2. Custom dashboards show cross-cluster resource utilization
+    3. Alert rules can reference metrics from any managed cluster
+  Timeframe: 2-3 sprints
diff --git a/plugins/jira/evals/cases/create/case-004-story-hypershift-aro/annotations.yaml b/plugins/jira/evals/cases/create/case-004-story-hypershift-aro/annotations.yaml
new file mode 100644
index 000000000..bca283cd9
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-004-story-hypershift-aro/annotations.yaml
@@ -0,0 +1,19 @@
+expected_issue_type: Story
+expected_project_key: CNTRLPLANE
+expected_conventions: [cntrlplane, hypershift]
+expected_labels: ["ai-generated-jira", "aro-hcp"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: "HyperShift / ARO"
+expected_version_field: customfield_10855
+expected_version_format: array
+expected_affects_version: null
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: story
+notes: >
+  CNTRLPLANE story with HyperShift ARO conventions layered. Tests that
+  "ARO HCP" in the summary triggers auto-detection of component
+  "HyperShift / ARO" and label "aro-hcp". Both cntrlplane and hypershift
+  conventions should be loaded. Version uses CNTRLPLANE array format.
diff --git a/plugins/jira/evals/cases/create/case-004-story-hypershift-aro/input.yaml b/plugins/jira/evals/cases/create/case-004-story-hypershift-aro/input.yaml
new file mode 100644
index 000000000..9d5f095ad
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-004-story-hypershift-aro/input.yaml
@@ -0,0 +1,8 @@
+type: story
+project_key: CNTRLPLANE
+summary: "Enable private API server for ARO HCP clusters"
+flags: "--version 4.22"
+interactive_responses: |
+  Who: ARO HCP cluster administrators
+  What: configure private API server endpoints for hosted control planes on Azure
+  Why: meet enterprise network isolation requirements for regulated workloads
diff --git a/plugins/jira/evals/cases/create/case-005-bug-ocpbugs-hypershift-rosa/annotations.yaml b/plugins/jira/evals/cases/create/case-005-bug-ocpbugs-hypershift-rosa/annotations.yaml
new file mode 100644
index 000000000..8de46977d
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-005-bug-ocpbugs-hypershift-rosa/annotations.yaml
@@ -0,0 +1,19 @@
+expected_issue_type: Bug
+expected_project_key: OCPBUGS
+expected_conventions: [ocpbugs, hypershift]
+expected_labels: ["ai-generated-jira", "rosa-hcp"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: "HyperShift / ROSA"
+expected_version_field: customfield_10855
+expected_version_format: string
+expected_affects_version: [{"name": "4.21"}]
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: bug
+notes: >
+  OCPBUGS bug with both ocpbugs and hypershift conventions layered. Tests
+  that "ROSA HCP" in summary triggers component "HyperShift / ROSA" and
+  label "rosa-hcp". Version fields use OCPBUGS format: versions as
+  [{"name": "4.21"}] and customfield_10855 as string "openshift-4.22".
diff --git a/plugins/jira/evals/cases/create/case-005-bug-ocpbugs-hypershift-rosa/input.yaml b/plugins/jira/evals/cases/create/case-005-bug-ocpbugs-hypershift-rosa/input.yaml
new file mode 100644
index 000000000..743b1ecde
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-005-bug-ocpbugs-hypershift-rosa/input.yaml
@@ -0,0 +1,15 @@
+type: bug
+project_key: OCPBUGS
+summary: "ROSA HCP cluster upgrade fails with OLM operator timeout"
+flags: ""
+interactive_responses: |
+  Affects version: 4.21
+  Target version: openshift-4.22
+  Reproducibility: Sometimes
+  Steps to reproduce:
+    1. Create ROSA HCP cluster on version 4.21
+    2. Initiate upgrade to 4.22
+    3. Wait for upgrade to complete
+  Actual results: Upgrade stalls at 60% with OLM operator timeout after 30 minutes
+  Expected results: Upgrade completes successfully within 45 minutes
+  Additional info: Affects approximately 30% of upgrade attempts on clusters with 5+ installed operators
diff --git a/plugins/jira/evals/cases/create/case-006-task-cntrlplane-plain/annotations.yaml b/plugins/jira/evals/cases/create/case-006-task-cntrlplane-plain/annotations.yaml
new file mode 100644
index 000000000..2c63b17c7
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-006-task-cntrlplane-plain/annotations.yaml
@@ -0,0 +1,19 @@
+expected_issue_type: Task
+expected_project_key: CNTRLPLANE
+expected_conventions: [cntrlplane]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: null
+expected_version_format: null
+expected_affects_version: null
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: task
+notes: >
+  Plain CNTRLPLANE task with no team conventions beyond project level.
+  No version or component specified. Description should follow task
+  template with what/why and acceptance criteria. Tests that the skill
+  correctly identifies this as non-user-facing technical work.
diff --git a/plugins/jira/evals/cases/create/case-006-task-cntrlplane-plain/input.yaml b/plugins/jira/evals/cases/create/case-006-task-cntrlplane-plain/input.yaml
new file mode 100644
index 000000000..ecc375ed8
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-006-task-cntrlplane-plain/input.yaml
@@ -0,0 +1,11 @@
+type: task
+project_key: CNTRLPLANE
+summary: "Update API documentation for cluster provisioning endpoints"
+flags: ""
+interactive_responses: |
+  What: Update the API documentation to reflect recent changes to the cluster provisioning endpoints
+  Why: Current documentation is outdated after the v2 API migration
+  Acceptance criteria:
+    1. All provisioning endpoints documented with request/response examples
+    2. Deprecation notices added for v1 endpoints
+    3. OpenAPI spec updated and validated
diff --git a/plugins/jira/evals/cases/create/case-007-feature-cntrlplane/annotations.yaml b/plugins/jira/evals/cases/create/case-007-feature-cntrlplane/annotations.yaml
new file mode 100644
index 000000000..4f505fb34
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-007-feature-cntrlplane/annotations.yaml
@@ -0,0 +1,19 @@
+expected_issue_type: Feature
+expected_project_key: CNTRLPLANE
+expected_conventions: [cntrlplane]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: null
+expected_version_format: null
+expected_affects_version: null
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: feature
+notes: >
+  CNTRLPLANE feature with strategic-level content. No version or component.
+  Description should follow feature template with market problem, strategic
+  value, and success criteria. Tests that the skill applies the feature
+  reference guide correctly.
diff --git a/plugins/jira/evals/cases/create/case-007-feature-cntrlplane/input.yaml b/plugins/jira/evals/cases/create/case-007-feature-cntrlplane/input.yaml
new file mode 100644
index 000000000..64d7dca5f
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-007-feature-cntrlplane/input.yaml
@@ -0,0 +1,12 @@
+type: feature
+project_key: CNTRLPLANE
+summary: "Advanced search capabilities for cluster fleet management"
+flags: ""
+interactive_responses: |
+  Market problem: Platform engineers managing 100+ clusters cannot efficiently find and filter clusters by custom attributes, leading to operational delays
+  Strategic value: Reduces mean time to cluster identification from 15 minutes to under 30 seconds, directly impacting incident response time
+  Success criteria:
+    - 80% of fleet management users adopt search within 3 months
+    - Average search latency under 500ms for fleets of 500+ clusters
+    - Support for at least 20 filterable attributes including custom labels
+  Timeline: Spans releases 4.22 through 4.23
diff --git a/plugins/jira/evals/cases/create/case-008-feature-request-rfe/annotations.yaml b/plugins/jira/evals/cases/create/case-008-feature-request-rfe/annotations.yaml
new file mode 100644
index 000000000..64fc47094
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-008-feature-request-rfe/annotations.yaml
@@ -0,0 +1,19 @@
+expected_issue_type: Feature Request
+expected_project_key: RFE
+expected_conventions: []
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: null
+expected_version_format: null
+expected_affects_version: null
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: feature-request
+notes: >
+  RFE project feature request. Tests default project key (RFE) and
+  issue type name ("Feature Request"). No project-specific conventions
+  apply. Description should follow the 4-question feature request
+  workflow (title, nature, business requirements, affected components).
diff --git a/plugins/jira/evals/cases/create/case-008-feature-request-rfe/input.yaml b/plugins/jira/evals/cases/create/case-008-feature-request-rfe/input.yaml
new file mode 100644
index 000000000..b4c2b56f0
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-008-feature-request-rfe/input.yaml
@@ -0,0 +1,9 @@
+type: feature-request
+project_key: RFE
+summary: "Support custom SSL certificates for ROSA HCP ingress"
+flags: ""
+interactive_responses: |
+  Title: Support custom SSL certificates for ROSA HCP ingress
+  Nature: Customers need to use their own CA-signed certificates for ingress controllers on ROSA HCP clusters instead of the default Let's Encrypt certificates
+  Business requirements: Enterprise customers in regulated industries (finance, healthcare) require custom certificates to meet compliance requirements. Currently they must use workarounds that break during cluster upgrades.
+  Affected packages/components: Ingress, Networking
diff --git a/plugins/jira/evals/cases/create/case-009-story-gcp-hcp/annotations.yaml b/plugins/jira/evals/cases/create/case-009-story-gcp-hcp/annotations.yaml
new file mode 100644
index 000000000..771a4f893
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-009-story-gcp-hcp/annotations.yaml
@@ -0,0 +1,22 @@
+expected_issue_type: Story
+expected_project_key: GCP
+expected_conventions: [gcp-hcp]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: "hypershift-operator-gcp"
+expected_version_field: null
+expected_version_format: null
+expected_affects_version: null
+expected_custom_fields:
+  customfield_10028: 2.0
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: gcp-story
+notes: >
+  GCP project story with GCP-HCP team conventions. Tests GCP-specific
+  template (user story + context + requirements + technical approach +
+  dependencies + acceptance criteria checklist). Should auto-estimate
+  Story Points (customfield_10028) using the GCP sizing guide — adding
+  a Prometheus metric is a 2-point task per the GCP examples. Component
+  provided via flag.
diff --git a/plugins/jira/evals/cases/create/case-009-story-gcp-hcp/input.yaml b/plugins/jira/evals/cases/create/case-009-story-gcp-hcp/input.yaml
new file mode 100644
index 000000000..a481b02a8
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-009-story-gcp-hcp/input.yaml
@@ -0,0 +1,16 @@
+type: story
+project_key: GCP
+summary: "Add Prometheus metric for hosted control plane CPU utilization"
+flags: "--component hypershift-operator-gcp"
+interactive_responses: |
+  Who: GCP HCP platform operations team
+  What: expose a Prometheus metric tracking CPU utilization per hosted control plane
+  Why: enable proactive scaling decisions and capacity planning for the management cluster
+  Context: Currently CPU utilization is only visible through GCP Cloud Monitoring, requiring context switching
+  Requirements: Metric must be available within 30 seconds of collection, labeled by cluster ID
+  Technical approach: Add a custom collector to the hypershift-operator-gcp that scrapes cgroup metrics
+  Dependencies: Requires prometheus-operator CRD access on the management cluster
+  Acceptance criteria:
+    - Metric hcp_cpu_utilization_ratio exposed on /metrics endpoint
+    - Grafana dashboard updated with CPU utilization panel
+    - Alert rule fires when utilization exceeds 80% for 5 minutes
diff --git a/plugins/jira/evals/cases/create/case-010-bug-ocpbugs-version-flag/annotations.yaml b/plugins/jira/evals/cases/create/case-010-bug-ocpbugs-version-flag/annotations.yaml
new file mode 100644
index 000000000..3e797ae65
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-010-bug-ocpbugs-version-flag/annotations.yaml
@@ -0,0 +1,19 @@
+expected_issue_type: Bug
+expected_project_key: OCPBUGS
+expected_conventions: [ocpbugs]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: customfield_10855
+expected_version_format: string
+expected_affects_version: [{"name": "4.21"}]
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: bug
+notes: >
+  OCPBUGS bug with --version flag. Tests that the version flag is applied
+  correctly with OCPBUGS conventions: customfield_10855 as plain string
+  "openshift-4.21" and versions as [{"name": "4.21"}]. No component
+  specified. Description should follow bug template.
diff --git a/plugins/jira/evals/cases/create/case-010-bug-ocpbugs-version-flag/input.yaml b/plugins/jira/evals/cases/create/case-010-bug-ocpbugs-version-flag/input.yaml
new file mode 100644
index 000000000..0682f2f04
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-010-bug-ocpbugs-version-flag/input.yaml
@@ -0,0 +1,15 @@
+type: bug
+project_key: OCPBUGS
+summary: "NodePort service intermittently unreachable on OVN-Kubernetes clusters"
+flags: "--version 4.21"
+interactive_responses: |
+  Affects version: 4.21
+  Target version: openshift-4.21
+  Reproducibility: Sometimes
+  Steps to reproduce:
+    1. Create a NodePort service on an OVN-Kubernetes cluster
+    2. Access the service via node IP and allocated port
+    3. Repeat access every 5 seconds for 10 minutes
+  Actual results: Approximately 5% of requests fail with connection refused
+  Expected results: All requests succeed consistently
+  Additional info: Issue is specific to OVN-Kubernetes; does not reproduce with OpenShift SDN
diff --git a/plugins/jira/evals/cases/create/case-011-story-hypershift-rosa/annotations.yaml b/plugins/jira/evals/cases/create/case-011-story-hypershift-rosa/annotations.yaml
new file mode 100644
index 000000000..2c1456a29
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-011-story-hypershift-rosa/annotations.yaml
@@ -0,0 +1,18 @@
+expected_issue_type: Story
+expected_project_key: CNTRLPLANE
+expected_conventions: [cntrlplane, hypershift]
+expected_labels: ["ai-generated-jira", "rosa-hcp"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: "HyperShift / ROSA"
+expected_version_field: customfield_10855
+expected_version_format: array
+expected_affects_version: null
+expected_custom_fields: {}
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: story
+notes: >
+  CNTRLPLANE story with ROSA HCP keywords triggering HyperShift conventions.
+  Tests auto-detection of "ROSA HCP" → component "HyperShift / ROSA" and
+  label "rosa-hcp". Version uses CNTRLPLANE array format.
diff --git a/plugins/jira/evals/cases/create/case-011-story-hypershift-rosa/input.yaml b/plugins/jira/evals/cases/create/case-011-story-hypershift-rosa/input.yaml
new file mode 100644
index 000000000..8f21fb9c7
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-011-story-hypershift-rosa/input.yaml
@@ -0,0 +1,8 @@
+type: story
+project_key: CNTRLPLANE
+summary: "Implement node auto-repair for ROSA HCP worker nodes"
+flags: "--version 4.22"
+interactive_responses: |
+  Who: ROSA HCP cluster operators
+  What: automatically detect and replace unhealthy worker nodes in hosted clusters on AWS
+  Why: minimize manual intervention and reduce cluster downtime caused by node failures
diff --git a/plugins/jira/evals/cases/create/case-012-epic-gcp-hcp/annotations.yaml b/plugins/jira/evals/cases/create/case-012-epic-gcp-hcp/annotations.yaml
new file mode 100644
index 000000000..2ed72606c
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-012-epic-gcp-hcp/annotations.yaml
@@ -0,0 +1,20 @@
+expected_issue_type: Epic
+expected_project_key: GCP
+expected_conventions: [gcp-hcp]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: null
+expected_version_format: null
+expected_affects_version: null
+expected_custom_fields:
+  customfield_10011: "Implement CMEK support for GCP hosted control plane storage"
+expected_validation_warnings: []
+expected_creation_blocked: false
+description_template: gcp-epic
+notes: >
+  GCP project epic with GCP-HCP team conventions. Tests Epic Name custom
+  field (customfield_10011) matching summary, and GCP epic template (use
+  case/context, current/desired state, scope, story breakdown, acceptance
+  criteria). No story points on epics.
diff --git a/plugins/jira/evals/cases/create/case-012-epic-gcp-hcp/input.yaml b/plugins/jira/evals/cases/create/case-012-epic-gcp-hcp/input.yaml
new file mode 100644
index 000000000..5b7de99e0
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-012-epic-gcp-hcp/input.yaml
@@ -0,0 +1,25 @@
+type: epic
+project_key: GCP
+summary: "Implement CMEK support for GCP hosted control plane storage"
+flags: ""
+interactive_responses: |
+  Use case: Enterprise customers require Customer-Managed Encryption Keys for all data at rest in hosted control planes
+  Current state: All etcd and PV storage uses Google-managed encryption keys with no option for customer control
+  Desired state: Customers can provide their own Cloud KMS keys for encrypting etcd data and persistent volumes in hosted control planes
+  Scope included:
+    - KMS key configuration during cluster creation
+    - Key rotation support
+    - Encryption status visibility in cluster dashboard
+  Scope excluded:
+    - Bring-your-own-HSM support
+    - Cross-region key replication
+  Story breakdown:
+    - Configure KMS key binding for etcd encryption
+    - Add PV encryption with customer KMS keys
+    - Build key rotation automation
+    - Add encryption status to cluster dashboard
+    - Write e2e tests for CMEK lifecycle
+  Acceptance criteria:
+    - Clusters created with customer KMS keys have all data encrypted with those keys
+    - Key rotation completes without cluster downtime
+    - Encryption status visible in cluster details API
diff --git a/plugins/jira/evals/cases/create/case-013-story-bad-summary/annotations.yaml b/plugins/jira/evals/cases/create/case-013-story-bad-summary/annotations.yaml
new file mode 100644
index 000000000..38e6cf3e8
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-013-story-bad-summary/annotations.yaml
@@ -0,0 +1,21 @@
+expected_issue_type: Story
+expected_project_key: CNTRLPLANE
+expected_conventions: [cntrlplane]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: null
+expected_version_format: null
+expected_affects_version: null
+expected_custom_fields: {}
+expected_validation_warnings: ["summary_starts_with_as_a"]
+expected_creation_blocked: false
+description_template: story
+notes: >
+  Tests summary validation — the summary starts with "As a" which is
+  an anti-pattern (user story format belongs in description, not summary).
+  The skill should detect this and warn in validation-results.json.
+  The interactive response says to use the suggested summary, so the
+  planned-call.json summary should be corrected to something concise like
+  "Enable ImageTagMirrorSet configuration in HostedCluster CRs".
diff --git a/plugins/jira/evals/cases/create/case-013-story-bad-summary/input.yaml b/plugins/jira/evals/cases/create/case-013-story-bad-summary/input.yaml
new file mode 100644
index 000000000..aaa5b9030
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-013-story-bad-summary/input.yaml
@@ -0,0 +1,9 @@
+type: story
+project_key: CNTRLPLANE
+summary: "As a cluster admin, I want to configure ImageTagMirrorSet in HostedCluster CRs so that I can enable tag-based image proxying"
+flags: ""
+interactive_responses: |
+  Use suggested summary: yes
+  Who: cluster administrators
+  What: configure ImageTagMirrorSet in HostedCluster custom resources
+  Why: enable tag-based image proxying for disconnected environments
diff --git a/plugins/jira/evals/cases/create/case-014-bug-credentials-in-content/annotations.yaml b/plugins/jira/evals/cases/create/case-014-bug-credentials-in-content/annotations.yaml
new file mode 100644
index 000000000..475b7295a
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-014-bug-credentials-in-content/annotations.yaml
@@ -0,0 +1,21 @@
+expected_issue_type: Bug
+expected_project_key: OCPBUGS
+expected_conventions: [ocpbugs]
+expected_labels: ["ai-generated-jira"]
+expected_security: "Red Hat Employee"
+expected_content_format: markdown
+expected_component: null
+expected_version_field: null
+expected_version_format: null
+expected_affects_version: null
+expected_custom_fields: {}
+expected_validation_warnings: ["credentials_detected"]
+expected_creation_blocked: true
+description_template: bug
+notes: >
+  Tests security validation — the interactive responses contain what
+  looks like AWS access key ID (AKIAIOSFODNN7EXAMPLE) and secret key.
+  The skill's Phase 6 security scan should detect these credentials and
+  BLOCK creation. planned-call.json should be empty ({}), and
+  validation-results.json should have creation_blocked: true with
+  security_warnings describing the type of credential found.
diff --git a/plugins/jira/evals/cases/create/case-014-bug-credentials-in-content/input.yaml b/plugins/jira/evals/cases/create/case-014-bug-credentials-in-content/input.yaml
new file mode 100644
index 000000000..681dcc95d
--- /dev/null
+++ b/plugins/jira/evals/cases/create/case-014-bug-credentials-in-content/input.yaml
@@ -0,0 +1,15 @@
+type: bug
+project_key: OCPBUGS
+summary: "OAuth token refresh fails silently on managed clusters"
+flags: ""
+interactive_responses: |
+  Affects version: 4.22
+  Target version: openshift-4.22
+  Reproducibility: Always
+  Steps to reproduce:
+    1. Configure OAuth with token AKIAIOSFODNN7EXAMPLE and secret wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY
+    2. Wait for token expiry (1 hour)
+    3. Attempt API call with expired token
+  Actual results: API call hangs indefinitely, no error returned
+  Expected results: Token refreshed automatically or clear error returned
+  Additional info: Cluster kubeconfig contains the service account token
diff --git a/plugins/jira/evals/eval-create.yaml b/plugins/jira/evals/eval-create.yaml
new file mode 100644
index 000000000..60f555648
--- /dev/null
+++ b/plugins/jira/evals/eval-create.yaml
@@ -0,0 +1,458 @@
+name: jira-create-eval
+description: Evaluate the jira:create skill's ability to prepare Jira issue creation calls with correct type-specific templates, project conventions, smart defaults, field formatting, and input validation
+skill: jira:create
+
+execution:
+  mode: case
+  arguments: "{type} {project_key} \"{summary}\" {flags}"
+  timeout: 300
+  max_budget_usd: 2.0
+
+runner:
+  type: claude-code
+  plugin_dirs:
+    - plugins/jira
+  system_prompt: |
+    Use the jira:create skill to prepare a Jira issue.
+
+    DRY RUN MODE: Do NOT call createJiraIssue or any Jira MCP tool.
+    Instead, after completing all preparation phases (load guidance, parse
+    arguments, apply defaults, interactive prompts, summary validation,
+    security validation), write the EXACT createJiraIssue call you would
+    have made to `planned-call.json` in the current working directory.
+
+    The JSON must match the createJiraIssue MCP tool signature:
+    {
+      "cloudId": "redhat.atlassian.net",
+      "projectKey": "...",
+      "issueTypeName": "...",
+      "summary": "...",
+      "description": "...",
+      "contentFormat": "markdown",
+      "additional_fields": {
+        "labels": [...],
+        "security": {"name": "..."},
+        "components": [...],
+        "customfield_10855": ...,
+        ...any other fields
+      }
+    }
+
+    Also write `validation-results.json` with:
+    {
+      "summary_warnings": [...],
+      "security_warnings": [...],
+      "creation_blocked": false
+    }
+
+    If security validation detects credentials, set creation_blocked to
+    true and leave planned-call.json empty (write {} only).
+
+    For interactive prompts, use these pre-filled responses instead of
+    prompting the user:
+
+    {interactive_responses}
+
+    For version ID lookups that would normally require
+    getJiraIssueTypeMetaWithFields, use placeholder ID "PLACEHOLDER_VERSION_ID".
+
+models:
+  judge: claude-opus-4-6
+
+permissions:
+  allow:
+    - "Skill"
+    - "Read"
+    - "Write"
+  deny:
+    - "Bash"
+    - "Agent"
+    - "mcp__*"
+
+mlflow:
+  experiment: jira-create-eval
+
+dataset:
+  path: plugins/jira/evals/cases/create
+  schema: |
+    Each case directory contains:
+    - input.yaml: YAML file with fields:
+      - 'type' — issue type: story, bug, epic, feature, task, or feature-request
+      - 'project_key' — Jira project key (e.g., CNTRLPLANE, OCPBUGS, GCP, RFE)
+      - 'summary' — issue title/summary text
+      - 'flags' — optional CLI flags (e.g., "--component HyperShift --version 4.22")
+      - 'interactive_responses' — pre-filled answers for the skill's interactive
+        prompts, formatted as the skill would expect them
+    - annotations.yaml: Expected outcomes for scoring:
+      - 'expected_issue_type': the Jira issue type name (Story, Bug, Epic, etc.)
+      - 'expected_project_key': the project key in the API call
+      - 'expected_conventions': list of convention files that should have been loaded
+        (e.g., [cntrlplane, hypershift])
+      - 'expected_labels': list of labels that must be present (always includes
+        "ai-generated-jira", may include team-specific labels)
+      - 'expected_security': security level name (always "Red Hat Employee")
+      - 'expected_content_format': always "markdown"
+      - 'expected_component': component name if one should be set, null otherwise
+      - 'expected_version_field': which custom field ID is used for target version
+        (customfield_10855), or null if no version expected
+      - 'expected_version_format': "array" for CNTRLPLANE-style [{"id": "..."}] or
+        "string" for OCPBUGS-style "openshift-X.Y", or null
+      - 'expected_affects_version': for OCPBUGS bugs, the versions field value, or null
+      - 'expected_custom_fields': dict of additional custom fields expected
+        (e.g., customfield_10011 for Epic Name, customfield_10028 for Story Points)
+      - 'expected_validation_warnings': list of expected summary/security warnings
+      - 'expected_creation_blocked': boolean, true if security validation should block
+      - 'description_template': which template the description should follow
+        (story, bug, epic, feature, task, feature-request, gcp-story, gcp-epic, etc.)
+      - 'notes': free-text explanation of the test case
+
+outputs:
+  - path: "output"
+    schema: |
+      The skill writes two files to the workspace root:
+      1. planned-call.json — the createJiraIssue arguments as a JSON object
+         with cloudId, projectKey, issueTypeName, summary, description,
+         contentFormat, and additional_fields.
+      2. validation-results.json — summary_warnings (array), security_warnings
+         (array), and creation_blocked (boolean).
+
+traces:
+  stdout: true
+  stderr: true
+  events: false
+  metrics: true
+
+judges:
+  - name: output_file_exists
+    description: Verify the skill produced planned-call.json in the workspace.
+    check: |
+      import os
+      files = outputs.get("files", {})
+      modified = outputs.get("modified_files", {})
+      all_f = {**files, **modified}
+      found = [k for k in all_f if os.path.basename(k) == "planned-call.json"]
+      if not found:
+          return (False, "No planned-call.json found")
+      return (True, f"planned-call.json found: {found[0]}")
+
+  - name: valid_json_structure
+    description: |
+      Verify planned-call.json is valid JSON with required top-level fields.
+      For blocked cases (security validation), {} is acceptable.
+    check: |
+      import json
+      import os
+      ann = outputs.get("annotations", {})
+      files = outputs.get("files", {})
+      modified = outputs.get("modified_files", {})
+      all_f = {**files, **modified}
+      found = {k: v for k, v in all_f.items()
+               if os.path.basename(k) == "planned-call.json"}
+      if not found:
+          return (False, "No planned-call.json found")
+      content = list(found.values())[0]
+      try:
+          data = json.loads(content)
+      except Exception as e:
+          return (False, f"Invalid JSON: {e}")
+      if ann.get("expected_creation_blocked"):
+          return (True, "Creation blocked — empty JSON is expected")
+      required = ["projectKey", "issueTypeName", "summary", "contentFormat"]
+      missing = [f for f in required if f not in data]
+      if missing:
+          return (False, f"Missing required fields: {', '.join(missing)}")
+      return (True, "Valid JSON with all required fields")
+
+  - name: universal_defaults_applied
+    description: |
+      Verify the universal defaults are present: labels include "ai-generated-jira",
+      security is "Red Hat Employee", contentFormat is "markdown".
+    check: |
+      import json
+      import os
+      ann = outputs.get("annotations", {})
+      if ann.get("expected_creation_blocked"):
+          return (True, "Creation blocked — defaults check skipped")
+      files = outputs.get("files", {})
+      modified = outputs.get("modified_files", {})
+      all_f = {**files, **modified}
+      found = {k: v for k, v in all_f.items()
+               if os.path.basename(k) == "planned-call.json"}
+      if not found:
+          return (False, "No planned-call.json found")
+      content = list(found.values())[0]
+      try:
+          data = json.loads(content)
+      except Exception as e:
+          return (False, f"Cannot parse JSON: {e}")
+      errors = []
+      cf = data.get("contentFormat", "")
+      if cf != "markdown":
+          errors.append(f"contentFormat is '{cf}', expected 'markdown'")
+      af = data.get("additional_fields", {})
+      labels = af.get("labels", data.get("labels", []))
+      if not isinstance(labels, list) or "ai-generated-jira" not in labels:
+          errors.append(f"labels missing 'ai-generated-jira': {labels}")
+      sec = af.get("security", data.get("security", {}))
+      sec_name = sec.get("name", "") if isinstance(sec, dict) else ""
+      if sec_name != "Red Hat Employee":
+          errors.append(f"security.name is '{sec_name}', expected 'Red Hat Employee'")
+      if errors:
+          return (False, "; ".join(errors))
+      return (True, "Universal defaults verified: ai-generated-jira label, "
+                     "Red Hat Employee security, markdown contentFormat")
+
+  - name: correct_issue_type
+    description: Verify issueTypeName matches the expected value from annotations.
+    check: |
+      import json
+      import os
+      ann = outputs.get("annotations", {})
+      if ann.get("expected_creation_blocked"):
+          return (True, "Creation blocked — issue type check skipped")
+      files = outputs.get("files", {})
+      modified = outputs.get("modified_files", {})
+      all_f = {**files, **modified}
+      found = {k: v for k, v in all_f.items()
+               if os.path.basename(k) == "planned-call.json"}
+      if not found:
+          return (False, "No planned-call.json found")
+      content = list(found.values())[0]
+      try:
+          data = json.loads(content)
+      except Exception as e:
+          return (False, f"Cannot parse JSON: {e}")
+      actual = data.get("issueTypeName", "")
+      expected = ann.get("expected_issue_type", "")
+      if actual.lower() == expected.lower():
+          return (True, f"Issue type correct: {actual}")
+      return (False, f"Issue type '{actual}', expected '{expected}'")
+
+  - name: correct_project_key
+    description: Verify projectKey matches the expected value from annotations.
+    check: |
+      import json
+      import os
+      ann = outputs.get("annotations", {})
+      if ann.get("expected_creation_blocked"):
+          return (True, "Creation blocked — project key check skipped")
+      files = outputs.get("files", {})
+      modified = outputs.get("modified_files", {})
+      all_f = {**files, **modified}
+      found = {k: v for k, v in all_f.items()
+               if os.path.basename(k) == "planned-call.json"}
+      if not found:
+          return (False, "No planned-call.json found")
+      content = list(found.values())[0]
+      try:
+          data = json.loads(content)
+      except Exception as e:
+          return (False, f"Cannot parse JSON: {e}")
+      actual = data.get("projectKey", "")
+      expected = ann.get("expected_project_key", "")
+      if actual == expected:
+          return (True, f"Project key correct: {actual}")
+      return (False, f"Project key '{actual}', expected '{expected}'")
+
+  - name: convention_fields_applied
+    description: |
+      Verify project/team-specific fields from annotations are present in the
+      planned call: component, version fields, custom fields, team-specific labels.
+    check: |
+      import json
+      import os
+      ann = outputs.get("annotations", {})
+      if ann.get("expected_creation_blocked"):
+          return (True, "Creation blocked — convention fields check skipped")
+      files = outputs.get("files", {})
+      modified = outputs.get("modified_files", {})
+      all_f = {**files, **modified}
+      found = {k: v for k, v in all_f.items()
+               if os.path.basename(k) == "planned-call.json"}
+      if not found:
+          return (False, "No planned-call.json found")
+      content = list(found.values())[0]
+      try:
+          data = json.loads(content)
+      except Exception as e:
+          return (False, f"Cannot parse JSON: {e}")
+      af = data.get("additional_fields", {})
+      errors = []
+      details = []
+      # Check component
+      exp_comp = ann.get("expected_component")
+      if exp_comp:
+          comps = af.get("components", data.get("components", []))
+          comp_names = []
+          if isinstance(comps, list):
+              comp_names = [c.get("name", c) if isinstance(c, dict) else str(c)
+                            for c in comps]
+          if exp_comp not in comp_names:
+              errors.append(f"Component '{exp_comp}' not found in {comp_names}")
+          else:
+              details.append(f"component={exp_comp}")
+      # Check version format
+      exp_vf = ann.get("expected_version_format")
+      if exp_vf:
+          tv = af.get("customfield_10855", data.get("customfield_10855"))
+          if tv is None:
+              errors.append("customfield_10855 (Target Version) not set")
+          elif exp_vf == "array" and not isinstance(tv, list):
+              errors.append(f"Target Version should be array, got {type(tv).__name__}: {tv}")
+          elif exp_vf == "string" and not isinstance(tv, str):
+              errors.append(f"Target Version should be string, got {type(tv).__name__}: {tv}")
+          else:
+              details.append(f"version_format={exp_vf}")
+      # Check affects version (OCPBUGS bugs)
+      exp_av = ann.get("expected_affects_version")
+      if exp_av:
+          vers = af.get("versions", data.get("versions"))
+          if vers is None:
+              errors.append("versions (Affects Version) not set")
+          else:
+              details.append(f"affects_version set")
+      # Check custom fields
+      exp_cf = ann.get("expected_custom_fields", {})
+      for field_id, expected_val in exp_cf.items():
+          actual_val = af.get(field_id, data.get(field_id))
+          if actual_val is None:
+              errors.append(f"Custom field {field_id} not set")
+          else:
+              details.append(f"{field_id} set")
+      # Check team-specific labels
+      exp_labels = ann.get("expected_labels", [])
+      actual_labels = af.get("labels", data.get("labels", []))
+      if not isinstance(actual_labels, list):
+          actual_labels = []
+      for lbl in exp_labels:
+          if lbl not in actual_labels:
+              errors.append(f"Label '{lbl}' missing from {actual_labels}")
+      if errors:
+          return (False, "; ".join(errors))
+      if not details:
+          details = ["no convention-specific fields expected"]
+      return (True, "Convention fields verified: " + ", ".join(details))
+
+  - name: description_quality
+    description: |
+      LLM judge assessing whether the description follows the correct type-specific
+      template and includes the required sections for that issue type.
+    prompt: |
+      You are evaluating whether a Jira issue description follows the correct
+      type-specific template from the jira:create skill.
+
+      The planned createJiraIssue call:
+
+      {% for path, content in outputs.files.items() if path.endswith('planned-call.json') %}
+      {{ content }}
+      {% endfor %}
+
+      Expected outcomes for this test case:
+
+      {{ annotations }}
+
+      The 'description_template' field indicates which template the description
+      should follow. Key template requirements:
+
+      - **story**: User story format (As a / I want / so that), acceptance criteria
+      - **bug**: Problem description, version, reproducibility, steps to reproduce,
+        actual results, expected results
+      - **epic**: Objective, scope, acceptance criteria (3-6 outcomes), Epic Name field
+      - **feature**: Market problem, strategic value, success criteria
+      - **task**: What/why, acceptance criteria, technical details
+      - **feature-request**: Title, nature/description, business requirements
+      - **gcp-story**: GCP team template — user story format, context/background,
+        requirements, technical approach, dependencies, acceptance criteria checklist
+      - **gcp-epic**: GCP template — use case/context, current/desired state, scope,
+        story breakdown, acceptance criteria
+
+      Evaluate on a 1-5 scale:
+
+      Score 1: No description, or description is a single line with no structure.
+      Score 2: Some structure but wrong template — e.g., a bug using story format,
+               or missing most required sections for the correct template.
+      Score 3: Correct template identified and partially followed — has some required
+               sections but is missing important ones (e.g., story without AC, bug
+               without steps to reproduce).
+      Score 4: Correct template fully followed with all required sections present.
+               Content is specific to the summary/context provided.
+      Score 5: Excellent — correct template, all sections present with specific and
+               actionable content, well-formatted Markdown.
+
+      If creation_blocked is true in annotations, evaluate only whether the skill
+      correctly identified the security issue — score 4 for correct detection,
+      5 for detection with helpful guidance about what was found.
+
+  - name: convention_compliance
+    description: |
+      LLM judge assessing whether the correct project and team conventions were
+      applied — version field formats, component selection, custom fields, labels —
+      matching the expected conventions listed in annotations.
+    prompt: |
+      You are evaluating whether the jira:create skill correctly applied project
+      and team conventions when preparing a Jira issue.
+
+      The planned createJiraIssue call:
+
+      {% for path, content in outputs.files.items() if path.endswith('planned-call.json') %}
+      {{ content }}
+      {% endfor %}
+
+      Validation results:
+
+      {% for path, content in outputs.files.items() if path.endswith('validation-results.json') %}
+      {{ content }}
+      {% endfor %}
+
+      Expected outcomes for this test case:
+
+      {{ annotations }}
+
+      The 'expected_conventions' field lists which convention files should have been
+      loaded. Key convention requirements by file:
+
+      - **cntrlplane**: Target Version (customfield_10855) in array format
+        [{"id": "VERSION_ID"}], version normalization (4.21 → openshift-4.21),
+        no fixVersions, components are team-specific not project-enforced
+      - **ocpbugs**: Bugs only, Affects Version (versions) as [{"name": "4.21"}],
+        Target Version (customfield_10855) as plain STRING "openshift-4.21" (NOT array),
+        no fixVersions
+      - **hypershift**: Component MUST be one of "HyperShift / ARO", "HyperShift / ROSA",
+        or "HyperShift" based on platform keywords; platform-specific labels
+        (aro-hcp, rosa-hcp)
+      - **gcp-hcp**: Story Points (customfield_10028) as float, GCP-specific components,
+        team templates for stories/epics/tasks, priority scheme
+
+      Evaluate on a 1-5 scale:
+
+      Score 1: Wrong conventions applied (e.g., OCPBUGS string version format used
+               for CNTRLPLANE, or no conventions loaded at all).
+      Score 2: Some conventions applied but with errors — wrong version field format,
+               missing required component for HyperShift, wrong custom field types.
+      Score 3: Correct conventions identified but incompletely applied — e.g., right
+               component but missing team labels, or right version field but wrong format.
+      Score 4: All expected conventions correctly applied — correct version field format,
+               correct component, correct labels, correct custom fields.
+      Score 5: Perfect convention compliance — all fields correct, version normalization
+               applied, team-specific fields present, no extraneous or incorrect fields.
+
+      If creation_blocked is true in annotations, score 4 if the skill stopped before
+      creating the call, 5 if it also correctly reported what it found.
+
+thresholds:
+  output_file_exists:
+    min_pass_rate: 1.0
+  valid_json_structure:
+    min_pass_rate: 1.0
+  universal_defaults_applied:
+    min_pass_rate: 1.0
+  correct_issue_type:
+    min_pass_rate: 1.0
+  correct_project_key:
+    min_pass_rate: 1.0
+  convention_fields_applied:
+    min_pass_rate: 0.85
+  description_quality:
+    min_mean: 3.5
+  convention_compliance:
+    min_mean: 3.5