redhat-developer RHDH Operator PR Test
Deploy the PR's CI-built bundle onto a live cluster and verify the change there.
GitHub code review belongs to /rhdh-pr-review.
Start here
Follow workflows/test-operator-pr.md from the top. Load
references/operator-pr-images.md when extracting or validating CI image URLs.
Gates
| Gate | Required check | If fail |
|---|---|---|
| Forge | /rhdh-forge is installed and returns the PR |
Stop; name /setup-rhdh-skills install |
oc |
oc is on PATH |
Stop; name /setup-rhdh-skills |
| Cluster | oc whoami succeeds |
Stop; the human runs oc login, or follows the rhdh-test-instance provision path in the workflow |
| Repository | The PR is redhat-developer/rhdh-operator |
Stop; this skill tests that repository only |
Boundaries
/rhdh-forgeowns GitHub reads. Invoke it by name with the PR URL or number and consume repository, number, state, files, diff, head and base refs, and comments. Do not parse the URL here or copy a fetch workflow./rhdh-pr-reviewowns analyzing the diff for merge comments and posting a GitHub review./rhdh-localowns a local compose runtime, not a live OpenShift cluster./rhdh-overlayowns overlay PR artifact checks, not operator bundles./rhdh-contextlocates an RHDH checkout when a localrhdh-test-instancetree is needed. Do not add it as a catalog dependency.
Writes
Posting a /test comment, applying or deleting cluster resources, and deploying
or rolling back an RHDH instance are external writes: invoke /mutation-gate
and follow it. A cluster operation's target names the namespace; a comment's
target names the repository and PR number. A request to test approves no write.
Deploy the full PR bundle or manifests, not only the operator binary image. Preserve the original cluster state and report it with cleanup.
Completion
Complete when the report names the PR and images used, one result per live check, the overall verdict, the deployed bundle or manifests, original and final cluster state, rollback commands, cleanup status, and the outcome of every approved write including skipped ones. Do not cache cluster state inside the skill directory.