No code required. You need a Google GKE cluster and about an hour.
CI runs every test on Kind only. Nobody has verified KubeIntellect on Google GKE, so we genuinely do not know whether it works there — and ADOPTERS.md is honestly empty rather than padded.
Why this platform
Workload Identity and Autopilot's restrictions on privileged workloads are the interesting parts — Autopilot in particular rejects things Kind allows.
What to do
Follow the documented install path on a Google GKE cluster and record what actually happens:
pip install kubeintellect
kubeintellect init # setup wizard
kubeintellect serve # API server on :8000
Then ask it something real about the cluster and try one read-only query end to end.
Report, as a comment on this issue:
- Platform + version —
kubectl version, distro/managed flavour, node count
- Did the documented steps work verbatim? If not, the exact command and the exact error
- Any step the docs do not mention but that turned out to be required
- Whether RBAC behaved — a
readonly key must be refused a mutating command before any approval prompt appears
- What you would change in the docs — even one sentence
Acceptance criteria
A report saying "it did not work, here is exactly where it stopped" is more valuable than one saying it worked. That is the outcome we cannot get any other way.
⚠️ Redact before pasting — cluster output contains node names, namespaces and registries. And please use a non-production cluster, or a readonly key if it must be a real one.
Credit
This counts. You are named in the release notes exactly like a code contributor
(GOVERNANCE.md), and if you
add yourself to ADOPTERS.md
your environment becomes one of the ones we protect and keep backward-compatible.
(Note: a comment alone will not appear on GitHub's Contributors graph, which counts merged
commits. A one-line docs PR from what you find will — and is usually warranted.)
Task metadata
|
|
| Difficulty |
Beginner |
| Estimated time |
~1 hour |
| Skills |
Kubernetes operations. Access to a Google GKE cluster. No Python, no code. |
| Mentor |
@MSKazemi — ask on this issue, or in Q&A |
| Status |
🟢 Unclaimed — comment "I'd like to take this" and it will be assigned to you |
No code required. You need a Google GKE cluster and about an hour.
CI runs every test on Kind only. Nobody has verified KubeIntellect on Google GKE, so we genuinely do not know whether it works there — and ADOPTERS.md is honestly empty rather than padded.
Why this platform
Workload Identity and Autopilot's restrictions on privileged workloads are the interesting parts — Autopilot in particular rejects things Kind allows.
What to do
Follow the documented install path on a Google GKE cluster and record what actually happens:
Then ask it something real about the cluster and try one read-only query end to end.
Report, as a comment on this issue:
kubectl version, distro/managed flavour, node countreadonlykey must be refused a mutating command before any approval prompt appearsAcceptance criteria
v4/docs/— or just say so and someone else can write itA report saying "it did not work, here is exactly where it stopped" is more valuable than one saying it worked. That is the outcome we cannot get any other way.
readonlykey if it must be a real one.Credit
This counts. You are named in the release notes exactly like a code contributor
(GOVERNANCE.md), and if you
add yourself to ADOPTERS.md
your environment becomes one of the ones we protect and keep backward-compatible.
(Note: a comment alone will not appear on GitHub's Contributors graph, which counts merged
commits. A one-line docs PR from what you find will — and is usually warranted.)
Task metadata