Como usar
- Copie o comando e verifique o pacote antes de instalar no seu ambiente de agente.
- Consulte no original as ferramentas, credenciais e ambientes necessários.
- Descreva os dados de entrada e a entrega esperada; confira a saída com as fontes.
npx skills add https://github.com/beclab/olares --skill olares-clusterAntes de começar
- A instalação não inclui assinatura de modelo, franquia API ou acesso aos seus arquivos.
- Estrelas e instalações indicam atividade, não um teste independente de qualidade ou segurança.
Documento original do Skill
Renderizado do repositório de origem, preservando títulos, exemplos, código, tabelas, links e imagens.
cluster (per-user K8s view)
Shared front door: load `../olares-shared/SKILL.md` for suite routing, active-profile selection, platform entry points, and the auth proceed/stop gate. Load its auth reference only when login, profile switching, token storage, or auth recovery is actually needed.
Load the shared platform model when the task depends on app namespaces, application spaces, cross-namespace DNS, or system middleware. Use olares-cli cluster <noun> <verb> --help for syntax.
When to use
- Inspect K8s runtime objects, YAML, events, logs, or identity context.
- Execute a one-shot command in a container.
- Scale, restart, stop, start, or delete a workload; suspend/resume cronjobs; rerun jobs.
- Watch pod, workload, application, or log state.
Diagnosing why an app is broken* (stuck install, crash loop,runningbut unreachable, image won't pull, resource pressure) is `../olares-doctor/SKILL.md` — it orchestrates theseclustercommands into symptom→root-cause routing.clusterstays the raw runtime view and the place that mutates K8s objects.
workload stop/start changes controller replicas; it does not update the Market lifecycle row. Use market stop/resume for app-level lifecycle.
Fast paths
| Task | Read | First command |
|---|---|---|
| See whether an app's pods are up | pod operations | olares-cli cluster pod list -n <namespace> -o json, then read each .status.phase |
| Read a container's logs | pod operations | olares-cli cluster pod logs <pod> -n <namespace> |
| Find what this profile can see at all | this file | olares-cli cluster context -o json |
| Restart a workload | workload operations | olares-cli cluster workload restart <name> -n <namespace> |
The identity-vs-server-decides principle (cross-cutting)
- Identity is the active profile; there is no per-invocation profile override.
- The server decides visibility and authorization.
cluster contextis identity display, not a permission preflight. - Surface a 403 as authoritative. The exception is
exec, whose namespace safety gate is described in its reference.
Verb index
| Noun | Verbs | Read when triggered |
|---|---|---|
context | (single verb) | The ControlHub-side identity and scope of the active profile — what this tree can see at all, before any noun below returns empty |
pod | list, get, yaml, events, logs, delete, restart, exec | exec requires Olares 1.12.7+; pod operations; exec safety |
container | list, env, logs, exec | exec requires Olares 1.12.7+; exec safety |
workload (wl) | list, images, get, yaml, rollout-status, scale, restart, stop, start, delete | workload operations |
application (app) | list, get, workloads, pods, status | application aggregation |
namespace (alias ns) | list, get | Scoped to the active profile, so this is the user's namespaces rather than the cluster's |
node (alias nodes) | list, get | The node names other trees ask for: files addresses cache/<node>/ and external/<node>/ by them, and a file task is retained per node |
job (jobs) | list, get, yaml, pods, events, rerun | job operations |
cronjob (cronjobs, cj) | list, get, yaml, jobs, suspend, resume | cronjob operations |
middleware (mw) | list | middleware model |
Watch and asynchronous semantics
- Watches and log follow poll; they are not streaming API watches.
- Transient network/5xx failures are retried within a bounded streak. Terminal 4xx responses stop immediately; 408 and 429 remain retryable.
- Ctrl-C stops the local poll cleanly. It does not undo a mutation already accepted by the server.
- A rollout/status watch ending is not proof that the application entrance is healthy; use
olares-doctorwhen runtime symptoms remain.
Safety and escalation
- Confirm every destructive runtime mutation before invoking it, even if
--yesis available. execis a remote code-execution boundary. Confirm namespace, pod, container, command, and whether writes are expected. Prefer one-shot execution; interactive TTY belongs to the human.- Do not delete or restart objects merely because a diagnostic found them unhealthy. Establish controller ownership and the app-level consequence first.
- A 404 may hide a namespace the profile cannot see. Compare against
cluster application list; do not broaden scope or switch identity without user approval. - For complete image-reference diagnostics use
doctor images, not a paginated workload listing. - Stop on ambiguous namespace/resource identity, missing permission, or any request to access another user's container.
