- Home
- Skills
- DevOps & Cloud
- kube readonly triage
Works with the AI Agents you already use
kube readonly triage
Perform safe, read-only triage for unhealthy Kubernetes and OpenShift workloads to find root causes fast.
Free
kube readonly triage
Example session with this skill installed
Why is the backend pod crashing in the staging namespace?
- Read your context and instructions
- Compiled the kube readonly triage
I have performed a read-only triage of the backend workload in the staging namespace.
Triage Report
| Field | Details |
|---|---|
| Target | context-staging-cluster / staging (prod: no) |
| Symptom | Pod backend-7f8d9b-x2k4 is in CrashLoopBackOff. |
| Cause | OOMKilled (Exit Code 137). The pod's memory limit is set to 256Mi, but the last state shows it exceeded this threshold during startup. |
| Confidence | Confirmed |
| Not checked | Application-level internal heap dumps (no exec permission). |
| Evidence | lastState: { terminated: { exitCode: 137, reason: OOMKilled, ... } } |
Suggested fix (not run):
Increase the memory limit in the deployment specification:
kubectl patch deployment backend -n staging --type='json' -p='[{"op": "replace", "path": "/spec/template/spec/containers/0/resources/limits/memory", "value":"512Mi"}]'
Next steps
- Review the memory usage trends using
kubectl top pod backend-7f8d9b-x2k4 -n staging. - Apply the suggested memory limit increase if the application requires a larger heap size for initialization.
- Check the application logs for potential memory leaks if this occurs frequently after startup.
Connects securely to your tools. The creator never sees your data.
About this skill
Kubernetes clusters are complex, and pinpointing why a workload is down often involves a repetitive, high-stakes hunt through logs and events. Human error during this process—like accidentally deleting a production pod or misconfiguring a live service—is a constant risk. This skill provides a systematic, read-only methodology for triaging unhealthy workloads on Kubernetes and OpenShift without the risk of altering cluster state.
What it does
- Health diagnostics scans pods, deployments, statefulsets, and jobs to identify specific failure states.
- Evidence gathering automatically executes
describeandlogs --previousto find exit codes and event-based causes. - Scheduling analysis identifies CPU/memory pressure, taint mismatches, and unbound PVCs for pending pods.
- Networking triage traces the connectivity chain from Ingress/Routes through Services to pod Endpoints and selectors.
- Security validation checks image pull secrets and OpenShift Security Context Constraints (SCC) without exposing sensitive values.
How it works
- Context Verification: The agent runs
auth whoamiandconfig current-contextto ensure the session is valid and safe. - Workload Scanning: It fetches pod statuses and sorts events by timestamp to find the immediate point of failure.
- Deep Inspection: Depending on the failure (e.g.,
CrashLoopBackOffvsPending), it follows a targeted branch to read specific logs or describe resources. - Structured Reporting: It delivers a triage report including the symptom, evidence-backed cause, and a suggested fix for the user to execute.
Frameworks & tools
Uses kubectl or oc (OpenShift) CLI tools. Compatible with any standard Kubernetes distribution.
Why this beats prompting it yourself
Generic prompts often lead to agents attempting exec commands or accidentally modifying the cluster. This skill enforces a strict read-only protocol and uses specific logic branches—like checking previous container logs for crashes—that standard LLM behavior frequently misses.
Use cases
- CrashLoop debugging finding specific application exit codes and OOMKill signals.
- Scheduling issues identifying why a pod is stuck in
Pendingdue to resource limits or node taints. - Service connectivity diagnosing why a
Runningpod is not receiving traffic due to selector mismatches.
Known limitations
Does not support reading Secret values or executing commands inside containers. Limited to the 60-minute default event history of the cluster.
Read-only commands only; no exec or edit. Cannot view Secret content. Triage is limited by the cluster's event retention period (typically 1 hour).
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 23 days ago
- Free with an Agensi account