verify-deploy

name	verify-deploy
description	Post-deployment health check. Verifies pod status, logs, events, Helm release, and ingress for a deployed application. Usage - /verify-deploy [application] [target]
user-invocable	true
allowed-tools	Bash(kubectl *), Bash(helm *), Bash(aws *), AskUserQuestion, Read, Grep, Glob
argument-hint	[application] [deployment-target]

Verify Deployment

Post-deployment health check that presents a GREEN/YELLOW/RED summary.

Steps

1. Determine parameters

If $ARGUMENTS is provided, parse application and deployment target from it. Otherwise, ask:

1. Application: Which service to check
2. Deployment target: Which namespace/deployment to check

2. Run health checks

Execute all checks and collect results:

2a. Pod Status

kubectl get pods -n {target} -l app={application} -o wide

• GREEN: All pods Running and Ready (READY column shows x/x)
• YELLOW: Pods Running but not all Ready, or recent restarts
• RED: CrashLoopBackOff, Error, Pending, ImagePullBackOff, or no pods found

2b. Recent Logs

kubectl logs -n {target} -l app={application} --tail=30 --since=5m

• GREEN: No ERROR/FATAL/CRITICAL lines
• YELLOW: WARNING lines present
• RED: ERROR/FATAL/CRITICAL lines present

2c. Warning Events

kubectl get events -n {target} --sort-by='.lastTimestamp' --field-selector type=Warning | tail -10

• GREEN: No warning events in last 10 minutes
• YELLOW: Warning events exist but unrelated to the target app
• RED: Warning events related to the target app

2d. Helm Release

helm list -n {target} --filter {application} -o json

• GREEN: Status "deployed", revision matches expected
• RED: Status "failed", "pending-install", or not found

3. Present summary

Deployment Health: {application} in {target}
================================================

  Pods:      GREEN  -- 3/3 Running, 0 restarts
  Logs:      YELLOW -- 2 WARNING lines (connection retry)
  Events:    GREEN  -- No warning events
  Helm:      GREEN  -- v1.2.3, revision 42, deployed

Overall: HEALTHY (1 advisory)

Advisories:
  - [YELLOW] Logs show connection retry warnings — may be transient

4. Offer next steps

• If RED: Offer to dig deeper into the failing component
• If YELLOW: Note the advisory items, offer to investigate
• If GREEN: Confirm deployment looks healthy

Safety

• Read-only skill — no state changes
• All commands are observational (get, list, logs, describe)
• If kubectl context is wrong, warn and ask to switch