cloudthinker-ai CAST AI Management Skill
Manage Kubernetes cluster optimization and cost reduction with CAST AI.
MANDATORY: Discovery-First Pattern
Always list connected clusters and their optimization status before querying specifics.
Phase 1: Discovery
#!/bin/bash
castai_api() {
local method="${1:-GET}"
local endpoint="$2"
local data="${3:-}"
if [ -n "$data" ]; then
curl -s -X "$method" -H "X-API-Key: $CASTAI_API_KEY" \
-H "Content-Type: application/json" \
"https://api.cast.ai/v1/${endpoint}" -d "$data"
else
curl -s -X "$method" -H "X-API-Key: $CASTAI_API_KEY" \
"https://api.cast.ai/v1/${endpoint}"
fi
}
echo "=== Connected Clusters ==="
castai_api GET "kubernetes/external-clusters" | jq -r '
.items[] | "\(.id)\t\(.name)\t\(.providerType)\t\(.agentStatus)\t\(.status)"
' | column -t
echo ""
echo "=== Optimization Status ==="
castai_api GET "kubernetes/external-clusters" | jq -r '
.items[] | "\(.name)\tAutoscaler: \(.autoscalerStatus // "unknown")\tSpot: \(.spotEnabled // false)\tPolicies: \(.optimizationEnabled // false)"
' | column -t
Core Helper Functions
#!/bin/bash
castai_api() {
local method="${1:-GET}"
local endpoint="$2"
local data="${3:-}"
if [ -n "$data" ]; then
curl -s -X "$method" -H "X-API-Key: $CASTAI_API_KEY" \
-H "Content-Type: application/json" \
"https://api.cast.ai/v1/${endpoint}" -d "$data"
else
curl -s -X "$method" -H "X-API-Key: $CASTAI_API_KEY" \
"https://api.cast.ai/v1/${endpoint}"
fi
}
Output Rules
- TOKEN EFFICIENCY: Target ≤50 lines per output
- Filter API responses with jq to extract cost and optimization fields
- Round costs and percentages to 2 decimal places
Common Operations
Cluster Cost Report
#!/bin/bash
CLUSTER_ID="${1:?Cluster ID required}"
echo "=== Cluster Cost Summary (last 30d) ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/cost-report?startDate=$(date -u -d '30 days ago' +%Y-%m-%d)&endDate=$(date -u +%Y-%m-%d)" | jq '{
totalCost: .totalCost,
computeCost: .computeCost,
spotSavings: .spotSavings,
optimizationSavings: .optimizationSavings,
savingsPercentage: .savingsPercentage
}'
echo ""
echo "=== Cost by Node Type ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/nodes" | jq -r '
.items[] |
"\(.name)\t\(.instanceType)\t\(.lifecycle)\t$\(.costPerHour | . * 100 | round / 100)/hr"
' | column -t | head -15
Spot Instance Management
#!/bin/bash
CLUSTER_ID="${1:?Cluster ID required}"
echo "=== Node Lifecycle Distribution ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/nodes" | jq '
.items | group_by(.lifecycle) | map({
lifecycle: .[0].lifecycle,
count: length,
totalCostPerHour: (map(.costPerHour) | add | . * 100 | round / 100)
})
'
echo ""
echo "=== Spot Interruption Risk ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/nodes" | jq -r '
.items[] | select(.lifecycle == "spot") |
"\(.name)\t\(.instanceType)\t\(.availabilityZone)\tAge: \(.age)"
' | column -t | head -15
Rebalancing Status
#!/bin/bash
CLUSTER_ID="${1:?Cluster ID required}"
echo "=== Rebalancing Plan ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/rebalancing-plan" | jq '{
status: .status,
nodesToAdd: (.nodesToAdd | length),
nodesToRemove: (.nodesToRemove | length),
estimatedSavings: .estimatedSavingsPercentage
}'
echo ""
echo "=== Nodes to Remove (underutilized) ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/rebalancing-plan" | jq -r '
.nodesToRemove[]? |
"\(.name)\t\(.instanceType)\tCPU: \(.cpuUtilization | . * 100 | round)%\tRAM: \(.memUtilization | . * 100 | round)%\tCost: $\(.costPerHour)/hr"
' | column -t | head -10
echo ""
echo "=== Recommended Replacements ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/rebalancing-plan" | jq -r '
.nodesToAdd[]? |
"\(.instanceType)\t\(.lifecycle)\t\(.availabilityZone)\tCost: $\(.costPerHour)/hr"
' | column -t | head -10
Workload Right-Sizing
#!/bin/bash
CLUSTER_ID="${1:?Cluster ID required}"
echo "=== Right-Sizing Recommendations ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/workload-optimization" | jq -r '
.items[] | select(.savingsAmount > 1) |
"\(.namespace)/\(.workloadName)\tCPU: \(.currentCpu)→\(.recommendedCpu)\tMem: \(.currentMemory)→\(.recommendedMemory)\tSavings: $\(.savingsAmount | . * 100 | round / 100)/mo"
' | sort -t'$' -k4 -rn | column -t | head -20
Optimization Policies
#!/bin/bash
CLUSTER_ID="${1:?Cluster ID required}"
echo "=== Autoscaler Policies ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/policies" | jq '{
enabled: .enabled,
spotEnabled: .spotInstances.enabled,
spotPercentage: .spotInstances.maxPercent,
unschedulablePods: .unschedulablePods.enabled,
nodeDownscaling: .nodeDownscaling.enabled,
cpuHeadroom: .clusterLimits.cpu.maxCores,
memoryHeadroom: .clusterLimits.memory.maxGiB
}'
echo ""
echo "=== Scaling Events (last 24h) ==="
castai_api GET "kubernetes/clusters/${CLUSTER_ID}/audit-log?fromDate=$(date -u -d '24 hours ago' +%Y-%m-%dT%H:%M:%SZ)" | jq -r '
.items[] | select(.action | contains("node")) |
"\(.createdAt[0:16])\t\(.action)\t\(.details[0:60])"
' | head -15
Safety Rules
- Review before enabling: Always review policies before enabling autoscaler or spot
- Gradual rollout: Enable optimization features one at a time (spot, downscaling, right-sizing)
- Node drain timeout: Ensure proper PDB and drain settings before enabling node removal
- Critical workloads: Mark critical workloads with anti-spot labels to keep them on on-demand nodes
Output Format
Present results as a structured report:
Managing Cast Ai Report
═══════════════════════
Resources discovered: [count]
Resource Status Key Metric Issues
──────────────────────────────────────────────
[name] [ok/warn] [value] [findings]
Summary: [total] resources | [ok] healthy | [warn] warnings | [crit] critical
Action Items: [list of prioritized findings]
Target ≤50 lines of output. Use tables for multi-resource comparisons.
Anti-Hallucination Rules
- NEVER assume resource names — always discover via CLI/API in Phase 1 before referencing in Phase 2.
- NEVER fabricate metric names or dimensions — verify against the service documentation or
--helpoutput. - NEVER mix CLI commands between service versions — confirm which version/API you are targeting.
- ALWAYS use the discovery → verify → analyze chain — every resource referenced must have been discovered first.
- ALWAYS handle empty results gracefully — an empty response is valid data, not an error to retry.
Counter-Rationalizations
| Shortcut | Counter | Why |
|---|---|---|
| "I'll skip discovery and check known resources" | Always run Phase 1 discovery first | Resource names change, new resources appear — assumed names cause errors |
| "The user only asked for a quick check" | Follow the full discovery → analysis flow | Quick checks miss critical issues; structured analysis catches silent failures |
| "Default configuration is probably fine" | Audit configuration explicitly | Defaults often leave logging, security, and optimization features disabled |
| "Metrics aren't needed for this" | Always check relevant metrics when available | API/CLI responses show current state; metrics reveal trends and intermittent issues |
| "I don't have access to that" | Try the command and report the actual error | Assumed permission failures prevent useful investigation; actual errors are informative |
Common Pitfalls
- Agent connectivity: CAST AI requires an in-cluster agent -- check agent pod health first
- Cloud permissions: Insufficient IAM permissions will silently prevent node provisioning
- PDB conflicts: Pod Disruption Budgets can block node draining during rebalancing
- Spot fallback: Without on-demand fallback, spot evictions can cause downtime
- Cost reporting delay: Cost data may lag by 1-2 hours behind actual usage
- Multi-AZ: Ensure instance type availability across all configured AZs