cost-optimisation

name	cost-optimisation
description	Use this skill to analyse, track, and reduce cloud infrastructure spend across Azure, AWS, and GCP. Triggers: any request to review cloud costs, identify waste or idle resources, right-size over-provisioned workloads, generate a cost report, set up budget alerts, optimise Reserved Instance or Savings Plan coverage, analyse cost per tenant, or produce recommendations for reducing monthly cloud spend.
tools	["bash","computer"]

Cloud Cost Optimisation Skill

Automate cloud spend analysis, anomaly detection, waste elimination, and right-sizing recommendations. Produce actionable cost reports by tenant, environment, and service — with one-click remediation where safe to do so.

---

Data Sources

Platform	Tool / API	Auth env var
Azure	Azure Cost Management + Advisor API	AZURE_SUBSCRIPTION_ID
AWS	Cost Explorer + Trusted Advisor	AWS_PROFILE
GCP	Cloud Billing API + Recommender API	GOOGLE_APPLICATION_CREDENTIALS
Multi	infracost (IaC cost estimation)	INFRACOST_API_KEY
Multi	Kubecost (Kubernetes cost allocation)	KUBECOST_URL

---

Core Workflows

1. Monthly Spend Summary

# Azure — total spend by resource group (last 30 days)
az consumption usage list \
  --start-date "$(date -d '-30 days' +%Y-%m-%d)" \
  --end-date "$(date +%Y-%m-%d)" \
  --query "sort_by([].{RG:resourceGroup, Cost:pretaxCost, Currency:currency}, &Cost)" \
  --output table

# AWS — top services by cost
aws ce get-cost-and-usage \
  --time-period Start=$(date -d '-30 days' +%Y-%m-%d),End=$(date +%Y-%m-%d) \
  --granularity MONTHLY \
  --metrics "UnblendedCost" \
  --group-by Type=DIMENSION,Key=SERVICE \
  --output json | jq '.ResultsByTime[0].Groups | sort_by(.Metrics.UnblendedCost.Amount | tonumber) | reverse'

2. Cost Per Tenant

# Kubecost — cost allocation by namespace (each namespace = one tenant)
curl -s "$KUBECOST_URL/model/allocation?window=30d&aggregate=namespace" | \
  jq '.data[0] | to_entries | sort_by(.value.totalCost) | reverse |
    .[] | {tenant: .key, cost_usd: .value.totalCost, cpu_cost: .value.cpuCost, memory_cost: .value.memoryCost}'

3. Idle & Orphaned Resource Detection

# Azure — unattached managed disks
az disk list --query "[?diskState=='Unattached'].{Name:name, SizeGB:diskSizeGb, SKU:sku.name}" \
  --output table

# Azure — unused public IPs
az network public-ip list \
  --query "[?ipConfiguration==null].{Name:name, RG:resourceGroup}" \
  --output table

# Azure — stopped (deallocated) VMs older than 30 days
az vm list --show-details \
  --query "[?powerState=='VM deallocated'].{Name:name, RG:resourceGroup, Size:hardwareProfile.vmSize}" \
  --output table

# Azure — unused load balancers (no backend pools)
az network lb list --query "[?backendAddressPools[0]==null].name" --output tsv

# Kubernetes — pods with no CPU/memory requests set
kubectl get pods -A -o json | \
  jq -r '.items[] | select(.spec.containers[].resources.requests == null) |
    "\(.metadata.namespace)/\(.metadata.name)"'

4. Right-Sizing Recommendations

# Azure Advisor — cost recommendations
az advisor recommendation list \
  --category Cost \
  --query "[].{Impact:impact, Resource:resourceMetadata.resourceId, Recommendation:shortDescription.solution}" \
  --output table

# AWS Compute Optimizer
aws compute-optimizer get-ec2-instance-recommendations \
  --query 'instanceRecommendations[?findingReasonCodes[0]==`CPUUnderprovisioned` ||
            findingReasonCodes[0]==`Overprovisioned`]' \
  --output json

5. Reserved Instance / Savings Plan Coverage

# AWS — RI coverage report
aws ce get-reservation-coverage \
  --time-period Start=$(date -d '-30 days' +%Y-%m-%d),End=$(date +%Y-%m-%d) \
  --granularity MONTHLY \
  --output json | jq '.CoveragesByTime[0].Total.CoverageHours.CoverageHoursPercentage'

# Target: ≥ 70% coverage for predictable workloads (dev/staging excluded)

---

Auto-Remediation Actions

Actions that can run automatically (with tagging/logging):

Action	Auto-safe	Estimated saving
Delete unattached disks (>14d)	✅	$5–50/disk/month
Release unused public IPs	✅	$3–7/IP/month
Scale down over-provisioned pods	✅	Varies
Delete orphaned snapshots (>90d)	✅	$0.05/GB/month
Stop idle dev VMs (after hours)	✅	65% of compute cost
Downsize right-size candidates	⚠️ Needs approval	20–40% per resource
Purchase Reserved Instances	⚠️ Needs approval	30–60% vs on-demand
Delete stopped VMs (>30d)	⚠️ Needs approval	100% of disk cost

---

Budget Alerting

# Azure — create budget with alert
az consumption budget create \
  --budget-name "monthly-budget-${ENV}" \
  --amount $BUDGET_AMOUNT \
  --category Cost \
  --time-grain Monthly \
  --start-date "$(date +%Y-%m-01)" \
  --end-date "$(date -d '+12 months' +%Y-%m-01)" \
  --threshold 80 \
  --contact-emails "$ALERT_EMAIL" \
  --contact-groups "$ACTION_GROUP_ID"

Alert thresholds:

• 80% of monthly budget → email warning
• 100% of monthly budget → email + Slack page
• 120% of monthly budget → P2 incident, escalate to leadership

---

FinOps Report Structure

Cloud Cost Report — [Month Year]
──────────────────────────────────

Total Spend:        $XX,XXX  (+/-X% vs last month)
Budget:             $XX,XXX  (XX% consumed)
Forecast (EOM):     $XX,XXX  (XX% of budget)

Top 5 Cost Drivers
  1. AKS Compute (prod)          $X,XXX  (XX%)
  2. Azure SQL (enterprise tier) $X,XXX  (XX%)
  ...

Cost by Tenant (Top 10)
  tenant-42 (enterprise):  $XXX/month  ($X.XX/user)
  ...

Waste Identified & Remediated
  Unattached disks deleted:    12  → saved $XXX/month
  Idle public IPs released:    8   → saved $XXX/month
  Dev VMs auto-stopped:        22  → saved $XXX/month

Recommendations (pending approval)
  Right-size 4 over-provisioned AKS node pools → save ~$X,XXX/month
  Purchase 12-month RIs for stable workloads   → save ~$X,XXX/month

RI/Savings Plan Coverage:  68%  (target 70%)

Cost Trend (6-month)
  [chart data]

---

IaC Cost Estimation (Pre-Deploy)

# Estimate cost of a Terraform plan before applying
infracost breakdown --path ./terraform \
  --var-file env/prod.tfvars \
  --format json > cost-estimate.json

infracost diff --path ./terraform \
  --compare-to cost-estimate-baseline.json

Surface cost delta in every PR that modifies infrastructure.

---

Examples

• "Show me our total cloud spend by tenant for the last 30 days"
• "Find all idle and orphaned resources across our Azure subscriptions"
• "Generate the monthly FinOps report with recommendations"
• "What is the cost difference if we right-size the AKS node pools?"
• "Set up budget alerts at 80% and 100% for the prod subscription"
• "What's our Reserved Instance coverage and where should we buy more?"

---

Output Format

{
  "period": { "start": "string", "end": "string" },
  "total_spend_usd": 0.0,
  "budget_usd": 0.0,
  "budget_consumed_pct": 0.0,
  "waste_identified_usd": 0.0,
  "waste_remediated_usd": 0.0,
  "top_tenants": [],
  "recommendations": [],
  "ri_coverage_pct": 0.0,
  "report_url": "string"
}