Cloud Cost Optimizer

SCOPE

Core Commands

analyze-spend

kilo cloud-cost-optimizer analyze-spend \
  --platform <aws|azure|gcp|all> \
  --days <integer> \
  --output <json|csv|html> \
  --threshold <percentage>

Real example: kilo cloud-cost-optimizer analyze-spend --platform all --days 30 --output html --threshold 15

idle-detector

kilo cloud-cost-optimizer idle-detector \
  --resource-type <ec2|ebs|elastic-ip|load-balancer|azure-vm|gcp-disk> \
  --idle-threshold-hours <integer> \
  --region <region-name>

Real example: kilo cloud-cost-optimizer idle-detector --resource-type ec2 --idle-threshold-hours 72 --region us-east-1

reservation-advisor

kilo cloud-cost-optimizer reservation-advisor \
  --platform <aws|azure|gcp> \
  --term <1-year|3-year> \
  --payment-option <all-upfront|partial-upfront|no-upfront> \
  --confidence <0.7-0.99>

Real example: kilo cloud-cost-optimizer reservation-advisor --platform aws --term 3-year --payment-option partial-upfront --confidence 0.85

right-size-recommendations

kilo cloud-cost-optimizer right-size-recommendations \
  --platform <aws|azure|gcp> \
  --metric <cpu|memory|network|all> \
  --utilization-threshold <0.1-0.5> \
  --instance-family <specific-family>

Real example: kilo cloud-cost-optimizer right-size-recommendations --platform aws --metric cpu --utilization-threshold 0.2

storage-analyzer

kilo cloud-cost-optimizer storage-analyzer \
  --service <s3|azure-blob|gcp-storage> \
  --bucket <bucket-name> \
  --days-access <integer> \
  --transition-to <standard|ia|archive>

Real example: kilo cloud-cost-optimizer storage-analyzer --service s3 --bucket my-app-logs --days-access 90 --transition-to ia

tag-compliance

kilo cloud-cost-optimizer tag-compliance \
  --required-tags <comma-separated> \
  --platform <aws|azure|gcp> \
  --resource-types <comma-separated>

Real example: kilo cloud-cost-optimizer tag-compliance --required-tags Environment,Owner,Project --platform aws --resource-types ec2,s3,lambda

forecast-spend

kilo cloud-cost-optimizer forecast-spend \
  --platform <aws|azure|gcp> \
  --days <30|60|90> \
  --include-commitments <true|false>

Real example: kilo cloud-cost-optimizer forecast-spend --platform aws --days 30 --include-commitments true

WORK PROCESS

1. Authentication & Configuration

# Set environment variables (loaded from ~/.openclaw/config/ext/custom-overrides.json)
export AWS_ACCESS_KEY_ID="AKIAIOSFODNN7EXAMPLE"
export AWS_SECRET_ACCESS_KEY="wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"
export AWS_DEFAULT_REGION="us-east-1"
export AZURE_SUBSCRIPTION_ID="12345678-1234-1234-1234-123456789012"
export GCP_PROJECT_ID="my-project-123456"
export COST_THRESHOLD_DAYS=30

2. Run Analysis (Real command sequence)

Multi-cloud spending analysis:

kilo cloud-cost-optimizer analyze-spend \
  --platform all \
  --days 30 \
  --output json \
  --threshold 15 > cost-analysis-$(date +%Y%m%d).json

Idle EC2 detection:

kilo cloud-cost-optimizer idle-detector \
  --resource-type ec2 \
  --idle-threshold-hours 168 \
  --region us-east-1 \
  --output csv > idle-ec2-us-east-1-$(date +%Y%m%d).csv

Reservation recommendations:

kilo cloud-cost-optimizer reservation-advisor \
  --platform aws \
  --term 3-year \
  --payment-option partial-upfront \
  --confidence 0.85 > reservations-recommendations-$(date +%Y%m%d).txt

3. Implementation Workflow

After receiving recommendations:

1. Review high-impact items (> $100/month potential savings)
2. Test non-destructive actions in staging environments first
3. Schedule changes during maintenance windows
4. Document every change with JIRA ticket reference
5. Monitor for 7 days post-implementation

4. Verification Steps

Before changes:

# Capture baseline
aws cloudwatch get-metric-statistics --namespace AWS/EC2 --metric-name CPUUtilization --start-time $(date -d '7 days ago' -Iseconds) --end-time $(date -Iseconds) --period 86400 --statistics Average
az monitor metrics list --resource <resource-id> --metric "Percentage CPU" --interval PT1H
gcloud monitoring time-series list --filter='metric.type="compute.googleapis.com/instance/cpu/utilization"' --project=$GCP_PROJECT_ID

After changes:

# Wait 24-48 hours, then compare costs
kilo cloud-cost-optimizer analyze-spend --platform aws --days 2 --output json | jq '.total_savings_expected'

GOLDEN RULES

1. Never delete resources without 7-day idle confirmation - Use --dry-run flag first, verify metrics manually, implement deletion schedule with 48-hour grace period
2. Never modify production resources during business hours - Restrict changes to 02:00-06:00 local timezone, require change management ticket approval
3. Always preserve audit trails - Tag every optimization action with CostOptimized=true and OptimizationDate=YYYY-MM-DD, create AWS Config rules to prevent deletion of tagged resources
4. Reservations require capacity reservation - Before purchasing 3-year reservations, verify aws ec2 describe-reserved-instances-offerings has capacity in target region, Azure az vm reservation list shows availability
5. Right-sizing requires performance testing - Load test target instance for 1 hour minimum in staging, compare p99 latency and throughput before production cutover
6. Storage transitions are irreversible - Archive to Glacier/Archive tiers has 90-day minimum retention, calculate egress costs before transition
7. Budget alerts must be set - After optimization, create/update AWS Budgets with 80%, 100%, 120% thresholds, integrate with Slack/PagerDuty webhooks
8. Cross-cloud coordination required - If optimizing AWS and Azure simultaneously, ensure application can handle mixed instance types and region failovers
9. Reserved instance flexibility check - Verify aws ec2 describe-reserved-instances --filters Name=scope,Values=Availability Zone matches actual deployment zones before purchase
10. Compliance validation mandatory - After tag cleanup, run aws resourcegroupstaggingapi get-resources to ensure no resources orphaned without required tags

EXAMPLES

Example 1: Detect unattached EBS volumes

$ kilo cloud-cost-optimizer idle-detector --resource-type ebs --idle-threshold-hours 24 --region us-west-2
Found 47 unattached EBS volumes in us-west-2:
  - vol-0a1b2c3d4e5f6g7h8 (100 GiB, gp3) - $10.30/month, unattached for 15 days
  - vol-1a2b3c4d5e6f7g8h9 (500 GiB, io2) - $75.00/month, unattached for 89 days
Total potential savings: $2,467/month

# Implementation command (review first):
aws ec2 delete-volume --volume-id vol-0a1b2c3d4e5f6g7h8

Example 2: Reserved instance purchase recommendation

$ kilo cloud-cost-optimizer reservation-advisor --platform aws --term 1-year --payment-option all-upfront
Based on 90-day usage analysis:

Instance: m5.large (2 vCPU, 8 GiB)
  Current on-demand spend: $1,872/year
  1-year All Upfront RI: $1,125 (40% savings)
  Savings: $747/year
  Risk: Low (95% utilization consistency)
  
Purchase command:
aws ec2 purchase-reserved-instances-offering \
  --reserved-instances-offering-id 1234abcd-12ab-34cd-56ef-1234567890ab \
  --instance-count 3 \
  --dry-run

Example 3: EC2 right-sizing analysis

$ kilo cloud-cost-optimizer right-size-recommendations --platform aws --metric cpu --utilization-threshold 0.15
Instance: i-0a1b2c3d4e5f6g7h8 (m5.xlarge)
  Avg CPU: 8% (max 12%)
  Avg Memory: 4.2 GiB / 16 GiB (26%)
  Current cost: $196/month
  Recommended: m5.large ($98/month)
  Savings: $98/month
  Confidence: 97% (based on 30-day metrics)
  
Downgrade command:
aws ec2 modify-instance-attribute --instance i-0a1b2c3d4e5f6g7h8 --instance-type m5.large

Example 4: S3 storage class transition

$ kilo cloud-cost-optimizer storage-analyzer --service s3 --bucket prod-backups --days-access 180
Bucket: prod-backups (2.4 TB total)
  Objects not accessed in 180+ days: 1.8 TB (75%)
  Current cost: $55/month (Standard)
  After transition to Intelligent-Tiering: $33/month
  Savings: $22/month ($264/year)
  
Batch transition command (dry-run first):
aws s3 cp s3://prod-backups s3://prod-backups --storage-class INTELLIGENT_TIERING --recursive --dryrun

Example 5: Multi-cloud cost anomaly detection

$ kilo cloud-cost-optimizer analyze-spend --platform all --days 7 --threshold 25
AWS: EC2 costs increased 34% ($2,100 → $2,814)
  Root cause: New m5.8xlarge instance launched 2024-01-15
  Action: Verify instance necessity, right-size or terminate

Azure: Storage costs increased 67% ($450 → $752)
  Root cause: Premium SSD provisioned for dev environment
  Action: Convert to Standard SSD

GCP: Compute Engine costs stable, BigQuery costs increased 22%
  Root cause: Unoptimized queries in analytics pipeline
  Action: Implement partitioned tables, add WHERE clauses
  
Total unexpected spend: $1,216

ROLLBACK COMMANDS

If idle resource deletion was incorrect:

# AWS - Restore deleted EBS volume (if snapshot exists)
aws ec2 create-volume --snapshot-id snap-0a1b2c3d4e5f6g7h8 --availability-zone us-east-1a

# Azure - Recreate deleted VM from latest snapshot
az vm create --resource-group my-group --name my-vm --attach-os-disk my-vm-osdisk --os-type linux

# GCP - Restore deleted persistent disk
gcloud compute disks create my-disk --source-snapshot my-snapshot --zone us-central1-a

If right-sizing caused performance issues:

# AWS - Revert instance type
aws ec2 modify-instance-attribute --instance i-0a1b2c3d4e5f6g7h8 --instance-type m5.xlarge

# Azure - Change VM size back
az vm resize --resource-group my-group --name my-vm --size Standard_D4s_v3

# GCP - Update machine type
gcloud compute instances set-machine-type my-instance --machine-type n1-standard-4 --zone us-central1-a

If reserved instance purchase was wrong:

# AWS - Cancel RI within 5 years (prorated refund)
aws ec2 modify-reserved-instances --reserved-instances-ids 1234abcd-5678-... --instance-count 1

# Azure - Exchange RI (if within 5 years)
az vm reservation exchange --reservation-order-id 11111111-1111-1111-1111-111111111111 --reservation-id 22222222-2222-2222-2222-222222222222

# GCP - Cancel commitment (minimum 1 year)
gcloud compute commitments remove-iam-policy-binding my-commitment --member='user:[email protected]' --role='roles/compute.viewer'

If storage transition errors occurred:

# Restore S3 objects from Standard-IA to Standard (within 30 days)
aws s3 copy s3://my-bucket/key s3://my-bucket/key --storage-class STANDARD --metadata-directive COPY

# Restore Azure blob from Archive (requires rehydration, 1-15 hours)
az storage blob rehydrate --container-name mycontainer --name myblob --tier Hot

# Restore GCP Nearline/Coldline to Standard (7-day minimum retention)
gsutil rewrite -s standard gs://my-bucket/gs://my-bucket/**

Complete rollback (undo all optimizations in last 24 hours):

# Generate rollback script from audit log
kilo cloud-cost-optimizer generate-rollback --since "24 hours ago" --platform all --output rollback-$(date +%Y%m%d-%H%M%S).sh

# Execute rollback with confirmation
bash rollback-$(date +%Y%m%d-%H%M%S).sh --dry-run
bash rollback-$(date +%Y%m%d-%H%M%S).sh --execute

Restore from backup before optimization:

# If pre-optimization backup created
kilo cloud-cost-optimizer restore-state --backup-file /backups/cloud-state-$(date -d 'yesterday' +%Y%m%d).json --region all

TROUBLESHOOTING

Issue: "InvalidClientTokenId" from AWS CLI

• Check: aws sts get-caller-identity
• Fix: Verify AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY are correct, not expired, have required IAM permissions (ce:GetCostAndUsage, ec2:Describe*, s3:ListAllMyBuckets)

Issue: "SubscriptionNotRegistered" from Azure CLI

• Check: az account show
• Fix: Ensure AZURE_SUBSCRIPTION_ID matches active subscription, run az account set --subscription $AZURE_SUBSCRIPTION_ID, register required providers: az provider register --namespace Microsoft.CostManagement

Issue: "Permission denied" on GCP

• Check: gcloud auth list
• Fix: Authenticate with gcloud auth activate-service-account --key-file=$GOOGLE_APPLICATION_CREDENTIALS, ensure service account has roles: roles/billing.viewer, roles/compute.viewer, roles/storage.objectViewer

Issue: No data returned for analysis

• Check: Billing export to S3/Storage bucket is configured, Cost and Usage Report (CUR) is active, data is at least 24 hours old
• Fix: Enable CUR in AWS Billing console, wait 24-48 hours for initial delivery, verify bucket permissions

Issue: High false-positive rate on idle detection

• Check: Verify --idle-threshold-hours matches business cycle (weekends vs weekdays), validate CloudWatch metrics have sufficient granularity
• Fix: Use --metric-average-period 300 for 5-minute granularity, cross-reference with application logs before termination

VERIFICATION CHECKLIST

After any optimization cycle, verify:

• All deleted resources confirmed unattached >72 hours
• Right-sized instances passed load test in staging
• Reserved instance coverage >70% of steady-state workload
• Storage transitions have lifecycle policies for cleanup
• Budget alerts configured at 80%, 100%, 120% thresholds
• Weekly cost report scheduled via cron: 0 6 * * 1 kilo cloud-cost-optimizer analyze-spend --platform all --days 7 --output html > /reports/cost-$(date +\%Y\%m\%d).html
• Slack notifications enabled with webhook URL
• All changes documented in change management system
• Rollback scripts tested in sandbox environment

DEPENDENCY INSTALLATION

# AWS CLI v2
curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip"
unzip awscliv2.zip
sudo ./aws/install

# Azure CLI
curl -sL https://aka.ms/InstallAzureCLIDeb | sudo bash

# GCloud SDK
curl -O https://dl.google.com/dl/cloudsdk/channels/rapid/downloads/google-cloud-cli-467.0.0-linux-x86_64.tar.gz
tar -xf google-cloud-cli-*.tar.gz
./google-cloud-sdk/install.sh

# Python dependencies
pip3 install --user pandas matplotlib

SECURITY REQUIREMENTS

• Use IAM roles with least privilege, never root credentials
• Rotate API keys every 90 days maximum
• Store credentials in AWS Secrets Manager / Azure Key Vault / GCP Secret Manager, not in shell history
• Enable MFA deletion on S3 buckets containing cost exports
• Encrypt all cost reports at rest with KMS keys
• Restrict skill execution to dedicated cost optimization service account with read-only billing permissions except for specific write actions

OUTPUT FORMATS

JSON:

{
  "timestamp": "2024-01-20T14:30:00Z",
  "platform": "aws",
  "total_monthly_savings": 4523.67,
  "recommendations": [
    {
      "type": "idle_resource",
      "resource_id": "vol-0a1b2c3d4e5f6g7h8",
      "region": "us-east-1",
      "monthly_cost": 10.30,
      "action": "delete",
      "confidence": "high",
      "justification": "Unattached for 15 days, no attachment history"
    }
  ]
}

CSV:

type,resource_id,region,monthly_cost,action,confidence
idle_resource,vol-0a1b2c3d4e5f6g7h8,us-east-1,10.30,delete,high
rightsizing,i-0a1b2c3d4e5f6g7h8,us-west-2,196.00,m5.large,medium
reservation,m5.large,us-east-1,747.00,1-year-all-upfront,high

HTML (with embedded charts): Interactive dashboard with savings by category, time series trends, implementation progress tracker

AUTOMATION INTEGRATION

GitHub Actions workflow: