AWS FinOps
Review AWS spend: find the biggest drivers, rightsizing and commitment opportunities, tagging gaps and quick wins, with savings estimates.
1 skillMITCódigo-fonte
Instalar
/plugin install aws-finops@pieroazambujaAinda não adicionou o marketplace? Rode antes /plugin marketplace add piazambuja/piazambuja.github.io.
npx skills add piazambuja/piazambuja.github.io --skill aws-cost-review -a codexInstala a skill deste plugin no projeto. Acrescente -g para todos os projetos.
npx skills add piazambuja/piazambuja.github.io --skill aws-cost-review -a gemini-cliInstala a skill deste plugin no projeto. Acrescente -g para todos os projetos.
npx skills add piazambuja/piazambuja.github.io --skill aws-cost-review -a github-copilotInstala a skill deste plugin no projeto. Acrescente -g para todos os projetos.
npx skills add piazambuja/piazambuja.github.io --skill aws-cost-review -a cursorInstala a skill deste plugin no projeto. Acrescente -g para todos os projetos.
npx skills add piazambuja/piazambuja.github.io --skill aws-cost-review -a kiro-cliInstala a skill deste plugin no projeto. Acrescente -g para todos os projetos.
Skills
O agente carrega uma skill quando o pedido combina com a descrição dela. No Claude Code, você também pode chamar direto com /aws-finops:<skill>. Abra cada uma para ler as instruções completas.
aws-cost-review Review AWS costs from Cost Explorer/CUR exports, the AWS CLI or Terraform, and produce a prioritised savings plan with estimates. Use when the user asks why the AWS bill is high, how to reduce cloud spend, or to review costs for an account.
AWS cost review
Read-only. Never stop, delete or modify resources; propose changes and let the user apply them.
1. Get the data
In order of preference:
- A CUR / Cost Explorer CSV the user provides.
- AWS CLI (read-only credentials):
aws ce get-cost-and-usage --time-period Start=<YYYY-MM-01>,End=<YYYY-MM-01> \
--granularity MONTHLY --metrics UnblendedCost \
--group-by Type=DIMENSION,Key=SERVICE
aws ce get-rightsizing-recommendation --service AmazonEC2
aws ce get-savings-plans-purchase-recommendation --savings-plans-type COMPUTE_SP \
--term-in-years ONE_YEAR --payment-option NO_UPFRONT --lookback-period-in-days SIXTY_DAYS- Terraform/IaC to estimate what is provisioned.
Look at three months to separate trend from spikes.
2. Find the drivers
Rank services by cost and by month-over-month growth. The usual suspects:
- EC2/EKS: low utilisation instances, on-demand where Spot fits (stateless, batch, CI), old generations (move to Graviton/current gen), oversized node groups, idle dev clusters running 24/7.
- Data transfer: NAT Gateway processing (add VPC endpoints for S3/DynamoDB/ECR), cross-AZ traffic between chatty services, egress to the internet.
- EBS: unattached volumes, old snapshots, gp2 → gp3.
- S3: no lifecycle rules, versioned buckets keeping old versions forever, wrong storage class.
- RDS/Aurora: oversized, Multi-AZ on non-prod, idle replicas, storage I/O on Aurora standard vs I/O-Optimized.
- CloudWatch: high-cardinality custom metrics, verbose logs without retention.
- Bedrock/AI: model choice per use case, prompt caching, batch inference for offline work.
3. Commitments
Only after rightsizing: Compute Savings Plans for the steady baseline, sized to the minimum stable usage, not the average.
4. Allocation
Check tag coverage (team, service, env). Without it, nobody owns the cost. Recommend a tag policy and cost allocation tags.
Output
A table ordered by estimated monthly savings: finding, evidence, action, estimated savings, effort (low/med/high), risk. Then quick wins (this week), medium-term work, and governance (tags, budgets, anomaly detection). State assumptions behind each estimate.