Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/aws-samples/sample-well-architected-skills-and-steering/llms.txt

Use this file to discover all available pages before exploring further.

The cost-optimization-review skill performs a focused assessment of your workload’s cost efficiency. It examines IaC configurations, scaling policies, and resource provisioning in your codebase to identify waste, right-sizing opportunities, and pricing model mismatches — with every finding tied to specific file paths and relative savings estimates.
Use cost-optimization-review when you need a dedicated cost review, spend analysis, or right-sizing assessment. For a multi-pillar review that includes cost alongside security, reliability, and other pillars, use wa-review instead. Note that cost findings that also improve sustainability are flagged explicitly — many optimizations deliver a double benefit.

​
What the Agent Analyzes

The skill runs a structured discovery across six cost domains before evaluating against WA Framework questions.
1

Compute Cost Discovery

The agent examines all compute resource configurations:
  • EC2 instance types and sizes in ASG/launch templates
  • ECS task definitions (CPU/memory allocations)
  • Lambda memory and timeout configurations
  • Fargate task sizes
  • EKS node group configurations
  • Batch compute environments
Automatically flagged as HIGH RISK:
  • Fixed-size compute (no auto-scaling) for variable workloads
  • Over-provisioned Lambda (>1024MB for simple operations)
  • x86 instance types where Graviton equivalents exist (cost + sustainability impact)
  • Large instance types where smaller instances with horizontal scaling would work
  • Dev/test environments with production-sized resources running 24/7
  • No scheduled scaling for environments with clear off-hours
2

Storage and Data Cost Discovery

The agent analyzes storage configurations:
  • S3 bucket configurations (storage class, lifecycle policies, versioning)
  • EBS volume types and sizes
  • RDS storage configurations (type, allocated size, auto-scaling)
  • DynamoDB capacity mode and provisioning
  • ElastiCache node types and cluster sizes
  • EFS configurations (throughput mode, lifecycle)
  • Backup retention policies
  • CloudWatch log group retention settings
Automatically flagged as HIGH RISK:
  • S3 buckets without lifecycle policies (accumulating indefinitely)
  • CloudWatch log groups with “never expire” retention
  • EBS volumes not using gp3 (gp2 is more expensive for the same performance)
  • Over-provisioned IOPS on EBS/RDS
  • S3 versioning enabled without lifecycle rules to expire old versions
  • Backup retention > 35 days without business justification
3

Data Transfer Cost Discovery

The agent analyzes network and data transfer patterns:
  • NAT Gateway usage (could VPC endpoints replace?)
  • Cross-region data transfer patterns
  • VPC endpoint configurations (or absence for S3/DynamoDB)
  • CloudFront distributions (or absence for static content)
  • Cross-AZ traffic patterns
  • API Gateway configurations (REST vs HTTP API pricing)
Automatically flagged as HIGH RISK:
  • S3/DynamoDB access going through NAT Gateway (VPC endpoint would be free)
  • No CloudFront for static content delivery
  • REST API Gateway where HTTP API would suffice (70% cheaper)
  • Cross-region replication without business justification
4

Pricing Model Assessment

The agent evaluates whether pricing models align with usage patterns:
  • Steady-state compute → Savings Plans or Reserved Instances opportunity
  • Variable/batch compute → Spot Instance opportunity
  • Serverless vs provisioned alignment
  • DynamoDB on-demand vs provisioned (on-demand for unpredictable, provisioned for steady)
  • Aurora Serverless v2 vs provisioned (for variable database load)
  • S3 Intelligent-Tiering for unknown access patterns
5

Environment and Lifecycle Management

The agent examines non-production environment configurations:
  • Dev/test/staging environment sizing vs production
  • Scheduled scaling or shutdown for non-production
  • Resource lifecycle policies (TTL on test resources)
  • Cost allocation tags on resources
  • AWS Budget and anomaly detection configurations
Automatically flagged as IMPROVEMENT OPPORTUNITY:
  • Non-production environments running 24/7 at production scale
  • No cost allocation tags on resources
  • No AWS Budget or Cost Anomaly Detection configured
  • No lifecycle policies on test/temporary resources

​
WA Framework Coverage: COST 1–11

After discovery, the agent evaluates your workload against all 11 Cost Optimization pillar questions.
QuestionFocus Area
COST 1Cloud financial management — cost allocation tags, budget configs, anomaly detection
COST 2Usage governance — SCPs limiting instance types, quotas, resource constraints
COST 3Cost monitoring — Budget alarms, Cost Anomaly Detection, billing alarms
COST 4Resource decommissioning — lifecycle policies, TTL configs, cleanup automation
COST 5Service selection — serverless for variable loads, provisioned for steady, Spot for batch
COST 6Resource type and size — instance types, scaling min/max, memory allocations, right-sizing
COST 7Pricing models — Savings Plans, Reserved capacity, Spot fleet configs
COST 8Data transfer costs — VPC endpoints, CloudFront distributions, regional placement
COST 9Demand management — demand shaping, throttling, buffering variable workloads
COST 10Optimization evolution — continuous improvement process, benchmark iteration
COST 11Cost culture — team cost awareness, showback/chargeback, optimization incentives

​
Output Format

The skill produces a structured cost optimization report including:
  • Cost Optimization Scorecard — 1–5 score across six domains (Compute Right-Sizing, Storage Lifecycle, Data Transfer, Pricing Models, Environment Management, Cost Visibility)
  • Critical and High Risk Findings — with domain, title, description, evidence (file:line), cost impact, recommendation, effort, and relevant AWS services
  • Medium and Low Risk Findings — in condensed format
  • Savings Summary table — current config, optimized config, relative savings %, and effort for each finding
  • Prioritized Remediation Plan — Quick Wins (< 1 week), Foundation (1–4 weeks), Strategic (1–3 months)

​
Savings Summary Format

## Savings Summary

| Category | Finding | Current Config | Optimized Config | Relative Savings | Effort |
|----------|---------|---------------|-----------------|-----------------|--------|
| Storage | EBS gp2 volumes | gp2 (3 IOPS/GB) | gp3 (3000 IOPS baseline) | ~20% per volume | Low |
| Transfer | S3 via NAT | NAT Gateway | VPC Endpoint (free) | 100% of S3 transfer costs | Low |
| Compute | No scheduled scaling | 24/7 prod-size dev env | Scale down off-hours | ~60% dev environment cost | Low |
| Logs | Never-expire log groups | Unlimited retention | 90-day retention | ~40% CloudWatch Logs cost | Low |
| API | REST API Gateway | REST (full price) | HTTP API | ~70% per-request cost | Medium |

​
How to Invoke

cost review
cost optimization review
check our spend
find cost savings opportunities

​
Cost Optimization Trade-off Awareness

Cost optimizations can introduce reliability or performance risks. The agent explicitly flags any recommendation that could degrade your availability or latency posture. For example, switching from Multi-AZ RDS to single-AZ reduces cost but introduces a SPOF — the agent will note this trade-off and recommend confirming it aligns with your availability target before making the change.
When the agent cannot determine actual resource utilization from IaC alone — for example, whether an instance is over-provisioned — it will report “Cannot Determine” and recommend using AWS Compute Optimizer or Cost Explorer to gather the utilization data needed to make a confident right-sizing decision.

​
Benchmark Results

Evaluated with Claude Opus 4.8, 16K output tokens, paired comparison (same prompt with and without skill):
BaselineWith SkillDelta
93%100%+7%
The skill brings a strong baseline model to 100% on behavioral assertions covering cost domain breadth, evidence specificity, savings estimation, and trade-off analysis.

Build docs developers (and LLMs) love

Get started for freeTalk to us