Purpose: Standardize how stakpak designs and optimizes AWS architectures. Scope: Greenfield: design new infrastructure. Brownfield: analyze existing architectures and propose improvements.
Purpose: Standardize how the agent designs and optimizes AWS architectures.
Scope:
Reference Frameworks:
Match requirements to AWS services (compute, storage, networking, database).
Consider Serverless-first designs when applicable:
Compute → Lambda, Step Functions, Fargate
API → API Gateway + AppSync
Storage → S3, DynamoDB
Messaging → SNS, SQS, EventBridge
Security → IAM, Cognito, WAF, KMS
5 Pillars Checklist
Operational Excellence: monitoring, IaC, automation.
Security: IAM least privilege, encryption, threat detection.
Reliability: HA, backup/restore, fault isolation.
Performance Efficiency: caching, scaling, right-sizing.
Cost Optimization: Spot, RIs, lifecycle rules, serverless.
Serverless Lens Focus:
Minimize undifferentiated ops.
Event-driven orchestration (Step Functions/EventBridge).
Use managed data stores (DynamoDB, Aurora Serverless).
Secure with IAM boundaries, managed identity (Cognito).
Draft architecture diagram.
For existing → generate recommendations table: Pillar, Current Gap, Recommendation, Expected Impact
Risks & mitigations.
Cost estimates (before/after).
Load test strategy
Write everything into Markdown architecture file.
Include: Overview, Requirements, Architecture, Diagrams, Well-Architected Review, Optimizations, Risks, Costs.
Enforce least privilege IAM policies.
Prefer IAM roles over static keys.
Use ABAC or RBAC (tags, groups, accounts) for scalable access control.
Require MFA for privileged accounts.
Use AWS SSO / IAM Identity Center for central identity management.
Encrypt all data at rest (S3, EBS, RDS, DynamoDB, etc.) with KMS CMKs.
Encrypt all data in transit (TLS 1.2+).
Enable S3 Block Public Access and least privilege bucket policies.
Use Secrets Manager / Parameter Store — no hardcoded credentials.
Use VPC with private subnets for workloads.
Restrict inbound/outbound traffic with Security Groups and NACLs.
Use VPC Endpoints for private service access (no public internet).
Add WAF/Shield for public-facing endpoints.
Prefer ALB/NLB with TLS termination over exposing EC2 directly.
Enable CloudTrail in all regions and send logs to a centralized S3 bucket.
Enable Config Rules for compliance enforcement.
Integrate GuardDuty, Security Hub, Inspector for threat detection.
Centralize logs (CloudWatch Logs / OpenSearch) and set retention policies.
Use CloudWatch alarms for anomalies, cost spikes, security events.
Apply multi-AZ deployments for critical data stores.
Enforce automated backups with retention policies.
Test disaster recovery scenarios (RTO/RPO compliance).
Use infrastructure as code (Terraform/CDK/CloudFormation) to rebuild environments securely.
Apply service control policies (SCPs) with AWS Organizations.
Enforce tagging standards for resources (cost, owner, env).
Align with compliance frameworks (ISO, SOC2, HIPAA, GDPR) when required.
Use Trusted Advisor and Well-Architected Tool for regular reviews.
NOTE: Dont implement anything until you generate the report and ask for my permission