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 AWS Well-Architected Framework defines six pillars and 57 standard questions that apply to every workload. Lenses extend that foundation with workload-specific best practices — additional questions, best practices, and risks that only matter when your architecture includes specific technologies or operates in a regulated industry. The wa-review skill ships with 27 lens packs, each sourced directly from the AWS public documentation and stored as flat reference files alongside the 57 core question files.

​
What lenses are

A lens is a curated set of additional best practices that the reviewing agent loads on top of the standard 57-question framework when it detects — or you explicitly request — a specific workload type. Each lens is stored as a directory under skills/wa-review/references/lenses/ and contains one reference file per lens question or topic area. The agent evaluates lens questions using the same evidence-gathering process as core framework questions: reading your code, IaC, and architecture documentation to determine compliance.

​
All 27 lenses

LensFilesSizeLoaded when
Framework questions572.2 MBFull review — one file per question evaluated
Serverless Lens6120 KBWorkload uses Lambda/API Gateway/Step Functions
Generative AI Lens29368 KBLLM, RAG, or fine-tuning workloads
Agentic AI Lens411.2 MBAI agent workloads
Responsible AI Lens28780 KBAI governance and fairness requirements
Hybrid Networking Lens30480 KBDirect Connect, VPN, Transit Gateway
Migration Lens676 KBMigration planning
DevOps Guidance Lens196820 KBCI/CD, automated governance, dev lifecycle, observability
Machine Learning Lens35852 KBML lifecycle (MLOPS), training/deployment, data engineering, responsible ML
Data Analytics Lens6180 KBData pipelines, governance, catalogs, lineage, analytics perf & cost
Games Industry Lens32316 KBGame backends, real-time multiplayer, player data, live ops
SaaS Lens6112 KBMulti-tenancy, tenant isolation, onboarding, metering, tiering
Financial Services Lens79432 KBFSI compliance, data residency, resilience, auditability
Life Sciences Lens56468 KBGxP, validated systems, clinical/research data, compliance
End User Computing Lens69372 KBVirtual desktops/apps, streaming, identity, endpoint delivery
Supply Chain Lens51244 KBSupply chain data, integration, traceability, resilience
Video Streaming & Advertising Lens43296 KBVideo pipelines, streaming delivery, ad tech, monetization
Telco Lens34272 KBTelecom workloads, 5G/edge, OSS/BSS, carrier-grade reliability
SAP Lens6380 KBSAP on AWS, S/4HANA, HANA databases, SAP landscape resilience
Modern Industrial Data Technology Lens34300 KBIndustrial data platforms, OT/IT convergence, manufacturing analytics
Microsoft Workloads Lens23368 KBWindows Server, SQL Server, Active Directory, .NET on AWS
Connected Mobility Lens6284 KBConnected vehicles, telematics, fleet data, automotive platforms
Healthcare Industry Lens692 KBHIPAA, clinical data, interoperability, patient privacy
Container Build Lens676 KBContainer image builds, supply chain security, registries, CI/CD
High Performance Computing Lens23104 KBHPC clusters, parallel workloads, scheduling, low-latency networking
Streaming Media Lens696 KBMedia streaming, live/VOD delivery, encoding, content workflows
IoT Lens59369 KBIoT devices, telemetry, edge computing, fleet provisioning, OTA updates
Government Lens646 KBPublic sector, privacy-by-design, compliance, real-time security

​
Auto-detection

The reviewing agent infers which lenses to apply by examining your workload before starting the review. Common heuristics include:
  • References to Bedrock, SageMaker, or third-party LLM APIs → Generative AI Lens
  • Agentic orchestration patterns (tool use, multi-step reasoning, LangChain, Strands) → Agentic AI Lens
  • Fairness, bias, explainability, or model governance requirements → Responsible AI Lens
  • SageMaker training pipelines, feature stores, or MLflow → Machine Learning Lens
  • Lambda functions, API Gateway, or Step Functions → Serverless Lens
  • ECS/EKS container image build pipelines → Container Build Lens
  • Slurm, AWS ParallelCluster, or HPC job schedulers → High Performance Computing Lens
  • Direct Connect, Site-to-Site VPN, or Transit Gateway → Hybrid Networking Lens
  • IoT Core, Greengrass, or fleet provisioning → IoT Lens
  • HIPAA or clinical data handling → Healthcare Industry Lens
  • GxP, clinical trials, or validated systems → Life Sciences Lens
  • PCI-DSS, financial data residency, or audit logging → Financial Services Lens
  • SAP workloads or S/4HANA → SAP Lens
  • Connected vehicle or telematics data → Connected Mobility Lens
  • Government or public sector compliance → Government Lens
  • CI/CD pipelines, DevOps toolchain, or GitOps → DevOps Guidance Lens
  • Multi-tenant SaaS with tenant isolation → SaaS Lens
  • Media encoding, live streaming, or VOD delivery → Streaming Media Lens or Video Streaming & Advertising Lens
  • Manufacturing, OT/IT convergence, or industrial IoT → Modern Industrial Data Technology Lens
  • Windows Server, SQL Server, Active Directory, or .NET → Microsoft Workloads Lens
  • Supply chain integration or traceability → Supply Chain Lens
  • Game backends, matchmaking, or live ops → Games Industry Lens
  • Telco infrastructure, 5G, or OSS/BSS → Telco Lens
  • AppStream, WorkSpaces, or end-user virtual desktops → End User Computing Lens
  • Data lakes, Glue, Athena, or data catalogs → Data Analytics Lens

​
Manually requesting a lens

You can always ask for a lens explicitly, even if the agent did not auto-detect it: 
Evaluate our architecture against the Serverless Lens.

Include the Financial Services Lens in this review — we process payments.

Run the Agentic AI Lens only, skip the core 57 questions.
Lens-only reviews are a cost-effective way to check a specific domain concern without reloading all 57 core question files. Combine with pillar-scoping for even tighter token budgets — for example: “run a security-only review with the Generative AI Lens.”

​
Where the files live

All lens reference data is committed to the repository and does not require any AWS credentials or network access at review time: 
skills/wa-review/references/
  questions/          57 framework question files (OPS01.md … SUS06.md)
  lenses/
    serverless-applications/
    generative-ai/
    agentic-ai/
    responsible-ai/
    devops-guidance/
    machine-learning/
    … (27 lens directories total)
To refresh lens content when AWS documentation updates, see Regenerating Reference Files.

Build docs developers (and LLMs) love

Get started for freeTalk to us