Automating agentic AI Well-Architected reviews with the AWS Agentic AI Lens

This article describes how Smarsh automated AWS Well-Architected reviews for agentic AI workloads using a custom Claude skill that evaluates 150 best practices and generates compliance workbooks.

• Built a Claude skill packaging the Agentic AI Lens as executable data with structured JSON best practices, assessment rubrics, and Python report generation
• Skill gathers evidence from GitHub repositories including architecture docs, infrastructure-as-code, CI/CD workflows, and test suites
• Assessed workload against all 150 lens items in hours instead of multi-day manual reviews, with five-value compliance rubric and mandatory evidence citations
• Validated skill using synthetic fixture repositories before trusting output on production workloads
• First review identified strengths like scoped IAM roles and SigV4-signed calls, plus gaps like disabled gateway authentication and missing prompt-injection screening
• Findings enabled focused AWS Solutions Architect engagements mapped to specific best-practice gaps rather than broad architecture reviews

Automating Well-Architected reviews as repeatable organizational capability accelerates agentic AI governance and time-to-market for regulated customers.