-
Things LLMs get wrong about AWS IAM
Agents writing AWS IAM have picked up weird habits: embellishing with Deny in identity policies, using permissions boundaries as a generic restriction, and adding sts:GetCallerIdentity in policy documents
Writing and tools
EngSecLabs is where Alex Smolen writes about security engineering and ships open-source security tools. It started as a consulting practice. Alex now works full-time elsewhere, so the site is the blog and the projects.
Still open to the occasional project or advisory engagement. Email me at alex@engseclabs.com to say hello.
Recent writing
Tools I've built
TrailTool CloudTrail for AI agents
Pre-aggregates events into entities (people, sessions, roles, services) so agents can reason over cloud activity without blowing the context window.
iam-agent-proxy Least privilege for AWS agents
Credential isolation and least privilege for AWS agents. Hands agents fake keys, re-signs each request with real credentials, and generates a least-privilege policy from observed behavior.
GraphGRC SOC 2 compliance in GitHub
Pre-written controls and policies in Markdown with semantic linking. Your compliance docs live in version control, queryable by AI, not locked in a SaaS platform.
Alex Smolen
I'm a security engineer with over a decade of experience building and breaking security-critical systems. I've worked across application security, cloud infrastructure, AI security, and compliance at startups and high-growth companies.
I write about security engineering here and on alexsmolen.com. If you're working through a hard security problem, I'd like to hear about it.