AI Agents
11 items across 2 sections
The Verdict, Four Months Later
Part 16 of 16 in the AWS Security Agent: From Zero to Hero series. Closing the series: the numbers this whole project is built on, the honest verdict distilled to its essentials, what AWS has shipped since my hands-on testing, and an evidence-based look at how AI pentesting stacks up against traditional consultancy today.
Attacking the Agent Itself
Part 14 of 16 in the AWS Security Agent: From Zero to Hero series. What AWS itself documents about AWS Security Agent as its own attack surface: the guardrails it publishes, how it limits its own blast radius, and how domain ownership and data handling are enforced. Sourced entirely from AWS's public documentation, not hands-on testing.
We Tried to Trick the Agent
Part 13 of 16 in the AWS Security Agent: From Zero to Hero series. Adversarial evasion experiments against AWS Security Agent: code obfuscation, prompt injection in design documents and pull requests, fake sanitizers, and misleading comments. Almost every obfuscated vulnerability was still detected (18 of 19), and every injection attempt in these runs was ignored.
What It Catches and What It Misses
Part 12 of 16 in the AWS Security Agent: From Zero to Hero series. An honest, data-driven account of where AWS Security Agent falls short, with a practical fix for each weakness, plus why results vary from run to run and why a single run undercounts.
How We Measured Detection Rate
Part 11 of 16 in the AWS Security Agent: From Zero to Hero series. The full measurement methodology behind this series. A purpose-built application with 39 intentional vulnerabilities, a JSON answer key, false-positive controls, and a scoring scheme you can reproduce.
Inside the Engine: What the Logs Reveal
Part 10 of 16 in the AWS Security Agent: From Zero to Hero series. A data-driven walk through the CloudWatch logs of two real penetration tests: the 14 tools the agent uses, the thousands of reasoning blocks and tool calls, and how the observed behavior maps onto AWS's published multi-agent stages.
Your First Penetration Test
Part 7 of 16 in the AWS Security Agent: From Zero to Hero series. Creating an on-demand penetration test in AWS Security Agent and watching each agent component work, from the TLS scanner to the crawler that discovered only 7 of the 48 scored endpoints on its own.
Code Reviews and GitHub
Part 5 of 16 in the AWS Security Agent: From Zero to Hero series. Connecting GitHub to AWS Security Agent, the analysis-type trap that produces silent reviews, what the agent caught in a deliberately vulnerable Flask app (9 of 10), and the new full-repository review that entered preview on May 12, 2026.
Design Reviews
Part 4 of 16 in the AWS Security Agent: From Zero to Hero series. Running a design review on an intentionally insecure architecture document, reading the four compliance statuses, and a finding the AWS docs do not mention: file format changes the results.
Setting Up From Scratch
Part 3 of 16 in the AWS Security Agent: From Zero to Hero series. A field-tested walkthrough of setting up AWS Security Agent: choosing an access method, creating an Agent Space, the two IAM roles, the service principal, and domain verification, including the one-click verification gotcha the docs do not warn you about.
Need Help with AWS Security?
Our AWS security experts can help you implement best practices across all these topics.
Contact Us