AWS Builder Center

Weekend Creative Agent Challenge: ShadowBreach — The Autonomous Daily Cyber Incident Agent

An always-on autonomous agent powered by Amazon Bedrock (Nova Lite) that synthesizes realistic, daily cloud security incident response challenges with mock audit logs and interactive tactical dilemmas. Built for the AWS Builder Weekend Creative Agent Challenge.

Weekend Creative Agent Challenge: ShadowBreach — The Autonomous Daily Cyber Incident Agent

Vision and What It Does

In cloud security and DevSecOps, real-world attack vectors evolve faster than traditional curriculum schedules. Security engineers and cloud practitioners often find themselves reading dry, CVE-formatted advisories with little contextual urgency.
The philosophy behind this challenge was simple: "The best tool is the one you never have to open."
ShadowBreach is an always-on, autonomous agent that turns threat intelligence into an engaging, daily interactive incident response simulation (RPG format). While the engineer sleeps, the agent wakes up, analyzes real-world vulnerability patterns, contextualizes the blast radius into a simulated AWS environment, generates realistic CloudTrail/VPC audit logs, and crafts a tactical crisis dilemma.
When the engineer starts their morning, an interactive terminal challenge is waiting for them—complete with actionable choices, real-time feedback, and architectural least-privilege takeaways. No human intervention or prompt engineering is needed at runtime; the agent works completely autonomously.



How I Built It

The development process focused on creating a resilient, hands-off pipeline that guarantees high-fidelity outputs every day without manual supervision:
  1. Autonomous Scheduling & Data Ingestion: An Amazon EventBridge cron trigger invokes an AWS Lambda function on a daily schedule.
  1. Context Enrichment via Amazon Bedrock: The payload is processed using Amazon Bedrock with the  amazon.nova-lite-v1:0  model via the Bedrock Converse API. A structured system prompt enforces strict JSON output, producing a scenario title, threat level, tactical dilemma, synthetic CloudTrail log snippet, three distinct tactical choices, and a security takeaway.
  1. Automated Publishing Pipeline: The generated scenario updates the front-facing dashboard, ensuring the terminal interface displays a fresh incident challenge every day.
  1. Key Challenges: Ensuring consistent JSON formatting from generative models without markdown noise. Implementing deterministic inference parameters and strict validation handling inside the Lambda handler resolved formatting inconsistencies.



AWS Services Used and Architecture Overview

[ Amazon EventBridge (Cron Trigger) ]
│
▼
[ AWS Lambda (Python Logic) ]
│
├──► [ Amazon Bedrock (Nova Lite Model) ]
│
▼
[ GitHub Pages / S3 (Terminal Interface & JSON Data) ]
│
▼
[ End-User / Browser (Interactive Roleplay Drill) ]
  • Amazon Bedrock (Nova Lite): Powers the core creative storytelling and technical synthesis, turning raw vulnerability concepts into educational, story-driven cyber drills.
  • AWS Lambda: Serverless execution environment orchestrating data processing and Bedrock model invocation.
  • Amazon EventBridge: Provides scheduled cron triggers that keep the agent completely autonomous.
  • GitHub Pages: Delivers a lightweight, responsive hacker-terminal web experience.



What I Learned

Building ShadowBreach highlighted how autonomous agents differ from standard request-response chatbots. The true power of Generative AI on AWS is unlocked when coupled with event-driven serverless architectures.
Using Amazon Bedrock's Converse API with Nova Lite proved to be exceptionally cost-effective and lightning-fast for structured reasoning tasks. Designing for the "zero-touch" paradigm reinforced the value of clean JSON schemas, resilient error handling, and automated delivery pipelines.



Links & Repository

Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article