Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
SecureVibes: Multi-Agent Security Scanner
See a live scan of a four-agent security system finding vulnerabilities. Learn orchestration, communication, prompt engineering, and real cost/performance results.
I built SecureVibes, an open-source security scanner that uses four AI agents working together to find vulnerabilities in code. Unlike traditional scanners that pattern-match or single-agent approaches that get overwhelmed, SecureVibes mimics how human security teams actually work: understand the architecture, model threats, then validate them in code.
In this talk, I’ll start a live scan and walk you through what’s happening under the hood as it runs:
• How the Claude Agent SDK orchestrates multiple agents
• Why file-based communication between agents beats in-memory state
• Real prompt engineering lessons (what worked, what failed spectacularly)
• How Sonnet found 17 vulnerabilities at $3.44 while Opus found only 12 at $7.64
• Why traditional SAST tools (Semgrep, Bandit) found zero issues on the same codebase
By the time the demo completes, you’ll see the full workflow in action and understand the multi-agent architecture. You’ll leave with practical patterns for building your own multi-agent systems and an open-source tool you can try immediately.