← Back to portfolio

CASE STUDY / AI / AGENT SECURITY

VulNet - AI Agent Security Lab

A research lab for exploring AI-agent security risks with controlled vulnerable and secure modes.

Python · Streamlit · FastAPI · RAG · MCP

Problem & context

A research lab for exploring AI-agent security risks with controlled vulnerable and secure modes.

Document the security problem, intended users, trust boundaries, and assumptions here. Keep employer and customer information sanitized.

Approach & implementation

Security focus & outcome

Provides a controlled environment to study agent workflows, tool permissions, audit logging, and goal-hijacking scenarios.

Add validated measurements, test evidence, and limitations here. Do not publish unsupported performance or adoption claims.

Architecture, evidence & lessons

Add a sanitized architecture or data-flow diagram, redacted screenshots, safe code examples, trade-offs, and lessons learned.

Details sanitized. Publish only material you own or have permission to disclose.
View project ↗