Hasan Demirkiran
Author, Kontext
Hasan Demirkiran writes about cyber risk classification, AI agent security, and evaluating guardrails for autonomous tool use.
Focus areas
- AI agent security and runtime authorization
- MCP authorization, OAuth, and credential management
- Least-privilege enforcement for autonomous systems
- Compliance and audit trails for agentic workflows
Selected writing
- Kestrel: a local classifier for the cyber risk of agent tool calls
Coding agents execute shell commands on your machine. We built a local classifier that screens every one of them in ~22 microseconds, and it beats seven frontier LLMs by ~0.40 F1 at a realistic base rate.