$ whoami
Daniel Alfasi
AI Security Research Lead at Reco
I break (and sometimes build) agentic systems, coding agents, and LLM applications - then ship the detection logic that catches the same attack in production. Findings that stay in a slide deck aren't findings.
selected work
Research that ships
Inside Claude Fable 5: red-team findings
431 adversarial evaluations across agent archetypes - what held, what failed, and why a production lifetime measured in days still matters for agent security.
VulnScopper: hidden links between unseen security entities
Graph neural networks + LLMs for CVE enrichment when the NVD can't keep up - presented at the GNNet workshop at ACM CoNEXT 2024.
writing
Latest from the blog
focus
What I work on
Agentic red teaming at scale: prompt injection, tool misuse, memory and RAG poisoning, scenario generation, and turning research into production detections.