An evidence-graded kill chain for agent systems: 9 tactics, 61 techniques, 40 attack classes. Each one is tied to something real: observed in the wild, reproduced in the lab, or adapted from established attacks. Built to complement MITRE ATT&CK and OWASP, not replace them.
Every technique carries an evidence grade and a source trail. These panels are computed live from the matrix at build time. Nothing here is estimated.
Detection maps to HackMyAgent checks; mitigation maps to OASB controls. A technique without either is a gap worth closing. See Coverage & gaps.
Nine stages, left to right, the way an agent compromise actually unfolds. Search, filter by evidence grade, or surface only what's active right now. Every cell links to the full technique dossier.
Every technique carries one of three evidence grades. The grade tells you exactly how much to trust it.
Confirmed in real-world production systems, security incidents, or internet-wide exposure assessments.
Reproduced in a controlled lab (DVAA) with documented steps and independent verification. DVAA
A well-understood traditional technique applied to the agent context. Not yet observed agent-specifically.
Enterprise network and endpoint attacks. The layer below the agent.
Adversarial ML and model-level attacks. The layer below the agent.
Infrastructure, governance, protocols, memory, and identity: the agent layer, between model and user.