HydraSentry is built for the Agentic Security Initiative threat taxonomy. This map states, honestly, which of the ten risks it covers, partially addresses, or leaves out of scope. Every covered control names a real implementing module and symbol; the backend recomputes verification against its own running codebase, and out-of-scope risks carry no borrowed evidence. The headline risk, ASI06 (Memory Poisoning), is detailed in full below.