Sparse Dynamics LLC builds safety tooling for the agents, skills, and runtimes that wrap today's AI — so harness users, developers, and everyday operators can trust what runs on their machine.
A tool designed to make your AI harness safer — scan skills, surface risk, and keep untrusted code from reaching your agent runtime.
Guardian sits at the boundary between your harness and the outside world: CLI-first, inspectable, built for developers who ship agents daily.
The reference backend for Guardian CLI — policy, scanning, and enforcement wired together so Guardian can run checks locally or in your stack.
Pair Guardian on the command line with Trypanophobe as the server: same safety model, deployable where your harness already lives.
Kernel-level antivirus safety built specifically for agent harnesses — enforcement at ring 0, below user-mode proxies and CLI filters.
Paladin hooks syscalls and driver IOCTL paths so malicious skills, injected tools, and compromised harness processes are stopped before they reach the OS kernel boundary.