Run agent-written code under a capability policy: see what a program can touch before it runs.
Free
Scanned where there is a package or repository to read, and every release diffed against the tool surface we already hold.
Run agent-written code under a capability policy: see what a program can touch before it runs.