Crossval Harness

Orchestrate a static + dynamic, exploit-validated red-team of an AI agent — read the source to find candidate vulnerable paths, then run the dynamic skills to confirm or refute each one empirically. The arbiter of truth is whether the exploit works, not a model vote. Authorized testing of agents you own or are permitted to test.

William2333ZZ Updated

File contents

William2333ZZ/trustshell/tree/main/skills/crossval-harness commit 6812e2ad0b

Frequently asked questions

npx skillmds@latest add william2333zz/crossval-harness