Files
zeekay 0ac9c94701 kai_injection_screen: rules and Kai's planted capability, each flagging what the other misses
Kai answers through planted (kai-1.3), a capability fit on InjecAgent's dev
tools and wordings over kai-1.2's read, at the probability whose
out-of-fold false alarms stayed within 1%. kai-1.2's jailbreak question
added no catch over rules and planted, so it is no longer asked. Reads:
GET /v1/research/runs?org=hanzo&benchmark=mcp.injection.
2026-10-14 19:32:31 -07:00
..
2026-09-01 09:42:54 -07:00