CapLite

Prompt-injection defense that derives write authorization from tool schemas and the user request before reading untrusted tool results.

Records citing CapLite

Nov 12, 2026
CapLite reports blocked benchmark injections with a substantial utility trade-off
DefensePaperThe University of Tokyo, Institute of Automation, Chinese Academy of Sciences

CapLite derives write permissions from tool schemas and values in the user request before reading untrusted tool results. The preprint’s authors report no successful tool-goal attacks in 13,935 AgentDojo episodes across 15 open-weight models, and 29.0% utility under attack on a broader task set. Its conditional write-integrity guarantee excludes output-only goals and does not establish read confidentiality or general agent security.