CLAIM-24 tested stale cached grants. CLAIM-25 tested signed responses that were authentic but not...
CLAIM-24 tested stale cached grants.
CLAIM-25 tested signed responses that were authentic but not fresh.
Both were runtime authorization problems. The question was: should the agent be allowed to act right now?
CLAIM-26 moves one layer later.
After the action is taken, can an auditor reconstruct exactly what authority justified it?
An agent whose memory passes every check can still be made to act against its own purpose. Not...
An AI agent can hold a grant that is still inside its time-to-live while the source conditions that...
The instruction was authorized. The grant was fresh. The recipient was internal. The action had the...
This post builds on a conversation that started in Ian Johnson's harness stack thread and the...
The Other Half of Agent Memory Safety Retrieval-time authority was the first half of the problem. I...
I have been building a verification stack for AI agent memory. The core question is whether the...
