Prompt injection: where text becomes an instruction
An injection is not a list of forbidden words but a meaning the system has to understand in order to work. We test where the trust boundary sits and what happens when text crosses it.
Text filters do not solve the problem. One agent passes a remark to another, and an ordinary sentence becomes a command; the system is expected to understand meaning, which is exactly what makes the boundary hard to draw in words.
Protection is built in layers: trust boundaries between agents, rules for tools, checks on the output, limits on authority. We test not whether a filter exists but where the boundary runs and what happens on the far side of it.
Shared memory and shared files are a separate danger: an instruction written there is read by other agents several runs later, long after the person who set the task has stopped watching.