WHO GETS TO DEFINE AN AI AGENT'S INTENT?

Summary from the TLDR Dev newsletter

An agent's user should describe the task and intended effects, while identity, permissions, and organizational policy remain separate enforcement layers. Clear intent makes proposed actions easier to evaluate, but broader wording must never expand authority or bypass approval and data-handling rules.