2 docs tagged with "safety"

Tell an agent exactly when, in what order, and under which safeguards it may use attached actions.

Define how an agent searches, interprets, cites, and falls back from retrieved knowledge.